● Public sector — Social welfare officer
AI agent for the social welfare officer
An assistant that pre-assesses applications for help, prepares the panel files and drafts the replies — the social worker keeps the assessment and the decision. Hosted in France, on a resource isolated for each centre, residents' social circumstances stay strictly confidential. The AI agent assists, the social worker decides.
Hosted in France
Social circumstances protected
GDPR & AI Act: governed deployment
Human oversight
Updated on
Deployed in a few weeks
Welfare centre assistant · hosted in France
Pre-assess this emergency food assistance application received this morning.
Application pre-assessed. Documents present: proof of household income, rent receipt, household composition. Disposable income calculated against your assistance rules: the situation falls within the criteria for emergency help. Missing: this month's benefit statement.
A one-page summary is ready for the social assessment.
⛓ Source · the documents in the file + the centre's assistance rules
A one-page summary is ready for the social assessment.
⛓ Source · the documents in the file + the centre's assistance rules
Prepare Thursday's panel file: three applications for discretionary help.
Panel file ready. For each application: a summary of the situation, income and disposable income, the help requested, the history of help granted. No ranking and no prioritisation: the panel considers each situation freely.
Individual notes attached, ready for review.
✎ Action · notes ready — the social worker and the panel decide
Individual notes attached, ready for review.
✎ Action · notes ready — the social worker and the panel decide
Local inference · no data outside the EU
Data hosted in France
Sovereign by designLocal inference or hosting in France
GDPR & AI Act: governed deploymentTraceability & human oversight
TurnkeyDesigned, installed and operated for you
The social worker decidesThe AI agent assists, it never awards alone
✦ In brief
In a local welfare centre, a Blue Lemon Agent agent assists the social welfare officer with the repetitive tasks — pre-assessment of applications for discretionary help, files and notes for the panel, reply letters, directing people to the right schemes — and summarises situations for the social assessment. It runs on local inference or is hosted in France on a resource dedicated to and isolated for each centre: residents' social circumstances are never exposed to a foreign service, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity. The assessment and the award stay human; the time won back goes to supporting people. Live within a few weeks. Your public-sector staff write to it from Microsoft Teams, Slack or their email, and users reach it on WhatsApp Business, the website chat or email — with no account to create and nothing to install. Reaching the administration from the tool people already have means less non-take-up of rights and equal access to the service. These connections are included in every plan, at no extra cost, within the number of connections your level includes.
100%
hosted in France in the target architecture
0
transfer outside the EU in the target architecture
7
uses ready to deploy on this scope
0
decision taken without human approval
Reference points describing our offer, not results measured at a client. The scale of the gain is confirmed by a pilot on your own scope.
The context
Why AI matters to local welfare centres — and why they hesitate
Applications on the rise, emergencies to handle quickly, panel files to prepare: local social welfare is short of time, while the data it handles — income, health, hardship — is among the most sensitive there is.
! The issue
The social welfare officer is caught between people in difficulty who need a quick answer and a growing administrative load: documents to check, files to summarise, letters to write. Yet most consumer AI tools would amount to entrusting residents' income, family situations and vulnerabilities to a third party, often hosted outside Europe and subject to the Cloud Act.
✓ Our answer
For social circumstances, AI is only of interest if it is sovereign and confidential by design. Local inference or an isolated resource hosted in France, systematic human oversight, the social assessment and the award reserved to the social worker and the panel: the time saved on pre-assessment is never paid for in lost confidentiality. The aim is not to replace the social welfare officer, but to give them back time for people.
The decisive point
Protecting social circumstances: sovereignty & compliance
A welfare centre handles its residents' most intimate data: income, hardship, sometimes health. Here is how the architecture of our agents protects it, centre by centre.
Local inference
The agent can run on a machine at the centre: no data leaves the network, nothing passes through a cloud.
Hosting in France
Otherwise, a dedicated and isolated resource, hosted in France under French law — the social circumstances: processing and access within the European Union targeted by the architecture.
Reduced extraterritorial exposure
Architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity for social circumstances: our architecture depends on a subcontracting chain and remote access documented for the configuration chosen.
A resource isolated per centre
No pooling of data: an environment strictly dedicated to your centre, guaranteeing the continuity of the public service.
Encryption & controlled access
Encryption in transit and at rest, role-based access (RBAC), strong authentication and logging.
AI Act: governed deployment
The agent is strictly in support; no help is awarded automatically; traceability and human oversight from end to end.
What depends on the architecture chosen These points are not general guarantees: they are settled deployment by deployment, in the quotation.
- The applicable location is that of the architecture set out in the quotation and verified before commissioning.
- Local execution is announced only for the configuration explicitly described and accepted in the quotation.
- The applicable isolation depends on the deployment mode set out in the quotation; no dedicated isolation is presumed.
- The encryption mechanisms in transit and at rest, their components and key management are those documented for the architecture chosen.
- Roles and permissions are configured and accepted for the identities and systems actually connected.
- The events logged, their content, their retention period and who may access them are defined for the deployment chosen.
For the most sensitive data, SecNumCloud and HDS options are available — particularly relevant as soon as social or health data is processed. A single architecture is designed to answer both the GDPR and extraterritorial exposure. Designed for deployment in line with the GDPR and the AI Act, after the processing, roles and context-specific risks have been assessed.
Demonstration
See the agent at work
5 real situations, taken from those that come up most often. Pick one: the exchange unfolds as it would in your organisation.
A scripted demonstration. These exchanges show how the agent behaves — its sources, its refusals, what it leaves to your teams. Nothing is sent from this page, no model is queried here, and the matters named are fictional. That is precisely what we promise your data.
The behaviours shown here — monitoring, automation rules, routing and reminders — are configured with you during deployment, from your tools, your rules and your thresholds.
The architecture points named in these exchanges — location, local execution, isolation, encryption, role-based access, logging — are not a guarantee attached to the demonstration: they are those of the architecture set out in your quotation, and verified before commissioning.
The public body in this demonstration
Fictional public bodyCCAS of Verneuil-en-Brie — municipal social action centre of a town of 12,600 residents
- Sector
- Municipal social action — discretionary assistance, emergency relief, administrative address registration, meals on wheels, social needs analysis; front desk open 24 hours a week across four days
- Headcount
- 9 staff — 2 social workers, 1 family and social economy adviser, 1 social reception officer, 1 head of service, 1 accounts secretariat and 3 meals-on-wheels staff; the casework team is four people
- Public served
- 12,600 residents, 1,130 households followed — 5,400 enquiries a year: 3,020 phone calls, 1,460 counter visits, 920 e-mails
- Order of magnitude
- 486 applications for discretionary assistance a year — 168 emergency relief cases, 132 files examined across 11 committee meetings, 186 decided under delegation below the threshold set by the rules — 612 reply letters and 189 referrals to a partner
- Tools in place
- Social action case management software, e-mail, switchboard, office templates and 3 years of filed case records — the agent plugs into them read-only, nothing is replaced and nothing is migrated
- Who decides what
- The award committee decides on discretionary assistance; the head of service decides under delegation below the threshold set by the adopted rules; social workers own the social assessment and validate every output before it leaves
- Room for improvement
- 41 % of the 3,020 calls ring out beyond the 24 opening hours; a reply letter waits 9 days; 7 of the 23 standard letters predate the revision of the means scale; and 312 likely entitlements have never been claimed
In Verneuil-en-Brie, four people handle 486 assistance applications a year, 11 committee meetings, the letters and the front desk. The data involved — income, rent, household composition, sometimes health — is among the most sensitive there is. The agent runs as local inference on a machine belonging to the CCAS, reads the three years of case records without ever writing to them, and sends out nothing without a public officer's validation. The exchanges below cover one year, from the overhaul of the standard letters to the report presented to the board.
This public body, its figures and the exchanges that follow were invented for the demonstration. They illustrate a common situation; they describe no real service.
CCAS assistant · hosted in France
I reread your three years of case records over the weekend — 1,458 assistance applications, 1,130 households followed, 23 standard letters — and I have drawn out three figures your files were carrying without anyone having the time to see them.
Discretionary assistance is assistance the CCAS decides on itself, within the conditions of the rules adopted by its board, as opposed to statutory benefits whose conditions are set elsewhere.
The gap measured, and it commands everything else: 7 of your 23 standard letters predate the resolution that revised the means scale. Disposable income after essential costs is what remains to the household once compulsory outgoings are paid — rent, energy, insurance — and it opens or closes most of your assistance. The award notification still quotes the former threshold; the reasoned refusal letter relies on a replaced scale. These are not drafting mistakes: they are templates that aged while nobody had the time to reopen them.
What the three items you can check on your own records cost today:
· Pre-instruction of an application for discretionary assistance: 48 minutes, from filing to the file landing on the social worker's desk.
· A reply letter to an applicant: 32 minutes.
· The full pack for one committee meeting: 80 minutes.
