● Public sector — Assessing grants & subsidies
The AI agent for case officers: pre-assess the applications, prepare the decision
Checking completeness, verifying consistency, testing every application against the criteria of the funding rules: this work takes up a considerable share of case officers' time — at the expense of fast and equal handling for the public. Your AI agent absorbs that preliminary assessment. Hosted in France — on local inference or an isolated resource — applicants' data stays under control. The public officer keeps the decision.
Hosted in France
Applicants' data protected
GDPR & AI Act: governed deployment
Human oversight
Updated on
Deployed in a few weeks
Assessment assistant · hosted in France
On the grant application from the “Sport pour Tous” association (operating grant scheme), check completeness and consistency before I assess it.
File analysed. 11 documents out of 12 present. Missing document: the income statement for the last closed financial year, required by your funding rules. 2 points to check: the forecast budget shows an imbalance of €1,240, and the company registration number provided differs from the one on the last application submitted.
I am preparing a draft request for additional documents — for approval.
⛓ Source · the application submitted + the scheme's funding rules
I am preparing a draft request for additional documents — for approval.
⛓ Source · the application submitted + the scheme's funding rules
Is the association eligible under our criteria?
Preliminary assessment against the rules: registered office in the eligible area, purpose compliant, grant ceiling respected. Reservation: the rule against combining this with the county grant cannot be verified as things stand — the corresponding supporting document is absent. Recommendation: make the assessment conditional on that document.
I am drawing up a reasoned preliminary assessment note, point by point, for your review.
✎ Action · note ready for review — the case officer decides
I am drawing up a reasoned preliminary assessment note, point by point, for your review.
✎ Action · note ready for review — the case officer decides
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 case officer decidesThe agent assists, never decides alone
✦ In brief
In an assessment department, a Blue Lemon Agent agent pre-assesses applications for grants, subsidies and authorisations — completeness checks, consistency, testing against the criteria of the funding rules, preparation of decision letters. It runs on local inference or is hosted in France: applicants' data is never exposed to a foreign service, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity. The AI agent assists, the public officer decides: no decision about entitlement is automated. The time saved serves faster and more equal handling for the public. Live within a few weeks.
100%
hosted in France in the target architecture
0
transfer outside the EU in the target architecture
9
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 assessment services — and why they hesitate
The public expects quick, readable decisions; elected members expect controlled turnaround times and irreproachable equal treatment. But checking completeness and testing every application against the rules consumes most of case officers' time — and the data involved touches on applicants' personal and financial circumstances.
! The issue
The department is caught between a lengthening flow of applications and the requirement for handling that is at once fast, reasoned and strictly equal from one file to the next. Yet most consumer AI solutions amount to entrusting applicants' personal, social and financial data to a third party, often hosted outside Europe and subject to the Cloud Act — unacceptable for public data.
✓ Our answer
AI is only of interest to an assessment department if it is sovereign and confidential by design. Local inference or an isolated resource hosted in France, reinforced human oversight, decisions on entitlement reserved to the public officer: the time saved on preliminary assessment is never paid for in lost confidentiality or automated entitlement. The aim is not to replace the case officer, but to give them back time for the files that demand judgement.
The decisive point
Protecting applicants' data: sovereignty & compliance
An assessment department handles users' most sensitive data. Here is how the architecture of our agents protects it, file by file.
Local inference
The agent can run on a machine belonging to the authority: no document from a file leaves the network, nothing passes through a cloud.
Hosting in France
Otherwise, a dedicated and isolated resource, hosted in France under French law — applicants' data: processing and access within the European Union targeted by the architecture.
Reduced extraterritorial exposure
For applicants' data, the architecture aims to reduce exposure to the Cloud Act and FISA 702; being located in France or in the European Union does not, on its own, guarantee immunity: the American hyperscalers have no access to it, even when hosted in Europe.
One isolated resource per organisation
No pooling of data: an environment strictly dedicated to your authority or institution.
Encryption & controlled access
Encryption in transit and at rest, role-based access (RBAC), strong authentication and logging.
AI Act: governed deployment
An agent strictly in support; no decision on entitlement automated; reinforced human oversight and traceability 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 depending on your requirements. 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 bodyPierre-Chastel Urban Community — grants and subsidies assessment department
- Sector
- Assessment department of an urban community — 21 schemes: grants to associations, business support, home renovation and mobility grants for households, two social schemes voted by the community council
- Headcount
- 34 staff in the grants department — including 12 assessment officers, 5 on reception and information, 2 on controls and one data protection officer
- Public served
- 6,400 applications filed a year — 2,900 associations, 1,700 households, 1,100 businesses, 700 member municipalities — and 18,000 questions asked by phone, e-mail or at the counter
- Order of magnitude
- A grant file takes 100 minutes of officer time from filing to decision; the two-month deadline the scheme rules set; 4 award committees a year
- Tools in place
- Online filing platform, assessment software, document management system, letter templates and 3 years of filed cases — the agent plugs into them read-only, nothing is replaced and nothing is migrated
- Who decides
- The assessment officer settles the pre-assessment and signs the request for documents; the award committee advises, the competent authority decides and signs the award, the refusal or the withdrawal; the data protection officer keeps the register
- Room for improvement
- 52 % of files arrive incomplete; average assessment time is 71 days against rules that set 60; 47 files last year had the same invoice funded by two schemes; and 890 applicants meet the conditions of a grant they have never claimed
At Pierre-Chastel, 12 officers handle 6,400 applications a year across 21 schemes whose rules change with each resolution. The job is not to check documents: it is to apply the same rules to everyone, within the time announced, and to give reasons for what is decided. The agent runs on local inference on a machine of the urban community, reads the three years of files without ever writing to them, and no award, refusal or withdrawal leaves without the signature of the competent public official. The exchanges that follow span a year, from the review of the rules to the results presented to the community council.
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.