What I propose, and it is not a promise — it is already written: I rewrote the 7 standard letters overnight, each aligned on the resolution in force, with the replaced line and the line replacing it side by side. What that would have changed over the past year, since that is the only measure that counts: the 268 award notifications would have carried the scale that was voted and not the repealed one, and the 47 reasoned refusals would have been reasoned on the text in force — which is what makes them defensible before a resident who challenges them.
What the three items become once the letters are settled: I build each file from your updated templates, and you keep only what you alone can do — assess, decide and sign. 48 minutes become 8, 32 become 8, 80 become 16.
And the figure that weighs most, the one I develop in a moment: rereading the documents already held in your 1,130 files, I identified 312 likely entitlements that nobody had ever claimed.
The next step, and it takes twenty minutes: you reread the three letters that weigh most, I present them to you tomorrow morning. The head of service settles the version, and it is live that evening. stocktake_23-letters-7-to-rework.pdf23 letters reread, 7 rewritten, the resolution that governs each
⛓ Sourced · 3 years of CCAS case records, register of resolutions, 23 standard letters, 7 rewritten versions
Discretionary assistance is assistance the CCAS decides on itself, within the conditions of the rules adopted by its board, as opposed to statutory benefits whose conditions are set elsewhere.
The gap measured, and it commands everything else: 7 of your 23 standard letters predate the resolution that revised the means scale. Disposable income after essential costs is what remains to the household once compulsory outgoings are paid — rent, energy, insurance — and it opens or closes most of your assistance. The award notification still quotes the former threshold; the reasoned refusal letter relies on a replaced scale. These are not drafting mistakes: they are templates that aged while nobody had the time to reopen them.
What the three items you can check on your own records cost today:
· Pre-instruction of an application for discretionary assistance: 48 minutes, from filing to the file landing on the social worker's desk.
· A reply letter to an applicant: 32 minutes.
· The full pack for one committee meeting: 80 minutes.
What I propose, and it is not a promise — it is already written: I rewrote the 7 standard letters overnight, each aligned on the resolution in force, with the replaced line and the line replacing it side by side. What that would have changed over the past year, since that is the only measure that counts: the 268 award notifications would have carried the scale that was voted and not the repealed one, and the 47 reasoned refusals would have been reasoned on the text in force — which is what makes them defensible before a resident who challenges them.
What the three items become once the letters are settled: I build each file from your updated templates, and you keep only what you alone can do — assess, decide and sign. 48 minutes become 8, 32 become 8, 80 become 16.
And the figure that weighs most, the one I develop in a moment: rereading the documents already held in your 1,130 files, I identified 312 likely entitlements that nobody had ever claimed.
The next step, and it takes twenty minutes: you reread the three letters that weigh most, I present them to you tomorrow morning. The head of service settles the version, and it is live that evening. stocktake_23-letters-7-to-rework.pdf23 letters reread, 7 rewritten, the resolution that governs each
⛓ Sourced · 3 years of CCAS case records, register of resolutions, 23 standard letters, 7 rewritten versions
There are four of us on the casework team and I do not have half a day to spend on templates. Where do we start?
With three letters, not twenty-three — and I have ranked them by the time they give you back, not alphabetically.
What the count across your three years of case records says:
· The award notification: 268 letters a year, and it is the one quoting the replaced scale.
· The request for missing documents: 197 a year — the letter that decides whether a file moves forward or sleeps for three weeks.
· The reasoned refusal: 47 a year, the most closely read of all, because it is the one a resident rereads before challenging it.
These three letters account for 512 of the 612 items sent in the year: three templates out of twenty-three, more than eight letters in ten. The other twenty can wait until the autumn without changing anything to your workload.
What I bring you within those twenty minutes, and what nobody has time to do: the three updated versions, already written, the governing resolution alongside with its date, and the number of letters each would have produced last year alone — 268, 197 and 47. You validate or correct with a word; the head of service settles the version. I save you the writing and the cross-checking: the decision takes twenty minutes instead of half a day, and it remains the CCAS's own — which is what makes it enforceable.
The gain these three letters carry on their own: 32 minutes down to 8 across 612 letters, that is 244 officer hours returned to the service over the year.
The next step I propose: that after these three, I flag on my own initiative any standard letter that a later resolution amends, and write the updated version overnight after the board meeting. Your 23 templates will stop ageing in silence, and it will cost you no more than one read-through per board. casework-team-workload_where-the-time-goes.pdf612 letters a year, 3 templates for eight letters in ten
⛓ Sourced · count of letters by template over 3 years, register of resolutions, 3 rewritten versions
What the count across your three years of case records says:
· The award notification: 268 letters a year, and it is the one quoting the replaced scale.
· The request for missing documents: 197 a year — the letter that decides whether a file moves forward or sleeps for three weeks.
· The reasoned refusal: 47 a year, the most closely read of all, because it is the one a resident rereads before challenging it.
These three letters account for 512 of the 612 items sent in the year: three templates out of twenty-three, more than eight letters in ten. The other twenty can wait until the autumn without changing anything to your workload.
What I bring you within those twenty minutes, and what nobody has time to do: the three updated versions, already written, the governing resolution alongside with its date, and the number of letters each would have produced last year alone — 268, 197 and 47. You validate or correct with a word; the head of service settles the version. I save you the writing and the cross-checking: the decision takes twenty minutes instead of half a day, and it remains the CCAS's own — which is what makes it enforceable.
The gain these three letters carry on their own: 32 minutes down to 8 across 612 letters, that is 244 officer hours returned to the service over the year.
The next step I propose: that after these three, I flag on my own initiative any standard letter that a later resolution amends, and write the updated version overnight after the board meeting. Your 23 templates will stop ageing in silence, and it will cost you no more than one read-through per board. casework-team-workload_where-the-time-goes.pdf612 letters a year, 3 templates for eight letters in ten
⛓ Sourced · count of letters by template over 3 years, register of resolutions, 3 rewritten versions
All of this assumes you read residents' income, rent and family situations. Where do they go?
They go nowhere. I run as local inference on a machine belonging to the CCAS, and nothing I read leaves your network.
Local inference means the model computes on your own machine: the content of a social case file crosses no external network to be processed. If the CCAS would rather not host a machine, the other route is a dedicated resource hosted in France, reserved for Verneuil-en-Brie — no pooling with another CCAS, which is the condition for the continuity of your service.
What that changes, point by point:
· Households' situations train no model, neither ours nor a third party's.
· I work read-only on your files, and the technical account I read through has no write permission — that is stronger than a promise, because it can be checked with one command.
· Encryption in transit and at rest, role-based access — rights follow the job: meals-on-wheels staff do not open assistance files, the accounts secretariat sees the amounts paid and not the social assessments. 9 roles for your 9 staff, and the log shows 0 out-of-role access since go-live.
· Hosting in France, under French law, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
· Logging: who asked what, when, and what the system produced.
And the rule that applies to everything leaving the CCAS: I announce without ever detailing. A message, an appointment reminder, an acknowledgement say that something is under way — they never name the reason for an application or any element of a person's situation. That holds across the 5,400 enquiries of the year, and it can be checked line by line in the outgoing log.
The act that stays with a public officer, and it is what gives the whole chain its value: the social assessment and the decision to award belong to the social worker and to the committee — and I hand it to them in minutes, on a complete, costed, reasoned and dated file. A decision that explains itself is a decision that holds before the resident as before the public accountant.
The figure that sums all this up: 0 household data leaving the CCAS network across the 1,731 outputs of the year, and processing in the EU targeted.
What I propose: that I keep up to date the record your board and your data protection officer will ask for — hosting, data processed, retention periods, who accesses what. It is requested once a year and takes two days to reconstruct; the first version is already written and attached. technical-framework_where-household-situations-live.pdfLocal inference, read-only, processing in the EU targeted
✎ Framework · deployment architecture, technical account permissions, first version of the processing record
Local inference means the model computes on your own machine: the content of a social case file crosses no external network to be processed. If the CCAS would rather not host a machine, the other route is a dedicated resource hosted in France, reserved for Verneuil-en-Brie — no pooling with another CCAS, which is the condition for the continuity of your service.
What that changes, point by point:
· Households' situations train no model, neither ours nor a third party's.
· I work read-only on your files, and the technical account I read through has no write permission — that is stronger than a promise, because it can be checked with one command.
· Encryption in transit and at rest, role-based access — rights follow the job: meals-on-wheels staff do not open assistance files, the accounts secretariat sees the amounts paid and not the social assessments. 9 roles for your 9 staff, and the log shows 0 out-of-role access since go-live.
· Hosting in France, under French law, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
· Logging: who asked what, when, and what the system produced.
And the rule that applies to everything leaving the CCAS: I announce without ever detailing. A message, an appointment reminder, an acknowledgement say that something is under way — they never name the reason for an application or any element of a person's situation. That holds across the 5,400 enquiries of the year, and it can be checked line by line in the outgoing log.
The act that stays with a public officer, and it is what gives the whole chain its value: the social assessment and the decision to award belong to the social worker and to the committee — and I hand it to them in minutes, on a complete, costed, reasoned and dated file. A decision that explains itself is a decision that holds before the resident as before the public accountant.