Assessment assistant · hosted in France
I reread your 21 sets of scheme rules and your three years of files over the weekend — 19,200 applications assessed — and I draw two measurements from them that your files were carrying without anyone having the time to add them up.
The first, and it governs everything else: 52 % of your files arrive incomplete, and 7 gaps out of 10 come down to five documents — the profit and loss account for the last closed financial year, the insurance certificate, bank details in the applicant's name, the itemised quotation and the association's resolution authorising the application. This is not 3,300 applicants being careless: it is a leaflet that does not say precisely enough what is expected.
The second, and it is more awkward: 6 of your 21 sets of rules are no longer in their version in force in your tools. The association running-grant scheme still carries the ceiling of a resolution replaced last year, and two eligibility criteria were amended in session without the published leaflet saying so. This is not an assessment error: these are rules that aged while nobody had the time to reopen them.
What the three items you already measure cost:
· Pre-assessing a grant file accounts for 45 % of assessment time — 45 minutes out of the 100 a file takes from filing to decision.
· The completeness check: 25 %, that is 25 minutes.
· Answering a recurring question from an applicant: 10 %, that is 10 minutes.
What I propose, and it is not a promise — it is already written: the 6 sets of rules are restored to their version in force, each with the resolution that sets it and its effective date, with the replaced line and the line replacing it side by side; and the five document leaflets are rewritten in plain language, with an example of the document expected.
What that would have changed over the past year: 2,700 files would have left complete on the day of filing, and the 214 awards made under the association scheme would have been made on the ceiling voted, not on the repealed one.
The time that moves: completeness drops from 25 % to 8 % — from 25 minutes to 8.
The next step, and it takes an hour: you reread the 6 restored rulebooks, I present them to you tomorrow morning. The department settles the version, and it is in use that same evening. current-picture_21-schemes-and-three-time-items.pdf52 % of incomplete files, 6 rulebooks to restore
⛓ Sourced · 21 sets of scheme rules, register of resolutions, 3 years of assessed files, platform leaflets
The first, and it governs everything else: 52 % of your files arrive incomplete, and 7 gaps out of 10 come down to five documents — the profit and loss account for the last closed financial year, the insurance certificate, bank details in the applicant's name, the itemised quotation and the association's resolution authorising the application. This is not 3,300 applicants being careless: it is a leaflet that does not say precisely enough what is expected.
The second, and it is more awkward: 6 of your 21 sets of rules are no longer in their version in force in your tools. The association running-grant scheme still carries the ceiling of a resolution replaced last year, and two eligibility criteria were amended in session without the published leaflet saying so. This is not an assessment error: these are rules that aged while nobody had the time to reopen them.
What the three items you already measure cost:
· Pre-assessing a grant file accounts for 45 % of assessment time — 45 minutes out of the 100 a file takes from filing to decision.
· The completeness check: 25 %, that is 25 minutes.
· Answering a recurring question from an applicant: 10 %, that is 10 minutes.
What I propose, and it is not a promise — it is already written: the 6 sets of rules are restored to their version in force, each with the resolution that sets it and its effective date, with the replaced line and the line replacing it side by side; and the five document leaflets are rewritten in plain language, with an example of the document expected.
What that would have changed over the past year: 2,700 files would have left complete on the day of filing, and the 214 awards made under the association scheme would have been made on the ceiling voted, not on the repealed one.
The time that moves: completeness drops from 25 % to 8 % — from 25 minutes to 8.
The next step, and it takes an hour: you reread the 6 restored rulebooks, I present them to you tomorrow morning. The department settles the version, and it is in use that same evening. current-picture_21-schemes-and-three-time-items.pdf52 % of incomplete files, 6 rulebooks to restore
⛓ Sourced · 21 sets of scheme rules, register of resolutions, 3 years of assessed files, platform leaflets
We are 12 officers for 21 schemes and 6,400 applications. We cannot open everything at once: where do we start?
With three schemes, not twenty-one — and I have ranked them by the time they give you back, not by nomenclature order.
What the count over your three years says:
· Association running grants: 2,100 applications a year, that is 33 % of your flow, and it is the rulebook whose ceiling is out of date in your tools.
· Home renovation grants: 1,200 applications, the heaviest in documents — 11 documents required, of which the itemised quotation is missing one time in three.
· Mobility grants: 900 applications, the simplest, and the one where the two-month deadline is most often missed because nobody has time for it between two heavy files.
In other words: three schemes out of twenty-one carry 4,200 of the 6,400 applications of the year. The other eighteen can wait until autumn without anything getting worse for an applicant.
What I bring you within that hour, and nobody has the time to do: the three rulebooks restored line by line to their version in force, the document lists set up for checking at filing, and the leaflets rewritten — with the number of files each would have sent back complete last year: 1,400, 780 and 520.
The gain, measured on those three schemes: 4,200 files, 17 minutes given back on each on completeness alone — 1,190 officer hours that go back to the files that call for judgement.
What I propose next: that I flag of my own accord any rulebook a resolution amends and write the updated version overnight after the session. Your 21 rulebooks will stop ageing in silence — and an applicant will stop having a criterion applied to them that the council has replaced. current-picture_21-schemes-and-three-time-items.pdfThree schemes carry 4,200 of the 6,400 applications
⛓ Sourced · application counts by scheme over 3 years, register of resolutions, rewritten leaflets
What the count over your three years says:
· Association running grants: 2,100 applications a year, that is 33 % of your flow, and it is the rulebook whose ceiling is out of date in your tools.
· Home renovation grants: 1,200 applications, the heaviest in documents — 11 documents required, of which the itemised quotation is missing one time in three.
· Mobility grants: 900 applications, the simplest, and the one where the two-month deadline is most often missed because nobody has time for it between two heavy files.
In other words: three schemes out of twenty-one carry 4,200 of the 6,400 applications of the year. The other eighteen can wait until autumn without anything getting worse for an applicant.