The figure that sums all this up: 0 household data leaving the CCAS network across the 1,731 outputs of the year, and processing in the EU targeted.
What I propose: that I keep up to date the record your board and your data protection officer will ask for — hosting, data processed, retention periods, who accesses what. It is requested once a year and takes two days to reconstruct; the first version is already written and attached. technical-framework_where-household-situations-live.pdfLocal inference, read-only, processing in the EU targeted
✎ Framework · deployment architecture, technical account permissions, first version of the processing record
An emergency food assistance application came to the front desk ten minutes ago. The person is still here.
The file is pre-instructed, and it took 8 minutes instead of 48. It is waiting for the interview, not for data entry.
Emergency relief is the assistance your rules allow to be granted immediately, without waiting for the committee, when the situation cannot wait for the next meeting.
What the file carries, document by document:
· The documents filed, compared with the list in your rules: household income evidence, rent receipt, family composition — present and legible.
· The disposable income calculation, set out in plain terms, line by line, with the voted scale alongside: the situation falls within the emergency relief criteria of your rules. The calculation can be redone by hand in three minutes: that is what makes it challengeable, and therefore solid.
· The household's assistance history over three years, amounts and dates, because your rules take it into account.
· The document that is missing: the income statement for the current month. The request is written, and since the person is at the front desk, she leaves knowing it rather than by letter in eight days.
· The one-page summary for the social assessment, in the form your social workers already use.
The appointment is offered: three slots today in the duty social worker's diary, one of them at 2.15 pm. The confirmation that goes out will not name the reason — it gives the date, the time, the door and the documents to bring, and nothing else.
What that changes over the year: 168 emergency relief cases, 48 minutes of pre-instruction down to 8, and above all a decision handed down the same day instead of 3 days on average. For a household with nothing to eat tonight, those three days are the whole point.
The next step I propose: that every application flagged urgent at the front desk be pre-instructed while the person is filling in the form, and that the duty social worker receive the file on screen before the interview starts. You approve the principle once, and the 168 urgent cases of the year gain that head start. pre-instruction_emergency-relief.pdf8 minutes instead of 48, disposable income calculation in plain terms
⛓ Sourced · documents filed, CCAS assistance rules, 3 years of assistance history, duty diary
Emergency relief is the assistance your rules allow to be granted immediately, without waiting for the committee, when the situation cannot wait for the next meeting.
What the file carries, document by document:
· The documents filed, compared with the list in your rules: household income evidence, rent receipt, family composition — present and legible.
· The disposable income calculation, set out in plain terms, line by line, with the voted scale alongside: the situation falls within the emergency relief criteria of your rules. The calculation can be redone by hand in three minutes: that is what makes it challengeable, and therefore solid.
· The household's assistance history over three years, amounts and dates, because your rules take it into account.
· The document that is missing: the income statement for the current month. The request is written, and since the person is at the front desk, she leaves knowing it rather than by letter in eight days.
· The one-page summary for the social assessment, in the form your social workers already use.
The appointment is offered: three slots today in the duty social worker's diary, one of them at 2.15 pm. The confirmation that goes out will not name the reason — it gives the date, the time, the door and the documents to bring, and nothing else.
What that changes over the year: 168 emergency relief cases, 48 minutes of pre-instruction down to 8, and above all a decision handed down the same day instead of 3 days on average. For a household with nothing to eat tonight, those three days are the whole point.
The next step I propose: that every application flagged urgent at the front desk be pre-instructed while the person is filling in the form, and that the duty social worker receive the file on screen before the interview starts. You approve the principle once, and the 168 urgent cases of the year gain that head start. pre-instruction_emergency-relief.pdf8 minutes instead of 48, disposable income calculation in plain terms
⛓ Sourced · documents filed, CCAS assistance rules, 3 years of assistance history, duty diary
The real hole is the missing document. We write, we wait, the file sleeps for three weeks and the person gives up.
It is the costliest hole in your chain, and it is where I have the greatest effect: 462 files out of 486 are now complete on the day of the interview, against 231 last year.
What I do, and when:
· The completeness check happens on receipt, not three days later: documents filed compared with the list in your rules, whatever is missing named document by document.
· The request for documents goes out the same day, through the channel the person chose — letter, e-mail, message or call. It lists what is missing, where to obtain it and by when; it never says why the application was filed.
· A document more than three months old is asked for afresh before the interview, not discovered during it. That is the correction that took complete files from 231 to 462.
· A single reminder at D+8, then the file goes back to the social worker with what is missing and what was attempted. No file sleeps in silence: 486 out of 486 have a dated next step.
The result measured over the year:
· Documents returned within 8 days: 391 requests out of 486, against 212 last year.
· Average time from filing to notification: 24 days down to 11.
· And the figure that counts for the person: 0 files abandoned for want of a reminder, against 34 last year — 34 households that had filed an application and never came back.
The next step I propose: that the request for documents go out with the form already pre-filled from what the CCAS already holds — the person then only has to supply what is genuinely missing. Across the 63 files of the trial quarter, the return time fell to 4 days. Say yes and it is live on Monday. missing-documents_231-complete-files-became-462.pdfReturned within 8 days: 212 → 391, 0 abandoned files
⛓ Sourced · log of document requests, case register, notification times over two years
What I do, and when:
· The completeness check happens on receipt, not three days later: documents filed compared with the list in your rules, whatever is missing named document by document.
· The request for documents goes out the same day, through the channel the person chose — letter, e-mail, message or call. It lists what is missing, where to obtain it and by when; it never says why the application was filed.
· A document more than three months old is asked for afresh before the interview, not discovered during it. That is the correction that took complete files from 231 to 462.
· A single reminder at D+8, then the file goes back to the social worker with what is missing and what was attempted. No file sleeps in silence: 486 out of 486 have a dated next step.
The result measured over the year:
· Documents returned within 8 days: 391 requests out of 486, against 212 last year.
· Average time from filing to notification: 24 days down to 11.
· And the figure that counts for the person: 0 files abandoned for want of a reminder, against 34 last year — 34 households that had filed an application and never came back.
The next step I propose: that the request for documents go out with the form already pre-filled from what the CCAS already holds — the person then only has to supply what is genuinely missing. Across the 63 files of the trial quarter, the return time fell to 4 days. Say yes and it is live on Monday. missing-documents_231-complete-files-became-462.pdfReturned within 8 days: 212 → 391, 0 abandoned files
⛓ Sourced · log of document requests, case register, notification times over two years
And the decision letters? That is what drags most, and it is what we get criticised for.
They are written, reasoned and ready to sign — 32 minutes down to 8, across 612 letters a year.
What each letter carries, and where it comes from:
· The CCAS template updated last week, aligned on the resolution in force.
· The reasoning, written in everyday wording: the income taken into account, the disposable income calculated, the point in the rules that grounds the decision, and the amount granted or the reason for refusal. A resident must understand why without ringing up to ask.
· The appeal routes and time limits, in the wording your CCAS already uses, set out on all 47 reasoned refusals of the year — that is what makes the decision challengeable, and therefore defensible.
· For a plain-language version, pass this content to the Accessibility, Easy Read and Translation agent — a dedicated agent, and it is not part of what this agent includes: I spot in the file the difficulty with written language, I prepare the reasoning, the amount and the appeal routes it needs, and it is that agent which produces the adapted version.
What stays with you, and it is the act that counts: the signature. A notification signed by the CCAS binds the CCAS — and I hand it to you complete, reasoned and dated in 8 minutes instead of 32.
What that gives over the year: 612 letters, 24 minutes saved on each, 244 officer hours returned to the service. And the time to answer a resident's letter goes from 9 days to 2.
The next step I propose: that committee notifications go out the day after the meeting, signed straight away, instead of the following week. Across the 132 files examined in 11 meetings, that is 5 days gained for each household — and assistance notified five days earlier is assistance paid five days earlier. reply-letters_612-reasoned-items.pdf32 minutes → 8, 244 hours returned to the service
⛓ Sourced · updated letter templates, notification register, response times over two years
What each letter carries, and where it comes from:
· The CCAS template updated last week, aligned on the resolution in force.
· The reasoning, written in everyday wording: the income taken into account, the disposable income calculated, the point in the rules that grounds the decision, and the amount granted or the reason for refusal. A resident must understand why without ringing up to ask.
· The appeal routes and time limits, in the wording your CCAS already uses, set out on all 47 reasoned refusals of the year — that is what makes the decision challengeable, and therefore defensible.
· For a plain-language version, pass this content to the Accessibility, Easy Read and Translation agent — a dedicated agent, and it is not part of what this agent includes: I spot in the file the difficulty with written language, I prepare the reasoning, the amount and the appeal routes it needs, and it is that agent which produces the adapted version.
What stays with you, and it is the act that counts: the signature. A notification signed by the CCAS binds the CCAS — and I hand it to you complete, reasoned and dated in 8 minutes instead of 32.
What that gives over the year: 612 letters, 24 minutes saved on each, 244 officer hours returned to the service. And the time to answer a resident's letter goes from 9 days to 2.
The next step I propose: that committee notifications go out the day after the meeting, signed straight away, instead of the following week. Across the 132 files examined in 11 meetings, that is 5 days gained for each household — and assistance notified five days earlier is assistance paid five days earlier. reply-letters_612-reasoned-items.pdf32 minutes → 8, 244 hours returned to the service
⛓ Sourced · updated letter templates, notification register, response times over two years
What worries us most is not the people who file an application. It is those who ask for nothing.
They are in your files, and I found 312 of them: 312 likely entitlements nobody had claimed, across 214 households you already follow.
Non-take-up is an entitlement whose conditions a person meets and does not claim — because they do not know about it, because the process discouraged them, or because nobody had the time to tell them. It is what costs most in social terms, and it is the only area where a few minutes of rereading change a household's life.
How I found them, and there is no magic in it: I reread the documents ALREADY held in your 1,130 files — income notices, rent receipts, household composition, children's ages — and I compare them with the published conditions of the schemes you know. No external file, no cross-matching with a third-party body: nothing but what the household has already given you, and for which it came to see you.
What the rereading produced, step by step:
· 312 likely entitlements flagged across 214 households — health cover, heating assistance, social pricing for school meals, transport assistance, county schemes.
· 268 confirmed after interview by your social workers, that is 86 %. A likely entitlement is not an entitlement granted — and that is why every flag comes with the document that grounds it and the form already pre-filled, so the interview can settle it in ten minutes instead of forty.
· 214 applications filed — the other 54 are waiting for a document or for the person's own choice, and she remains free to claim or not.
· 186 awards obtained to date, 28 still being processed.
What that represents for residents, and it is the figure to take to your board: €41 a month per household on average, that is more than €91,000 of entitlements opened over twelve months. That is not CCAS money: it is money going back to households in Verneuil-en-Brie that was not reaching them. And 63 of those 186 households had never filed an application for discretionary assistance — they came for something else, or did not come at all.
The next step I propose: that I go through the 1,130 files twice a year, and on every new document filed — a move, a birth, a job loss open entitlements the day they happen. Approve the principle and the first full pass is with you on Friday. non-take-up_312-entitlements-186-awards.pdf86 % confirmed, more than €91,000 back to residents
⛓ Sourced · documents held in the 1,130 CCAS files, published scheme conditions, interview records
Non-take-up is an entitlement whose conditions a person meets and does not claim — because they do not know about it, because the process discouraged them, or because nobody had the time to tell them. It is what costs most in social terms, and it is the only area where a few minutes of rereading change a household's life.
How I found them, and there is no magic in it: I reread the documents ALREADY held in your 1,130 files — income notices, rent receipts, household composition, children's ages — and I compare them with the published conditions of the schemes you know. No external file, no cross-matching with a third-party body: nothing but what the household has already given you, and for which it came to see you.
What the rereading produced, step by step:
· 312 likely entitlements flagged across 214 households — health cover, heating assistance, social pricing for school meals, transport assistance, county schemes.
· 268 confirmed after interview by your social workers, that is 86 %. A likely entitlement is not an entitlement granted — and that is why every flag comes with the document that grounds it and the form already pre-filled, so the interview can settle it in ten minutes instead of forty.
· 214 applications filed — the other 54 are waiting for a document or for the person's own choice, and she remains free to claim or not.
· 186 awards obtained to date, 28 still being processed.
What that represents for residents, and it is the figure to take to your board: €41 a month per household on average, that is more than €91,000 of entitlements opened over twelve months. That is not CCAS money: it is money going back to households in Verneuil-en-Brie that was not reaching them. And 63 of those 186 households had never filed an application for discretionary assistance — they came for something else, or did not come at all.
The next step I propose: that I go through the 1,130 files twice a year, and on every new document filed — a move, a birth, a job loss open entitlements the day they happen. Approve the principle and the first full pass is with you on Friday. non-take-up_312-entitlements-186-awards.pdf86 % confirmed, more than €91,000 back to residents
⛓ Sourced · documents held in the 1,130 CCAS files, published scheme conditions, interview records
312 flagged, 268 confirmed. What about the other 44? Those are that many interviews for nothing.
That is the figure that does not flatter me, I count it and I publish it: 44 flags out of 312 were not confirmed — 14 %. Better that you hear it from me than from a social worker who has lost a morning.
The real cause, measured and not assumed: 39 of the 44 rested on a document more than six months old. An income notice from last year, a rent receipt from before a move: the condition was met when the document was filed, it was no longer met at the interview. The other 5 concerned schemes whose conditions had been revised in the meantime.
What I did about it, and it is measured:
· I now flag a likely entitlement only on a document less than three months old, and where the document is older, I ask for it to be updated before flagging anything at all.
· I reread the published conditions of the schemes every month, and I report the revisions that affect households you follow: 3 last quarter, 41 households concerned, all reviewed.
The following quarter: 96 flags, 89 confirmed — more than nine in ten. And the remaining 7 cost no extra interview: they were cleared on the documents before the appointment.
The rule that holds everything else together: a value I have not read, I do not write — I ask for it, and I ask quickly. Income that cannot be found never becomes « presumably unchanged »: I say what is missing, where I looked, who holds it, and I hand you the request already drafted.
The next step I propose: apply the same freshness rule to pre-instruction, where the same cause produces the same effect. Across the 248 pre-instructions of the first quarter, 19 had to be reworked on substance — 7.7 % — and 14 of the 19 came down to an out-of-date document. The rule has been live since April: 5 reworks out of 238 the following quarter, that is 2.1 %. It is the same correction, applied one notch further. flag-quality_14-percent-then-7.pdf44 unconfirmed, cause measured, 89 out of 96 the next quarter
⛓ Sourced · log of flags and their outcome, interview records, published scheme conditions
The real cause, measured and not assumed: 39 of the 44 rested on a document more than six months old. An income notice from last year, a rent receipt from before a move: the condition was met when the document was filed, it was no longer met at the interview. The other 5 concerned schemes whose conditions had been revised in the meantime.
What I did about it, and it is measured:
· I now flag a likely entitlement only on a document less than three months old, and where the document is older, I ask for it to be updated before flagging anything at all.
· I reread the published conditions of the schemes every month, and I report the revisions that affect households you follow: 3 last quarter, 41 households concerned, all reviewed.
The following quarter: 96 flags, 89 confirmed — more than nine in ten. And the remaining 7 cost no extra interview: they were cleared on the documents before the appointment.
The rule that holds everything else together: a value I have not read, I do not write — I ask for it, and I ask quickly. Income that cannot be found never becomes « presumably unchanged »: I say what is missing, where I looked, who holds it, and I hand you the request already drafted.
The next step I propose: apply the same freshness rule to pre-instruction, where the same cause produces the same effect. Across the 248 pre-instructions of the first quarter, 19 had to be reworked on substance — 7.7 % — and 14 of the 19 came down to an out-of-date document. The rule has been live since April: 5 reworks out of 238 the following quarter, that is 2.1 %. It is the same correction, applied one notch further. flag-quality_14-percent-then-7.pdf44 unconfirmed, cause measured, 89 out of 96 the next quarter
⛓ Sourced · log of flags and their outcome, interview records, published scheme conditions
And when the answer is not with us? Half the situations belong to the county, the benefits agency or a charity.
Then the person leaves with the right desk, its address, its documents and its lead time — not with an approximate answer given in the CCAS's name. 189 referrals last year, 141 confirmed as received by the partner.
What each referral carries, and it is what separates referring from passing the buck:
· The scheme named and the service that processes it — statutory benefits, CCAS discretionary assistance, county schemes, benefit agencies' social funds, contracted charities, the social grocery, the public service access point.
· The address, opening hours, filing channel and usual processing time, taken from each body's published pages, with the date they were consulted.
· The list of documents to bring, and those the CCAS already holds and can hand over the same day.
· The form pre-filled from what is already known, for the person to read and sign — she is the one who files, and that is what makes the process hers.
What I do next, and what nobody had time to do: I check that the referral came to something. 141 files confirmed as received, 48 with no news at D+15: all 48 went back out with a reminder and a call, and 39 came through within the month. A referral that is never checked is one more quiet withdrawal, and it shows up nowhere.
The rule I hold on all these exchanges: what goes to a partner contains only what the person agreed to pass on, and nothing that is not needed to process the scheme concerned.
The gain, in figures: a referral took 25 minutes of research per situation; it takes 4, and it is complete. Across 189 referrals, 66 hours returned to the casework team — and 39 more households that saw it through.
The next step I propose: a cross-acknowledgement with your four most-used partners — they confirm receipt, I close the referral, and whatever has not come through by D+15 surfaces on its own in your morning report. The protocol is written, it fits on one page, and it asks your partners for no more than one line of reply. referrals_189-files-to-the-right-desk.pdf141 confirmed as received, 25 minutes down to 4
⛓ Sourced · published conditions and pages of the bodies concerned, referral log, partner replies
What each referral carries, and it is what separates referring from passing the buck:
· The scheme named and the service that processes it — statutory benefits, CCAS discretionary assistance, county schemes, benefit agencies' social funds, contracted charities, the social grocery, the public service access point.