What I bring you within that hour, and nobody has the time to do: the three rulebooks restored line by line to their version in force, the document lists set up for checking at filing, and the leaflets rewritten — with the number of files each would have sent back complete last year: 1,400, 780 and 520.
The gain, measured on those three schemes: 4,200 files, 17 minutes given back on each on completeness alone — 1,190 officer hours that go back to the files that call for judgement.
What I propose next: that I flag of my own accord any rulebook a resolution amends and write the updated version overnight after the session. Your 21 rulebooks will stop ageing in silence — and an applicant will stop having a criterion applied to them that the council has replaced. current-picture_21-schemes-and-three-time-items.pdfThree schemes carry 4,200 of the 6,400 applications
⛓ Sourced · application counts by scheme over 3 years, register of resolutions, rewritten leaflets
All of this assumes you read association accounts, tax assessments, quotations. Where does that data go?
It goes nowhere. I run on local inference on a machine of the urban community, and nothing I read leaves your network.
Local inference means the model computes on your machine: the text of a tax assessment or a profit and loss account crosses no outside network to be processed. If the authority would rather not host a machine, the other route is an isolated resource hosted in France, dedicated to Pierre-Chastel — no pooling with another authority.
What that changes, point by point:
· Your applicants' data trains 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 right to write — that is verified with one command, not on trust.
· Encryption in transit and at rest, role-based access — rights follow the job: a reception officer sees how far a file has got, not the financial documents; controls see the documents, not the reception exchanges. 6 roles for your 34 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. It is that log which answers the public accountant as well as an applicant who challenges a decision.
And the act the law reserves to a person, which is exactly what gives your decisions their value: awarding, refusing or withdrawing a grant has effects on the applicant — it cannot rest on automated processing alone. The decision is taken by the competent public official, reasoned, dated, traced, and the applicant can challenge it. Everything leading up to it, I have already done: documents read, criteria of the rules in force checked one by one, reasoned note written, and the draft letter ready with its routes and time limits for appeal. A decision nobody had signed could be challenged by nobody, and would be worth nothing before an applicant or before the administrative court.
The figure that sums this up: 0 applicant data out of the authority's network across the 6,400 applications of the year, and processing in the EU targeted.
What I propose: that I keep up to date the record your data protection officer and your council will ask for — hosting, data processed, retention periods, who accesses what. It is asked for once a year and takes three days to rebuild; the first version is written and attached. technical-framework_where-applicant-data-lives.pdfLocal inference, read-only, processing in the EU targeted
✎ Framework · deployment architecture, technical account rights, access log, first version of the register record
Local inference means the model computes on your machine: the text of a tax assessment or a profit and loss account crosses no outside network to be processed. If the authority would rather not host a machine, the other route is an isolated resource hosted in France, dedicated to Pierre-Chastel — no pooling with another authority.
What that changes, point by point:
· Your applicants' data trains 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 right to write — that is verified with one command, not on trust.
· Encryption in transit and at rest, role-based access — rights follow the job: a reception officer sees how far a file has got, not the financial documents; controls see the documents, not the reception exchanges. 6 roles for your 34 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. It is that log which answers the public accountant as well as an applicant who challenges a decision.
And the act the law reserves to a person, which is exactly what gives your decisions their value: awarding, refusing or withdrawing a grant has effects on the applicant — it cannot rest on automated processing alone. The decision is taken by the competent public official, reasoned, dated, traced, and the applicant can challenge it. Everything leading up to it, I have already done: documents read, criteria of the rules in force checked one by one, reasoned note written, and the draft letter ready with its routes and time limits for appeal. A decision nobody had signed could be challenged by nobody, and would be worth nothing before an applicant or before the administrative court.
The figure that sums this up: 0 applicant data out of the authority's network across the 6,400 applications of the year, and processing in the EU targeted.
What I propose: that I keep up to date the record your data protection officer and your council will ask for — hosting, data processed, retention periods, who accesses what. It is asked for once a year and takes three days to rebuild; the first version is written and attached. technical-framework_where-applicant-data-lives.pdfLocal inference, read-only, processing in the EU targeted
✎ Framework · deployment architecture, technical account rights, access log, first version of the register record
The association Sport for All has just filed an application for a running grant. What exactly do you do with it?
The file is read, checked and pre-sorted in four minutes, and the association's chair knows the same day what is missing.
Completeness is the check that every document required by the scheme rules is present; consistency is the matching of amounts, dates and identifiers between those documents.
What I hand over, document by document:
· The 12 documents required by the scheme, ticked one by one: 11 present, 1 missing — the profit and loss account for the last closed financial year, required by the article of the rules that I quote in plain terms in the note.
· The values extracted and verified: amount requested, forecast budget, grants already obtained, number of members, the association's identification number. Each value carries the document and page it comes from, and can be checked with one click.
· Three consistency checks, and two call for your eye: the forecast budget shows an imbalance of €1,240 between costs and income, and the identification number differs from last year's file — a declared change of registered office or a typing error; both documents are side by side and you settle it in ten seconds.
· The scheme's ceiling compared with the amount requested, on the version of the rules in force, with the resolution that sets it and its effective date.
· The summary handed to the applicant, in plain language: what has been received, what is missing, by when, what happens next.
The time that moves: the completeness check goes from 25 % to 8 % of assessment time — from 25 minutes to 8. Across 6,400 applications a year, that is 1,813 hours, and above all it is the back-and-forth that disappears, not the check.
The next step I propose: that this check run at filing, on the platform, before the file even reaches you — the applicant sees what is missing while their documents are still to hand. Over last year, 2,700 files would have left complete on the day of filing; approve the principle and the check is live tomorrow. reading-the-documents_sport-for-all-file.pdf12 documents ticked, 5 values extracted and sourced, 2 gaps flagged
⛓ Sourced · documents filed, scheme rules in force and their resolution, previous year's file
Completeness is the check that every document required by the scheme rules is present; consistency is the matching of amounts, dates and identifiers between those documents.