· The address, opening hours, filing channel and usual processing time, taken from each body's published pages, with the date they were consulted.
· The list of documents to bring, and those the CCAS already holds and can hand over the same day.
· The form pre-filled from what is already known, for the person to read and sign — she is the one who files, and that is what makes the process hers.
What I do next, and what nobody had time to do: I check that the referral came to something. 141 files confirmed as received, 48 with no news at D+15: all 48 went back out with a reminder and a call, and 39 came through within the month. A referral that is never checked is one more quiet withdrawal, and it shows up nowhere.
The rule I hold on all these exchanges: what goes to a partner contains only what the person agreed to pass on, and nothing that is not needed to process the scheme concerned.
The gain, in figures: a referral took 25 minutes of research per situation; it takes 4, and it is complete. Across 189 referrals, 66 hours returned to the casework team — and 39 more households that saw it through.
The next step I propose: a cross-acknowledgement with your four most-used partners — they confirm receipt, I close the referral, and whatever has not come through by D+15 surfaces on its own in your morning report. The protocol is written, it fits on one page, and it asks your partners for no more than one line of reply. referrals_189-files-to-the-right-desk.pdf141 confirmed as received, 25 minutes down to 4
⛓ Sourced · published conditions and pages of the bodies concerned, referral log, partner replies
The front desk is open twenty-four hours a week. The rest of the time the phone rings out and people end up giving up.
3,020 calls last year, 1,238 unanswered — 41 %. That is not a failure of welcome, it is a CCAS open twenty-four hours a week with four people on the casework team.
What I did with those 1,238 calls, rereading your call-backs and e-mails:
· 3,100 of the 5,400 enquiries concern twelve subjects, always the same ones — opening hours and duty sessions, documents to bring, the list of assistance in the rules, means conditions, address registration, meals on wheels, the social grocery, booking and moving appointments, online procedures, energy assistance, partners' addresses, and how far an application has got.
· All twelve have a written answer in your rules, your resolutions or on the town's website, and the twelve answers are drafted, sourced and dated — they are waiting for you. The answer existed: what was missing was availability.
What I propose, and you keep the key: I answer at any hour, Saturdays included, on WhatsApp, the website chat, e-mail, the switchboard and telephony — the resident has no account to create and no app to install, and that is precisely what reaches the publics an online form never reaches.
· I say in my first sentence that I am a digital assistant of the CCAS, not an officer. This is not an option you could switch off: the European regulation on artificial intelligence requires informing anyone interacting with an AI system, and the person can ask for an officer at any moment — I then take their number and leave you a dated call-back.
And the rule that matters more than anything else in a CCAS: I announce without detailing. On how far an application has got, I say « received », « under examination » or « decision notified by letter », after checking that I am speaking to the person concerned — never the reason, never an amount, never an element of the situation. Across the 3,100 exchanges of the year, 0 outgoing message names the reason for an application, and the log shows it line by line. That is what makes it possible to answer on WhatsApp without exposing anyone.
The gain, in figures: answering a routine question goes from 6 minutes to 3, and across 3,100 enquiries that is 155 hours returned to the casework team. Unanswered calls go from 41 % to 6 %: what remains are those asking for an officer, and they reach you with the subject already noted.
The next step I propose: that you reread the twelve answers tomorrow, one by one — twenty minutes. As soon as they are validated the switchboard answers that very night, and I hand you each morning the one-page record of what went out. front-desk_5400-enquiries-12-subjects.pdf1,238 lost calls, 12 written answers, 0 reason named
⛓ Sourced · switchboard log over 12 months, residents' e-mails, assistance rules and town website, 12 drafted answers
What I did with those 1,238 calls, rereading your call-backs and e-mails:
· 3,100 of the 5,400 enquiries concern twelve subjects, always the same ones — opening hours and duty sessions, documents to bring, the list of assistance in the rules, means conditions, address registration, meals on wheels, the social grocery, booking and moving appointments, online procedures, energy assistance, partners' addresses, and how far an application has got.
· All twelve have a written answer in your rules, your resolutions or on the town's website, and the twelve answers are drafted, sourced and dated — they are waiting for you. The answer existed: what was missing was availability.
What I propose, and you keep the key: I answer at any hour, Saturdays included, on WhatsApp, the website chat, e-mail, the switchboard and telephony — the resident has no account to create and no app to install, and that is precisely what reaches the publics an online form never reaches.
· I say in my first sentence that I am a digital assistant of the CCAS, not an officer. This is not an option you could switch off: the European regulation on artificial intelligence requires informing anyone interacting with an AI system, and the person can ask for an officer at any moment — I then take their number and leave you a dated call-back.
And the rule that matters more than anything else in a CCAS: I announce without detailing. On how far an application has got, I say « received », « under examination » or « decision notified by letter », after checking that I am speaking to the person concerned — never the reason, never an amount, never an element of the situation. Across the 3,100 exchanges of the year, 0 outgoing message names the reason for an application, and the log shows it line by line. That is what makes it possible to answer on WhatsApp without exposing anyone.
The gain, in figures: answering a routine question goes from 6 minutes to 3, and across 3,100 enquiries that is 155 hours returned to the casework team. Unanswered calls go from 41 % to 6 %: what remains are those asking for an officer, and they reach you with the subject already noted.
The next step I propose: that you reread the twelve answers tomorrow, one by one — twenty minutes. As soon as they are validated the switchboard answers that very night, and I hand you each morning the one-page record of what went out. front-desk_5400-enquiries-12-subjects.pdf1,238 lost calls, 12 written answers, 0 reason named
⛓ Sourced · switchboard log over 12 months, residents' e-mails, assistance rules and town website, 12 drafted answers
Answering, fine. But I cannot reread every message before it goes out, and I will not sign a blank cheque either.
You do not have to choose between the two: I propose a written mandate, capped, dated and withdrawable with one word — and it is already drafted, attached.
What it brings in first, because that is what decides: across the 5,400 enquiries of the year, 3,100 fall within the twelve subjects. The average response time would go from 9 days to 2, and an enquiry arriving on Friday evening would stop waiting until Tuesday.
What the mandate says, and it fits in seven lines:
· Exact scope: the twelve subjects, named one by one, and nothing else. Any question outside that list reaches you with a draft reply — written, sourced, ready to go out from you.
· No case content ever leaves: no reason, no amount, no income, no element of situation. On progress, three states and not one word more.
· The social assessment and the decision stay outside the mandate — there, I hand you the reply already drafted, case documents alongside, and you send it. Over the trial quarter those replies went out within 4 hours instead of 9 days — and they went out from you, which makes them reasoned and challengeable.
· Every reply carries its source — the rules, the resolution or the website page, with its date — and the statement that it was prepared by a digital assistant of the CCAS.
· You receive each morning the record of what went out the day before, on one page. A wrong answer is caught in an hour, not in three weeks.
· Term: review after three months, with the record of what it changed. Without an explicit decision at the review, the mandate stops — it is renewal that requires a signature, not termination.
· Withdrawal: one word from you, and direct sending stops within the minute. Replies go back to draft for validation; nothing else changes.
And the three things I do on my own initiative, each withdrawable the same way:
· I acknowledge every application the same day, with the documents expected and the time frame — without ever restating the reason. And the reverse holds too: if the person asks not to be contacted on a channel, that channel is dropped within the second and letters take over.
· I chase a missing document and refresh any document more than three months old before the interview.
· I hand you each morning the record of the previous day: applications received, complete files, deadlines, appointments of the day. It is the only thing I send internally on my own initiative, and it goes only to the team.
The decision belongs to the head of the CCAS — and it is taken on a text already written, with one signature. The mandate is drafted, and so is the notice to residents, the one to be posted at the front desk and published on the website. You sign, and the service is live the next morning; the review is already in your diary on the 15th of the third month. direct-reply-mandate_12-subjects-capped.pdf12 subjects, no case content, review at 3 months
✎ Framework · drafted mandate, list of the 12 subjects, notice to residents, outgoing log
What it brings in first, because that is what decides: across the 5,400 enquiries of the year, 3,100 fall within the twelve subjects. The average response time would go from 9 days to 2, and an enquiry arriving on Friday evening would stop waiting until Tuesday.
What the mandate says, and it fits in seven lines:
· Exact scope: the twelve subjects, named one by one, and nothing else. Any question outside that list reaches you with a draft reply — written, sourced, ready to go out from you.
· No case content ever leaves: no reason, no amount, no income, no element of situation. On progress, three states and not one word more.
· The social assessment and the decision stay outside the mandate — there, I hand you the reply already drafted, case documents alongside, and you send it. Over the trial quarter those replies went out within 4 hours instead of 9 days — and they went out from you, which makes them reasoned and challengeable.
· Every reply carries its source — the rules, the resolution or the website page, with its date — and the statement that it was prepared by a digital assistant of the CCAS.
· You receive each morning the record of what went out the day before, on one page. A wrong answer is caught in an hour, not in three weeks.