What I hand over, document by document:
· The 12 documents required by the scheme, ticked one by one: 11 present, 1 missing — the profit and loss account for the last closed financial year, required by the article of the rules that I quote in plain terms in the note.
· The values extracted and verified: amount requested, forecast budget, grants already obtained, number of members, the association's identification number. Each value carries the document and page it comes from, and can be checked with one click.
· Three consistency checks, and two call for your eye: the forecast budget shows an imbalance of €1,240 between costs and income, and the identification number differs from last year's file — a declared change of registered office or a typing error; both documents are side by side and you settle it in ten seconds.
· The scheme's ceiling compared with the amount requested, on the version of the rules in force, with the resolution that sets it and its effective date.
· The summary handed to the applicant, in plain language: what has been received, what is missing, by when, what happens next.
The time that moves: the completeness check goes from 25 % to 8 % of assessment time — from 25 minutes to 8. Across 6,400 applications a year, that is 1,813 hours, and above all it is the back-and-forth that disappears, not the check.
The next step I propose: that this check run at filing, on the platform, before the file even reaches you — the applicant sees what is missing while their documents are still to hand. Over last year, 2,700 files would have left complete on the day of filing; approve the principle and the check is live tomorrow. reading-the-documents_sport-for-all-file.pdf12 documents ticked, 5 values extracted and sourced, 2 gaps flagged
⛓ Sourced · documents filed, scheme rules in force and their resolution, previous year's file
And when a document is unreadable, or an amount cannot be read? That is where we make mistakes, and the applicant pays for them.
A value I have not read, I do not write — I ask for it, the same day, from the right person, with the request already drafted on your template.
What that gives, measured over the quarter:
· 1,060 values unreadable or absent — cropped scan, truncated financial table, quotation with no pre-tax amount.
· 1,060 requests for documents prepared, each naming the document expected, the line that is missing and the reason in one sentence — never “file not compliant”, which brings back the wrong paper and costs everyone a fortnight.
· 931 replies back within eight days, against an average of 24 days last year.
What I add and nobody has the time to do: when the document already exists in an earlier file of the same applicant and is still valid, I bring it up instead of asking for it. Over the quarter, 380 documents did not have to be requested again — 380 applicants who were not asked for a paper they had already provided, which is the leading cause of a grant application being abandoned.
And what that does to the deadline, the only commitment the applicant remembers: average assessment time goes from 71 days to 38, and the two-month deadline your rules set is met 96 times out of 100 against 58.
Where those 33 days come from, item by item: the request for documents goes out on the day of filing instead of the twelfth day — 12 days; the halfway reminder brings documents back within eight days instead of 24 — 11 days; the file reaches the officer pre-assessed — 10 days.
The next step I propose: that no file spends 45 days without an officer being alerted by name, with the exact reason for the wait and the action that unblocks it. Over the past year, 780 files would have been picked up before that threshold — you approve the threshold, I set it tonight and hand you the list every Monday. completeness-and-consistency_1060-values-requested.pdf380 documents not asked again, 71 days brought down to 38
⛓ Sourced · log of unread values, document requests of the quarter, applicants' earlier files, assessment times over 12 months
What that gives, measured over the quarter:
· 1,060 values unreadable or absent — cropped scan, truncated financial table, quotation with no pre-tax amount.
· 1,060 requests for documents prepared, each naming the document expected, the line that is missing and the reason in one sentence — never “file not compliant”, which brings back the wrong paper and costs everyone a fortnight.
· 931 replies back within eight days, against an average of 24 days last year.
What I add and nobody has the time to do: when the document already exists in an earlier file of the same applicant and is still valid, I bring it up instead of asking for it. Over the quarter, 380 documents did not have to be requested again — 380 applicants who were not asked for a paper they had already provided, which is the leading cause of a grant application being abandoned.
And what that does to the deadline, the only commitment the applicant remembers: average assessment time goes from 71 days to 38, and the two-month deadline your rules set is met 96 times out of 100 against 58.
Where those 33 days come from, item by item: the request for documents goes out on the day of filing instead of the twelfth day — 12 days; the halfway reminder brings documents back within eight days instead of 24 — 11 days; the file reaches the officer pre-assessed — 10 days.
The next step I propose: that no file spends 45 days without an officer being alerted by name, with the exact reason for the wait and the action that unblocks it. Over the past year, 780 files would have been picked up before that threshold — you approve the threshold, I set it tonight and hand you the list every Monday. completeness-and-consistency_1060-values-requested.pdf380 documents not asked again, 71 days brought down to 38
⛓ Sourced · log of unread values, document requests of the quarter, applicants' earlier files, assessment times over 12 months
On consistency in particular: we have little time to cross-check amounts from one file to another, and that is where errors slip through.
The cross-checks are done on all 6,400 files, not on the ones there is time to look at — and they cover five points, every one verifiable with the document in hand.
The five cross-checks, and what they gave over twelve months:
· Amount requested against the ceiling in force — 112 files above the ceiling, of which 79 through not knowing a ceiling amended during the year. Those 79 got the right information before assessment rather than a refusal afterwards.
· Forecast budget balanced — 240 imbalances, each flagged with the two lines at issue.
· Applicant identifier consistent with earlier files — 61 discrepancies, of which 44 genuine changes of registered office and 17 typing errors corrected before assessment.
· Quotations and invoices matched from one file to another — the check that brought the most, and it has a document of its own.
· Dates of filing and of committing the expenditure — 134 files where the expenditure was committed before filing, which several of your rulebooks exclude: the applicant knew before incurring the costs of applying.
What that changes for equal treatment, and it is the argument your council will remember: the same five cross-checks, on all 6,400 files, whoever the officer and whatever the month. On 300 files rechecked by your internal control, the interpretation gap between officers has gone from 11 % to 3 % — and the remaining 3 % are assumed judgements, traced and reasoned, which is exactly what a judge expects.
The next step I propose: that the summary of the cross-checks go to the applicant with the acknowledgement of receipt, when it flags nothing as much as when it flags something. An applicant who knows their file is complete and consistent does not call three times to check — and over the trial quarter, follow-up calls fell by 41 %. completeness-and-consistency_1060-values-requested.pdf5 cross-checks on 6,400 files, interpretation gap from 11 % to 3 %
⛓ Sourced · 12 months of filed applications, scheme rules and their resolutions, earlier files, 300 files rechecked by internal control
The five cross-checks, and what they gave over twelve months:
· Amount requested against the ceiling in force — 112 files above the ceiling, of which 79 through not knowing a ceiling amended during the year. Those 79 got the right information before assessment rather than a refusal afterwards.
· Forecast budget balanced — 240 imbalances, each flagged with the two lines at issue.
· Applicant identifier consistent with earlier files — 61 discrepancies, of which 44 genuine changes of registered office and 17 typing errors corrected before assessment.
· Quotations and invoices matched from one file to another — the check that brought the most, and it has a document of its own.
· Dates of filing and of committing the expenditure — 134 files where the expenditure was committed before filing, which several of your rulebooks exclude: the applicant knew before incurring the costs of applying.
What that changes for equal treatment, and it is the argument your council will remember: the same five cross-checks, on all 6,400 files, whoever the officer and whatever the month. On 300 files rechecked by your internal control, the interpretation gap between officers has gone from 11 % to 3 % — and the remaining 3 % are assumed judgements, traced and reasoned, which is exactly what a judge expects.
The next step I propose: that the summary of the cross-checks go to the applicant with the acknowledgement of receipt, when it flags nothing as much as when it flags something. An applicant who knows their file is complete and consistent does not call three times to check — and over the trial quarter, follow-up calls fell by 41 %. completeness-and-consistency_1060-values-requested.pdf5 cross-checks on 6,400 files, interpretation gap from 11 % to 3 %
⛓ Sourced · 12 months of filed applications, scheme rules and their resolutions, earlier files, 300 files rechecked by internal control
The Sport for All file is complete. What is on my desk when I open it to assess it?
A pre-assessment note reasoned criterion by criterion, the version of the rules in force alongside, and the draft letter already written — you read and settle your position in twelve minutes.
What the note carries, in this order:
· The eligibility criteria of the rules, one by one, with the file's value opposite and the document it comes from. Registered office in the eligible area: met. Statutory purpose consistent with the scheme: met, statutes on page 4. Grant ceiling respected: €4,200 requested against a ceiling of €6,000 — met.
· The criterion that calls for your judgement, set apart: the no-double-funding rule with the departmental grant cannot be verified from the file's documents alone. I neither settle it nor leave it hanging: the certificate to request is identified, the letter is written, and it goes out from you with one click. Over the quarter, that document came back within eight days in 9 cases out of 10.
· The calculation of the amount proposed, in plain terms, with each coefficient of the scale and its source, and the effective date of the resolution that sets it.
· The three comparable files assessed in the last twelve months, with the decision taken and its reasoning — that is what lets you stay consistent from one file to the next without having to remember.
The decision to award stays with the competent authority — and I hand it over in twelve minutes, complete file, criteria ticked and reasoning already written, where the same assessment used to take three quarters of an hour of reconstruction.
The time that moves: pre-assessment goes from 45 % to 12 % of the file's time — from 45 minutes to 12. Across 6,400 applications a year, that is 3,520 officer hours given back to the files that call for judgement.
The next step I propose: that the note be produced for all 21 schemes and not only the first three. The other eighteen rulebooks are already read and their criteria set out line by line — all I need is two hours of your rereading, and next year's 6,400 files will all arrive pre-assessed. pre-assessment-note_sport-for-all.pdfCriteria ticked one by one, amount calculated, 3 comparable files
⛓ Sourced · scheme rules in force and their resolution, documents in the file, 3 comparable files from the last 12 months
What the note carries, in this order:
· The eligibility criteria of the rules, one by one, with the file's value opposite and the document it comes from. Registered office in the eligible area: met. Statutory purpose consistent with the scheme: met, statutes on page 4. Grant ceiling respected: €4,200 requested against a ceiling of €6,000 — met.
· The criterion that calls for your judgement, set apart: the no-double-funding rule with the departmental grant cannot be verified from the file's documents alone. I neither settle it nor leave it hanging: the certificate to request is identified, the letter is written, and it goes out from you with one click. Over the quarter, that document came back within eight days in 9 cases out of 10.
· The calculation of the amount proposed, in plain terms, with each coefficient of the scale and its source, and the effective date of the resolution that sets it.
· The three comparable files assessed in the last twelve months, with the decision taken and its reasoning — that is what lets you stay consistent from one file to the next without having to remember.
The decision to award stays with the competent authority — and I hand it over in twelve minutes, complete file, criteria ticked and reasoning already written, where the same assessment used to take three quarters of an hour of reconstruction.
The time that moves: pre-assessment goes from 45 % to 12 % of the file's time — from 45 minutes to 12. Across 6,400 applications a year, that is 3,520 officer hours given back to the files that call for judgement.
The next step I propose: that the note be produced for all 21 schemes and not only the first three. The other eighteen rulebooks are already read and their criteria set out line by line — all I need is two hours of your rereading, and next year's 6,400 files will all arrive pre-assessed. pre-assessment-note_sport-for-all.pdfCriteria ticked one by one, amount calculated, 3 comparable files
⛓ Sourced · scheme rules in force and their resolution, documents in the file, 3 comparable files from the last 12 months
What then takes us the most time are the letters: award, reasoned refusal, request for documents. And a poorly reasoned refusal comes back as an appeal.