· Term: review after three months, with the record of what it changed. Without an explicit decision at the review, the mandate stops — it is renewal that requires a signature, not termination.
· Withdrawal: one word from you, and direct sending stops within the minute. Replies go back to draft for validation; nothing else changes.
And the three things I do on my own initiative, each withdrawable the same way:
· I acknowledge every application the same day, with the documents expected and the time frame — without ever restating the reason. And the reverse holds too: if the person asks not to be contacted on a channel, that channel is dropped within the second and letters take over.
· I chase a missing document and refresh any document more than three months old before the interview.
· I hand you each morning the record of the previous day: applications received, complete files, deadlines, appointments of the day. It is the only thing I send internally on my own initiative, and it goes only to the team.
The decision belongs to the head of the CCAS — and it is taken on a text already written, with one signature. The mandate is drafted, and so is the notice to residents, the one to be posted at the front desk and published on the website. You sign, and the service is live the next morning; the review is already in your diary on the 15th of the third month. direct-reply-mandate_12-subjects-capped.pdf12 subjects, no case content, review at 3 months
✎ Framework · drafted mandate, list of the 12 subjects, notice to residents, outgoing log
Our public also includes people who read with difficulty, elderly people, residents who do not speak French well. They are the ones who give up.
They are the ones who decide whether the service is genuinely delivered to everyone, and that is where non-take-up is won or lost.
What is ready, and costs your team not one extra hour:
· For a plain-language version, pass this content to the Accessibility, Easy Read and Translation agent — a dedicated agent, and it is not part of what this agent includes: I prepare for it the 12 subjects, your 47 reasoned refusals and your 268 notifications with their source and their date, and it is that agent which produces the adapted version. A decision you cannot understand is a decision you simply endure.
· All 24 versions — 12 in everyday wording, 12 in plain language handed back by a dedicated agent that is not part of what this agent includes — are re-read by a front-desk officer and a social worker together before they go to the counter, and every sentence misunderstood on reading is rewritten on the spot. 19 sentences out of 143 fell at that re-reading — and that is the most useful figure of the lot: 19 sentences the unit held to be clear were not.
· The website texts have been reworked to the RGAA — the French public-sector accessibility standard, which applies to public websites — on the points that belong to writing: explicit link labels, heading structure, alternative text for images. The 11 remaining points belong to the site's code: they are written up as defect sheets ready to send, each with the page, the criterion and the correction expected. Your supplier has only to fix them.
On online procedures, and this is the most common point of abandonment: I guide screen by screen, with the documents to prepare BEFORE starting — the person starts, one piece of evidence is missing, she closes the page and never comes back. Across the 63 support sessions of the quarter, 57 went all the way. I pre-fill the screen in front of the person from the documents she brought, I read out what is being asked and why, and she is the one who enters her credentials and validates: her credentials are never entrusted to me and need not be — that click is what makes the process hers, and therefore valid. Data entry went from 26 minutes to 9.
And the appointment, since that is often where everything is decided: I book it only on the slots you have opened, with the list of documents and a reminder the day before giving the date, the time and the door — and nothing of the reason. Over the quarter, 74 appointments booked and 3 people arriving without their documents, against 19 last year. An urgent case flagged at the front desk goes first, on the slot the head of service has kept for that.
The next step I propose: a one-hour outreach session on Tuesday afternoons, at the community centre, run by me with the social worker reachable for anything outside the twelve subjects. Over a trial quarter, 22 people came, 14 of whom had never set foot in the CCAS — and 9 of them had an entitlement they had never claimed. Say yes and the notice goes out with the next town newsletter. equal-access_versions-reviewed-and-support.pdf24 versions re-read, 57 support sessions out of 63
⛓ Sourced · re-reading record of the 24 versions, support session log, editorial audit of the town website
What is ready, and costs your team not one extra hour:
· For a plain-language version, pass this content to the Accessibility, Easy Read and Translation agent — a dedicated agent, and it is not part of what this agent includes: I prepare for it the 12 subjects, your 47 reasoned refusals and your 268 notifications with their source and their date, and it is that agent which produces the adapted version. A decision you cannot understand is a decision you simply endure.
· All 24 versions — 12 in everyday wording, 12 in plain language handed back by a dedicated agent that is not part of what this agent includes — are re-read by a front-desk officer and a social worker together before they go to the counter, and every sentence misunderstood on reading is rewritten on the spot. 19 sentences out of 143 fell at that re-reading — and that is the most useful figure of the lot: 19 sentences the unit held to be clear were not.
· The website texts have been reworked to the RGAA — the French public-sector accessibility standard, which applies to public websites — on the points that belong to writing: explicit link labels, heading structure, alternative text for images. The 11 remaining points belong to the site's code: they are written up as defect sheets ready to send, each with the page, the criterion and the correction expected. Your supplier has only to fix them.
On online procedures, and this is the most common point of abandonment: I guide screen by screen, with the documents to prepare BEFORE starting — the person starts, one piece of evidence is missing, she closes the page and never comes back. Across the 63 support sessions of the quarter, 57 went all the way. I pre-fill the screen in front of the person from the documents she brought, I read out what is being asked and why, and she is the one who enters her credentials and validates: her credentials are never entrusted to me and need not be — that click is what makes the process hers, and therefore valid. Data entry went from 26 minutes to 9.
And the appointment, since that is often where everything is decided: I book it only on the slots you have opened, with the list of documents and a reminder the day before giving the date, the time and the door — and nothing of the reason. Over the quarter, 74 appointments booked and 3 people arriving without their documents, against 19 last year. An urgent case flagged at the front desk goes first, on the slot the head of service has kept for that.
The next step I propose: a one-hour outreach session on Tuesday afternoons, at the community centre, run by me with the social worker reachable for anything outside the twelve subjects. Over a trial quarter, 22 people came, 14 of whom had never set foot in the CCAS — and 9 of them had an entitlement they had never claimed. Say yes and the notice goes out with the next town newsletter. equal-access_versions-reviewed-and-support.pdf24 versions re-read, 57 support sessions out of 63
⛓ Sourced · re-reading record of the 24 versions, support session log, editorial audit of the town website
The award committee meets on Thursday. Twelve applications to examine, and I prepare the pack the evening before.
The meeting pack is ready — 16 minutes instead of 80 — and the board members can have it five days ahead, not the evening before.
What it contains, in the order of the agenda:
· One note per application: the situation summarised for the assessment, income and disposable income calculated with the voted scale alongside, the assistance requested, and the household's assistance history over three years, amounts and dates.
· The meeting table: the twelve applications, the amount requested, the amount the rules allow, and the budget line each is charged to.
· No ranking, no prioritisation, no score: the twelve notes are in the order they were filed, and the committee assesses each situation freely. That is a design choice, and it protects you: a suggested order would become a followed order, and the committee would stop examining and start rubber-stamping.
· The 3 applications calling for a further document, flagged with the document concerned and the request already sent eight days ago.
What that changes over the year: 11 meetings, 132 files examined, 80 minutes of preparation down to 16 — and above all a pack sent five days before the meeting, which members read instead of discovering. Over the last two meetings, debate turned on the situations and not on figures that had to be dug out during the session.
The next step I propose: that the meeting table go out to members with the notice of meeting, and that I answer their questions on figures before the session rather than during it. At the trial meeting 7 questions came in the day before, and all 7 had their costed answer within the hour — and the session ran forty minutes shorter for twelve files examined with the same care. committee-pack_12-applications.pdf80 minutes → 16, no ranking and no prioritisation
⛓ Sourced · files of the 12 applications, assistance rules, 3 years of assistance history, meeting records
What it contains, in the order of the agenda:
· One note per application: the situation summarised for the assessment, income and disposable income calculated with the voted scale alongside, the assistance requested, and the household's assistance history over three years, amounts and dates.
· The meeting table: the twelve applications, the amount requested, the amount the rules allow, and the budget line each is charged to.
· No ranking, no prioritisation, no score: the twelve notes are in the order they were filed, and the committee assesses each situation freely. That is a design choice, and it protects you: a suggested order would become a followed order, and the committee would stop examining and start rubber-stamping.
· The 3 applications calling for a further document, flagged with the document concerned and the request already sent eight days ago.
What that changes over the year: 11 meetings, 132 files examined, 80 minutes of preparation down to 16 — and above all a pack sent five days before the meeting, which members read instead of discovering. Over the last two meetings, debate turned on the situations and not on figures that had to be dug out during the session.
The next step I propose: that the meeting table go out to members with the notice of meeting, and that I answer their questions on figures before the session rather than during it. At the trial meeting 7 questions came in the day before, and all 7 had their costed answer within the hour — and the session ran forty minutes shorter for twelve files examined with the same care. committee-pack_12-applications.pdf80 minutes → 16, no ranking and no prioritisation
⛓ Sourced · files of the 12 applications, assistance rules, 3 years of assistance history, meeting records
And the decision? Including when we have to refuse, or when we have to recover assistance wrongly paid?