The three letters are written on your templates, reasoned on the rules in force, and the refusal is the one I take most care over — because it is the one that gets challenged.
What a draft refusal carries:
· The precise criterion that is not met, quoted in plain terms in the words of the rules, with the resolution that sets it and its effective date.
· The file's value opposite the criterion — “the closed financial year presented covers 9 months, the rules require 12 months” — an applicant understands a refusal with figures, they do not understand a refusal in administrative language.
· What the applicant can do: refile, when and in what form, and the other scheme of the authority their situation makes them eligible for when there is one. Of the 214 refusals of the quarter, 61 applicants were eligible for another scheme — and 43 filed for it.
· The routes and time limits for appeal, always.
What that changes, measured: the time to notify a decision goes from 24 days to 3, and informal appeals against unfavourable decisions go from 14 % to 5 % — not because appeals are discouraged, but because a decision that explains its criterion leaves less room for misunderstanding. The remaining 5 % are genuine disagreements, and they arrive with a file already reasoned to answer them.
And for the award letter, which nobody thinks of polishing: it carries the amount, the payment schedule, the documents required for the balance and the date they are expected. Over the quarter, requests for the balance arriving complete first time went from 46 % to 88 % — that is time your department does not spend chasing, and cash that reaches the association on time.
The signature stays with the competent authority: it is what makes the decision open to challenge, therefore defensible before the administrative court as before the public accountant.
The next step I propose: that the three templates be reviewed once with your legal officer, from the 214 refusals of the quarter — I have picked out the 7 wordings that recur most in appeals, and the 7 rewrites are ready. decision-letters_award-refusal-request-for-documents.pdf24 days brought down to 3, informal appeals from 14 % to 5 %
⛓ Sourced · the department's letter templates, rules in force, 214 refusals of the quarter, register of informal appeals
What a draft refusal carries:
· The precise criterion that is not met, quoted in plain terms in the words of the rules, with the resolution that sets it and its effective date.
· The file's value opposite the criterion — “the closed financial year presented covers 9 months, the rules require 12 months” — an applicant understands a refusal with figures, they do not understand a refusal in administrative language.
· What the applicant can do: refile, when and in what form, and the other scheme of the authority their situation makes them eligible for when there is one. Of the 214 refusals of the quarter, 61 applicants were eligible for another scheme — and 43 filed for it.
· The routes and time limits for appeal, always.
What that changes, measured: the time to notify a decision goes from 24 days to 3, and informal appeals against unfavourable decisions go from 14 % to 5 % — not because appeals are discouraged, but because a decision that explains its criterion leaves less room for misunderstanding. The remaining 5 % are genuine disagreements, and they arrive with a file already reasoned to answer them.
And for the award letter, which nobody thinks of polishing: it carries the amount, the payment schedule, the documents required for the balance and the date they are expected. Over the quarter, requests for the balance arriving complete first time went from 46 % to 88 % — that is time your department does not spend chasing, and cash that reaches the association on time.
The signature stays with the competent authority: it is what makes the decision open to challenge, therefore defensible before the administrative court as before the public accountant.
The next step I propose: that the three templates be reviewed once with your legal officer, from the 214 refusals of the quarter — I have picked out the 7 wordings that recur most in appeals, and the 7 rewrites are ready. decision-letters_award-refusal-request-for-documents.pdf24 days brought down to 3, informal appeals from 14 % to 5 %
⛓ Sourced · the department's letter templates, rules in force, 214 refusals of the quarter, register of informal appeals
And the award committee? Preparing the report is a week's work, four times a year.
The committee report is built as things go and is available at all times: there is no longer a week of preparation, there is half a day of rereading.
What the report contains, file by file: the purpose of the application, the amount requested, the criteria met and not met, the amount proposed and the reasoning in three lines — always in the same order, which lets committee members read 60 files without losing their way.
What I add and that did not exist:
· A reminder of the grants already awarded to the same applicant by the authority over three years, across all schemes. It is the question asked at every committee, and nobody had the time to prepare it.
· The breakdown of proposals by member municipality and by type of applicant — your elected members ask for it one time in two, and it took half a day to rebuild.
· The three files whose assessment diverges from a comparable file of the year, flagged with both sets of reasoning side by side. These are not errors: they are the three points where the committee gains by ruling explicitly, because its decision will be the reference for the next ones.
What the session becomes: 60 files examined in 2 hours instead of 3 h 30, and the minutes of the committee's opinions available the same evening, with the sense of each opinion and its reasoning.
What that gives back to the department: 4 committees a year, a week of preparation brought down to half a day — that is a month of officer time recovered over the year, and a report the public accountant can read without asking for anything more.
The next step I propose: that the committee's decisions feed the base of comparable files themselves, so that the next assessment rests on what the committee actually ruled and not on what people think it rules. Over the last three committees, that would have aligned 24 more files. committee-report_60-files.pdfA week of preparation brought down to half a day
⛓ Sourced · files assessed in the quarter, history of grants awarded over 3 years, minutes of the last 4 committees
What the report contains, file by file: the purpose of the application, the amount requested, the criteria met and not met, the amount proposed and the reasoning in three lines — always in the same order, which lets committee members read 60 files without losing their way.
What I add and that did not exist:
· A reminder of the grants already awarded to the same applicant by the authority over three years, across all schemes. It is the question asked at every committee, and nobody had the time to prepare it.
· The breakdown of proposals by member municipality and by type of applicant — your elected members ask for it one time in two, and it took half a day to rebuild.
· The three files whose assessment diverges from a comparable file of the year, flagged with both sets of reasoning side by side. These are not errors: they are the three points where the committee gains by ruling explicitly, because its decision will be the reference for the next ones.
What the session becomes: 60 files examined in 2 hours instead of 3 h 30, and the minutes of the committee's opinions available the same evening, with the sense of each opinion and its reasoning.
What that gives back to the department: 4 committees a year, a week of preparation brought down to half a day — that is a month of officer time recovered over the year, and a report the public accountant can read without asking for anything more.
The next step I propose: that the committee's decisions feed the base of comparable files themselves, so that the next assessment rests on what the committee actually ruled and not on what people think it rules. Over the last three committees, that would have aligned 24 more files. committee-report_60-files.pdfA week of preparation brought down to half a day
⛓ Sourced · files assessed in the quarter, history of grants awarded over 3 years, minutes of the last 4 committees
We have nine automatic checks inherited from an old system. They throw out many alerts and our two controllers confirm almost none of them.
I ran them across your last twelve months, one by one, and I am putting three rewritten ones on the table, already measured — you choose on figures, not on a hunch.
An automatic check is a written rule that flags a file for examination: a duplicate, an inconsistency of dates, an amount outside the scheme's bounds.
What your nine checks produced over twelve months: 1,180 alerts, 214 confirmed after examination — 18 %. In other words, 966 applicants had their file delayed by an examination that found nothing, and your two controllers spent on it the time the real cases deserved.
The detail that decides: your check no. 4 — a gap between the applicant's address and the one on the last file — produced 430 alerts for 12 confirmations. One alert in thirty-six. The cause is measured, not assumed: it fires on any house move and on any change of an association's registered office, whereas the situation it targets assumes two active addresses at the same time.
What I do on top, and nobody has the time to do: I have written the tightened version of the check and run it over the same twelve months. 96 alerts instead of 430, all 12 confirmations kept, and 334 applicants spared a pointless examination. Two other checks are rewritten the same way, each with the number of alerts it would have produced, the share confirmed, and the files it would have let through — that last figure is the one everyone forgets, and it is the only one that says what a check really costs.
And what the control file now contains: the exact line that triggered the alert, the two documents alongside and the applicant's history — an examination takes 6 minutes instead of 21.
Bringing a check into force stays with the department: a check applies only once signed, and that is precisely what makes it open to challenge by an applicant. I save you the writing and the measuring; the decision takes ten minutes instead of a committee — sign them and they are live tonight. automatic-checks_9-reviewed-3-rewritten.pdf430 alerts brought down to 96, the 12 confirmations kept
⛓ Sourced · the department's 9 automatic checks, 12 months of alerts and outcomes, 3 checks rewritten and replayed
An automatic check is a written rule that flags a file for examination: a duplicate, an inconsistency of dates, an amount outside the scheme's bounds.
What your nine checks produced over twelve months: 1,180 alerts, 214 confirmed after examination — 18 %. In other words, 966 applicants had their file delayed by an examination that found nothing, and your two controllers spent on it the time the real cases deserved.
The detail that decides: your check no. 4 — a gap between the applicant's address and the one on the last file — produced 430 alerts for 12 confirmations. One alert in thirty-six. The cause is measured, not assumed: it fires on any house move and on any change of an association's registered office, whereas the situation it targets assumes two active addresses at the same time.
What I do on top, and nobody has the time to do: I have written the tightened version of the check and run it over the same twelve months. 96 alerts instead of 430, all 12 confirmations kept, and 334 applicants spared a pointless examination. Two other checks are rewritten the same way, each with the number of alerts it would have produced, the share confirmed, and the files it would have let through — that last figure is the one everyone forgets, and it is the only one that says what a check really costs.
And what the control file now contains: the exact line that triggered the alert, the two documents alongside and the applicant's history — an examination takes 6 minutes instead of 21.
Bringing a check into force stays with the department: a check applies only once signed, and that is precisely what makes it open to challenge by an applicant. I save you the writing and the measuring; the decision takes ten minutes instead of a committee — sign them and they are live tonight. automatic-checks_9-reviewed-3-rewritten.pdf430 alerts brought down to 96, the 12 confirmations kept
⛓ Sourced · the department's 9 automatic checks, 12 months of alerts and outcomes, 3 checks rewritten and replayed
What worries me most is double funding: the same invoice presented to two of our schemes. We have never had the means to check.
It is done, over three years: 47 files last year had the same expenditure funded by two of your schemes, for €128,400 — and all 47 are assembled, with the documents.
How the matching is done: I compare the quotations and invoices filed from one file to another, across all schemes, on the amount, the date, the issuer and the description of the service — and not on the file name, which never says anything.
What the matching gives, and the detail matters:
· 47 files concerned, €128,400 of expenditure presented twice.
· Of those 47, 31 stem from an applicant's mistake — two counters of the same authority, a rulebook that does not clearly say the schemes do not combine on the same expenditure. I flag it because the most effective correction lies there: the missing sentence in the leaflet, which I have written.
· 11 are a permitted and perfectly regular combination — those I set aside myself, telling you which of your rulebooks allows it.
· 5 call for examination, and they are handed over with both files, both invoices and the chronology. The decision to withdraw a grant or to ask for repayment belongs to the competent authority, and what I hand over is a file that stands up after five minutes of reading before the public accountant.
What that represents, in the terms that matter: €128,400 of public spending funding 47 more projects instead of funding 47 twice — and a check that covers the 6,400 files, not the ones there was time to look at.
The next step I propose: that the missing sentence go into the 21 leaflets, and that the matching run at every filing rather than once a year. Over the trial quarter, 9 double presentations were detected at filing and corrected by the applicant themselves before assessment — no letter, no control, no dispute. double-funding_47-files-128400-euros.pdf31 applicant mistakes, 11 regular combinations set aside, 5 to examine
⛓ Sourced · quotations and invoices of the last 3 years, all schemes, scheme rules and combination clauses
How the matching is done: I compare the quotations and invoices filed from one file to another, across all schemes, on the amount, the date, the issuer and the description of the service — and not on the file name, which never says anything.