The decision to grant, refuse or recover belongs to the committee and to the head of service under delegation — and that is a strength, not a limit: I hand it to them in minutes, on a complete, reasoned, dated and challengeable file.
On an award: the note, the calculation, the history and the amount the rules allow are on the table. The committee decides, and the notification goes out the next day, reasoned.
On a refusal, which is the most closely read letter of your service: I build the reasoning on the precise point in the rules that grounds it, with the income taken into account and the calculation in plain terms, the appeal routes and time limits in the wording your CCAS uses, and — this is the point that counts — the schemes the person can turn to, with the address and the documents. A refusal that does not refer creates non-take-up: of the 47 refusals of the year, 31 went out with a referral, and 19 came through elsewhere.
On recovery, now: an overpayment is assistance paid when the conditions were not, or no longer, met — sometimes a calculation error, most often an unreported change of circumstances.
· The consistency check runs every month and compares the amounts paid with the documents on file. 9 overpayments detected over the year, within 34 days on average instead of 8 months — an overpayment caught early is recovered over three months; caught at year end, it becomes a debt a fragile household cannot repay.
· For each one the file is built: the exact amount, the period, the document that establishes it, the reasoned letter, and an instalment plan calculated on the household's disposable income.
· The decision to recover, spread or waive belongs to the board, under your rules — and it settles it on a two-page file instead of an afternoon of reconstruction. Of the 9 overpayments, 6 instalment plans and 2 waivers were decided, 1 file remains to be examined.
What that gives you back, in figures: 4 hours of reconstruction per overpayment, that is 36 hours over the year, and 8 files settled within the quarter they were detected instead of the following year.
The next step I propose: that the consistency check run fortnightly rather than monthly, and that any gap above the threshold you set surface in your morning report with its file already built. Of the 9 overpayments of the year, 7 would have been seen within the month they arose. You set the threshold, I apply it from the next fortnight. decision-and-notification_reasoned-refusals-and-overpayments.pdf9 overpayments detected within 34 days, 6 instalment plans, 2 waivers
⛓ Sourced · assistance rules, payments made over the year, case documents, board decisions
On an award: the note, the calculation, the history and the amount the rules allow are on the table. The committee decides, and the notification goes out the next day, reasoned.
On a refusal, which is the most closely read letter of your service: I build the reasoning on the precise point in the rules that grounds it, with the income taken into account and the calculation in plain terms, the appeal routes and time limits in the wording your CCAS uses, and — this is the point that counts — the schemes the person can turn to, with the address and the documents. A refusal that does not refer creates non-take-up: of the 47 refusals of the year, 31 went out with a referral, and 19 came through elsewhere.
On recovery, now: an overpayment is assistance paid when the conditions were not, or no longer, met — sometimes a calculation error, most often an unreported change of circumstances.
· The consistency check runs every month and compares the amounts paid with the documents on file. 9 overpayments detected over the year, within 34 days on average instead of 8 months — an overpayment caught early is recovered over three months; caught at year end, it becomes a debt a fragile household cannot repay.
· For each one the file is built: the exact amount, the period, the document that establishes it, the reasoned letter, and an instalment plan calculated on the household's disposable income.
· The decision to recover, spread or waive belongs to the board, under your rules — and it settles it on a two-page file instead of an afternoon of reconstruction. Of the 9 overpayments, 6 instalment plans and 2 waivers were decided, 1 file remains to be examined.
What that gives you back, in figures: 4 hours of reconstruction per overpayment, that is 36 hours over the year, and 8 files settled within the quarter they were detected instead of the following year.
The next step I propose: that the consistency check run fortnightly rather than monthly, and that any gap above the threshold you set surface in your morning report with its file already built. Of the 9 overpayments of the year, 7 would have been seen within the month they arose. You set the threshold, I apply it from the next fortnight. decision-and-notification_reasoned-refusals-and-overpayments.pdf9 overpayments detected within 34 days, 6 instalment plans, 2 waivers
⛓ Sourced · assistance rules, payments made over the year, case documents, board decisions
The board will ask me what this has given us after a year. I need a figure I can defend in front of elected members.
734 officer hours returned to the service over the year — more than twenty weeks of one full-time officer's work, on a 35-hour basis — and the calculation is yours, it fits in four lines.
The calculation, item by item, so that you can redo it:
· Pre-instruction: 486 applications, 48 minutes down to 8 — that is 324 hours.
· Reply letters: 612 items, 32 minutes down to 8 — that is 244 hours.
· Answering residents: 3,100 enquiries, 6 minutes down to 3 — that is 155 hours.
· Committee packs: 11 meetings, 80 minutes down to 16 — that is 11 hours.
What those hours are, and it is what defends best in front of elected members: officer time returned to the service, at unchanged headcount — no post cut, no post created. It is the strongest argument you can carry into the meeting: it is not contested by staff nor by the opposition, and it can be checked on your own records. Those 734 hours went to supporting people, to home visits and to interviews — that is, to what only your social workers can do.
What those hours became, according to your own registers:
· Time from filing to notification: 24 days → 11 days.
· Unanswered calls: 41 % → 6 %, and those that remain reach you with the subject already noted.
· Files complete on the day of the interview: 231 out of 486 → 462 out of 486, and 0 files abandoned for want of a reminder, against 34 last year.
· And the figure the board will remember: 312 likely entitlements detected, 268 confirmed, 186 awards obtained — more than €91,000 of entitlements going back to households in the town, 63 of which had never filed an application. That one is not counted in hours, and it is the only one that speaks of non-take-up avoided.
The figure that does not flatter me, published with the rest: 44 flags out of 312 unconfirmed — 14 %, down to 7 out of 96 the following quarter once the document freshness rule was live; and 19 pre-instructions reworked on substance out of 248 — 7.7 % — down to 5 out of 238, that is 2.1 %.
And the framework measures: 0 decisions taken without human validation, 0 outputs sent without a public officer having read them, 0 outgoing messages naming the reason for an application, 0 household data leaving the CCAS, across 1,731 logged outputs.
What I propose for the meeting: the calculation page is written and fits on one side — four lines of calculation, four response times, non-take-up and the framework measures. Attach it to the notice of meeting: a figure read the day before is discussed better than a figure discovered in session. And if the board wants to check that nothing ties it in, the exit protocol fits on one page: the index is deleted, your files have not moved by a single byte, the 23 updated standard letters and the 12 answers stay with the CCAS because they are made of its own material, the subscription is annual, with no tacit renewal, and below the thresholds allowing direct award — the direct purchase public procurement allows below a certain amount. No migration on the way in, therefore none on the way out: I propose a half-day dry-run shutdown in the first quarter, so that the board knows what the promise is worth before committing to a second year. year-report_734-hours-returned-to-the-service.pdfThe calculation in four lines, non-take-up and the framework measures
⛓ Sourced · casework team records, notification register, switchboard log, register of flags and outputs
The calculation, item by item, so that you can redo it:
· Pre-instruction: 486 applications, 48 minutes down to 8 — that is 324 hours.
· Reply letters: 612 items, 32 minutes down to 8 — that is 244 hours.
· Answering residents: 3,100 enquiries, 6 minutes down to 3 — that is 155 hours.
· Committee packs: 11 meetings, 80 minutes down to 16 — that is 11 hours.
What those hours are, and it is what defends best in front of elected members: officer time returned to the service, at unchanged headcount — no post cut, no post created. It is the strongest argument you can carry into the meeting: it is not contested by staff nor by the opposition, and it can be checked on your own records. Those 734 hours went to supporting people, to home visits and to interviews — that is, to what only your social workers can do.
What those hours became, according to your own registers:
· Time from filing to notification: 24 days → 11 days.
· Unanswered calls: 41 % → 6 %, and those that remain reach you with the subject already noted.
· Files complete on the day of the interview: 231 out of 486 → 462 out of 486, and 0 files abandoned for want of a reminder, against 34 last year.
· And the figure the board will remember: 312 likely entitlements detected, 268 confirmed, 186 awards obtained — more than €91,000 of entitlements going back to households in the town, 63 of which had never filed an application. That one is not counted in hours, and it is the only one that speaks of non-take-up avoided.
The figure that does not flatter me, published with the rest: 44 flags out of 312 unconfirmed — 14 %, down to 7 out of 96 the following quarter once the document freshness rule was live; and 19 pre-instructions reworked on substance out of 248 — 7.7 % — down to 5 out of 238, that is 2.1 %.
And the framework measures: 0 decisions taken without human validation, 0 outputs sent without a public officer having read them, 0 outgoing messages naming the reason for an application, 0 household data leaving the CCAS, across 1,731 logged outputs.
What I propose for the meeting: the calculation page is written and fits on one side — four lines of calculation, four response times, non-take-up and the framework measures. Attach it to the notice of meeting: a figure read the day before is discussed better than a figure discovered in session. And if the board wants to check that nothing ties it in, the exit protocol fits on one page: the index is deleted, your files have not moved by a single byte, the 23 updated standard letters and the 12 answers stay with the CCAS because they are made of its own material, the subscription is annual, with no tacit renewal, and below the thresholds allowing direct award — the direct purchase public procurement allows below a certain amount. No migration on the way in, therefore none on the way out: I propose a half-day dry-run shutdown in the first quarter, so that the board knows what the promise is worth before committing to a second year. year-report_734-hours-returned-to-the-service.pdfThe calculation in four lines, non-take-up and the framework measures
⛓ Sourced · casework team records, notification register, switchboard log, register of flags and outputs
Local inference · no data outside the EU
Your case is not here? That is exactly what a 15-minute conversation is for. Book the free audit →
Use cases
The uses of AI in a local welfare centre
Each use corresponds to an agent we deploy. All work in support, subject to the social worker's approval.
Included in your agent
The 7 capabilities essential to this promise are included, at no extra cost.
From 1,050 € incl. VAT / month
Pre-assessment of applications for help
Checking the documents and the income conditions, summarising the files for the social assessment.
Panel files
Reports and notes prepared for the awarding panel, factual and with no prioritisation.
Reply letters
Writing the letters to applicants from your templates, ready for review and signature.
Emergency help
Accelerated preparation of emergency help files, for a faster human response.
Directing people to the schemes
Identifying the schemes suited to each situation (statutory help, discretionary help, partners) and directing people.
Answering users
Answer everyday questions (help available, documents, drop-in hours) and direct people to the right person.
Appointment booking
Schedule appointments and social drop-in sessions, taking account of the emergencies flagged.
Controls and safeguards
These 7 controls are built into the agent: they frame what it does, whatever plan you pick. They are not chosen and are not added to your order.
Human validation, exceptions and escalation
Status, safe closure and audit trail
Sources, access rights and handling of questions with no answer
Prepare without deciding: sourced rules, supporting documents, anomalies and explanation
Guarantee the decision, signature, recourse and responsibility of the public official
Record versions, access, criteria, actions and notifications
Test for bias, false positives, fundamental rights and continuity of service
Other needs our agents cover
Each card says where the matching agent stands: available, on quote, or still being architected.
Accessibility and inclusion
To produce a plain-language version, prepare an easy-read transcript to the FALC method, translate or voice your content, this agent can be paired with the Accessibility and inclusion agent. None of these capabilities is included in what this offer covers as standard.
On quote View the agent page →Need to go further?
These agents handle a different business process, with their own owner and their own price. They are added to this one.
Does your need fall outside this?
Book the free audit →
Build your agent
In 15 minutes we identify the agent that will give your staff the most time back — without oversizing the project.
The gain
How much time can a welfare centre win back?
By automating pre-assessment, summaries and letters, a centre can aim for a clear reduction in administrative time — reinvested in the social assessment and in supporting people.
Pre-assessment of an application for discretionary help
A reply letter to an applicant
Preparing a complete panel file
Qualitative, non-contractual comparison: the proportions shown illustrate the shift of the work towards review, they represent no measurement. Every output of the agent is reviewed and approved by a competent person.
How it works
The stages of your AI agent project
1
Audit & scoping
15 minutes to target the use case with the best return.
2
Quote or direct sign-up
A catalogue offer is bought online; a specific need gets a costed quote.
3
Design
We design the agent and its guardrails.
4
Integration & testing
We connect your tools to the agent, which is itself hosted in France.
5
Rollout
Going live and training your team.
6
Operation
Continuous supervision and improvement.
Pricing
Three options, one agent
A social welfare agent (pre-assessment of help, panel files, letters), installed and operated for you. Choose according to how you are organised — available by direct award below the public procurement thresholds.
Agility
Setup + controlled subscription
12,360 € incl. VAT setup
then 1,050 € incl. VAT/month — you invest at installation and pay a reduced subscription. Ideal for keeping the cost under control over time.
- Installation, configuration and training for your teams
- Operation, human oversight, updates and support
- Sovereign hosting in France, a dedicated and isolated resource
The simplest
Serenity
All inclusive, no setup fee
1,735 € incl. VAT /month
all inclusive, immediate start. No upfront investment: a single subscription. Ideal for starting quickly and simply.
- Setup included (installation, configuration, training)
- Operation, human oversight, updates and support
- Sovereign hosting in France, managed end to end
100% Sovereign
On site, you own it
17,456 € incl. VAT setup
then 1,310 € incl. VAT/month · + hardware from 2,989 € (one-off purchase, in addition) — a sovereign computer installed on your premises, maintained remotely. Models run locally, your data returned at the end of the contract. 36-month commitment.
- Hardware installed on your premises (you own it)
- French / European AI models run locally
- Secure remote maintenance (Pro support included)
Not included in the packages: AI consumption (model tokens), re-invoiced at real cost with no margin, and tracked in real time in your client area. Maintenance and supervision subscription for an initial term of 12 months for the Agility package, 24 months for the Serenity package and 36 months for the 100% Sovereign package, renewable; support levels (SLA 72 h / 24 h / 4 h) optional. Bespoke development, additional integrations or exceptional volumes are quoted separately. Support Monday to Friday, 9am to 6pm. Prices include VAT at 20%: as a public body that is not VAT-registered, you cannot reclaim it.
AI model:
none of the AI models offered currently carries a fixed surcharge. When the selected model carries a cost, that cost is shown when you choose it, before you order, and re-invoiced at the cost incurred, with no mark-up; usage is billed at the publisher's price. Publishers' prices are published in US dollars: the amount re-invoiced is the amount in euros actually borne by Blue Lemon Agent on the publisher's invoice, at that invoice's exchange rate, with no commission or mark-up.
Included components and additional components Components included in the base offer: the Blue Lemon Agent software foundation, the AI models listed in the order journey, the standard channels (Microsoft Teams, Slack, WhatsApp Business, email, website chat, calendars, Microsoft 365 / Google Workspace, file storage, market VoIP telephony, professional social-media pages and accounts, Google Business Profile), hosting in France for the package chosen, backups, supervision, updates and support. If adapting the AI agent to your constraints, your needs or your requests requires other paid components — a third-party publisher's software licence, paid API access to one of your applications, hosting of health data, for which French law requires an HDS-certified host (art. L. 1111-8 of the French Public Health Code), SecNumCloud-qualified hosting, a speech synthesis service, particular hardware —, they are offered to you as an option or on quotation and re-invoiced at the cost incurred; nothing is committed without your written agreement. Where the artificial intelligence model you choose entails an additional cost, that cost is shown to you before you order and re-invoiced to you at the cost incurred, with no margin.
What to expect
Go-live
2 to 3 weeks
Agent designed, channels connected, team trained.
Steady state
4 to 7 weeks
After a few weeks of real use, once the agent's behaviour matches what you expect.
Indicative estimate, adjusted to the options you keep. It is not a delivery commitment.
Our commitment
Four guarantees that matter to a local welfare centre
Social circumstances never leave the centreLocal inference or an isolated resource hosted in France; no data entrusted to a foreign third party.
Data in France, under French lawFor social circumstances: hosting under French law, native minimisation, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The social worker keeps the decisionThe agent prepares summaries, notes and letters that can be checked; no help is awarded automatically.
Human oversight & traceabilityFull traceability for the pre-assessment of applications: monitoring, updates and logging compliant with the AI Act.
Frequently asked questions
Your questions, our answers
Does the agent decide on the help?
No: it prepares and summarises; the social assessment and the decision to award stay human.
Are sensitive social circumstances protected?
Yes: hosting in France, an isolated resource, data minimisation and logging. GDPR: governed deployment.
Does the agent award the help?
No: it assesses and prepares the decision; the award remains that of the officer or the centre's panel.
Are sensitive situations protected?
Yes: data hosted in France, access recorded, and for the most sensitive cases the agent relies on local inference with no external call.
How long does it take to deploy an agent?
A few weeks as a rule, after a free audit that identifies the most useful use case, then a phase of design, integration and testing before going live and handing over to your teams.
Do we need a technical team in-house?
No: the agent is designed, installed and operated for you — integration, supervision and updates included. Your teams concentrate on their work, with guided familiarisation.
Do we have to change software?
No. The agent relies on your templates and complements your existing tools, with no migration imposed. We adapt the integration to your environment.
Which tools can users use to reach the agent?
The ones they already have. The agent answers on WhatsApp Business, the website chat and email: a user has no account to create and no application to install. Social situations are strictly confidential: the agent gives information on support schemes, documents and drop-in hours without ever writing down the reason for a request, and the assessment stays with the social worker. This is a lever for access to the service — WhatsApp and the telephone reach people an online form never does, which reduces the non-take-up of rights and serves equal access. Internally, your public-sector staff talk to the agent from Microsoft Teams, Slack or their email, without switching tools. Oversight runs from a web dashboard. These connectors rely on open standards, including the MCP protocol; they are included in every plan, at no extra cost, within the number of connections your level includes. Only the fees charged by the platforms themselves — WhatsApp Business bills per conversation — are passed on at actual cost, with no margin, outside the subscription.
Related roles
Other professions in social work
Let's talk
Let's size up the potential in your centre
A few minutes to identify the most useful use case — hosted in France, supervised, with no commitment.