What the matching gives, and the detail matters:
· 47 files concerned, €128,400 of expenditure presented twice.
· Of those 47, 31 stem from an applicant's mistake — two counters of the same authority, a rulebook that does not clearly say the schemes do not combine on the same expenditure. I flag it because the most effective correction lies there: the missing sentence in the leaflet, which I have written.
· 11 are a permitted and perfectly regular combination — those I set aside myself, telling you which of your rulebooks allows it.
· 5 call for examination, and they are handed over with both files, both invoices and the chronology. The decision to withdraw a grant or to ask for repayment belongs to the competent authority, and what I hand over is a file that stands up after five minutes of reading before the public accountant.
What that represents, in the terms that matter: €128,400 of public spending funding 47 more projects instead of funding 47 twice — and a check that covers the 6,400 files, not the ones there was time to look at.
The next step I propose: that the missing sentence go into the 21 leaflets, and that the matching run at every filing rather than once a year. Over the trial quarter, 9 double presentations were detected at filing and corrected by the applicant themselves before assessment — no letter, no control, no dispute. double-funding_47-files-128400-euros.pdf31 applicant mistakes, 11 regular combinations set aside, 5 to examine
⛓ Sourced · quotations and invoices of the last 3 years, all schemes, scheme rules and combination clauses
We also assess two social schemes for residents — heating and school meals. Those are no longer grants, they are people's entitlements.
And that is precisely where the chain tightens: the same mechanics apply, with one more requirement on who decides and on what the applicant can challenge.
What I do on these two schemes:
· The income conditions are ticked against the documents, each value with its tax assessment and page, and the scale applied line by line with its effective date.
· The amount proposed is calculated in plain terms — no aggregated total, each line stands on its own.
· The draft notice is written in plain language, with the reasoning, the routes and time limits for appeal and a note of the right to obtain an explanation of the calculation.
What never changes, and it is what gives the entitlement its value: awarding, refusing, revising or withdrawing a social benefit has effects on a person — the decision is taken by the competent public official, reasoned, dated and open to challenge. I hand it over in 9 minutes instead of 40, complete file and reasoning written.
What that changed, measured over the year:
· 1,100 social applications assessed, average time down from 52 days to 21.
· Files complete at filing: 44 % → 89 %, because the leaflet finally says which tax assessment is expected.
· And on the heating grant, applied for in autumn: 214 households received their notice before 15 November against 61 last year — for a household waiting on that grant, this is not a management indicator, it is a bill they can pay.
The next step I propose: that these two schemes be assessed together with your benefits department rather than alongside it — the documents are the same in 8 files out of 10, and an applicant provides them twice today. The alignment is written, it fits on one page, and it gives you back 340 hours a year while removing one procedure for 1,100 households. social-schemes_1100-applications-52-days-down-to-21.pdf89 % of files complete at filing, 214 notices before 15 November
⛓ Sourced · rules and scales of the 2 social schemes, 1,100 applications assessed, notification log
What I do on these two schemes:
· The income conditions are ticked against the documents, each value with its tax assessment and page, and the scale applied line by line with its effective date.
· The amount proposed is calculated in plain terms — no aggregated total, each line stands on its own.
· The draft notice is written in plain language, with the reasoning, the routes and time limits for appeal and a note of the right to obtain an explanation of the calculation.
What never changes, and it is what gives the entitlement its value: awarding, refusing, revising or withdrawing a social benefit has effects on a person — the decision is taken by the competent public official, reasoned, dated and open to challenge. I hand it over in 9 minutes instead of 40, complete file and reasoning written.
What that changed, measured over the year:
· 1,100 social applications assessed, average time down from 52 days to 21.
· Files complete at filing: 44 % → 89 %, because the leaflet finally says which tax assessment is expected.
· And on the heating grant, applied for in autumn: 214 households received their notice before 15 November against 61 last year — for a household waiting on that grant, this is not a management indicator, it is a bill they can pay.
The next step I propose: that these two schemes be assessed together with your benefits department rather than alongside it — the documents are the same in 8 files out of 10, and an applicant provides them twice today. The alignment is written, it fits on one page, and it gives you back 340 hours a year while removing one procedure for 1,100 households. social-schemes_1100-applications-52-days-down-to-21.pdf89 % of files complete at filing, 214 notices before 15 November
⛓ Sourced · rules and scales of the 2 social schemes, 1,100 applications assessed, notification log
Our five reception officers take 18,000 questions a year, and the same ones come back constantly: where is my file, which documents, which criteria.
18,000 questions, of which 12,600 on thirteen subjects, always the same — the thirteen answers are written, sourced on your rules and dated, and they are waiting for you.
The thirteen subjects, drawn from your own exchanges: how far a file has got, documents to provide, criteria of a scheme, ceiling and amount, assessment time, date of the next committee, payment date, documents for the balance, reason for a refusal, routes of appeal, online filing, combination with another grant, and who one is speaking to.
What I propose, and you keep the key: I answer at any hour, Saturdays included, and I say in my first sentence that I am the urban community's digital assistant, not an officer. This is not an option you could switch off: the European regulation on artificial intelligence requires that anyone interacting with an AI system be informed, and the applicant can ask for an officer at any moment — I take their details and leave a dated call-back.
On the status of a file, the answer is the file's own: stage reached, document awaited if one is missing, date of the last request for documents, committee it is listed for. I read your assessment software, not a reception promise.
What is accessible to everyone, and not only to those who can read administrative language:
· Every answer and every leaflet exists in an easy-read version — “easy to read and understand”: short sentences, one message per sentence, everyday words rather than administrative ones. It is the applicant who chooses their version.
· The draft easy-read version is submitted to the head of service, who dates and settles it — a plain-language version is approved, it is not inferred, and it is the same dated version that serves the portal, the phone, the front desk and the letter.
· On the filing platform, I guide screen by screen, with the documents to prepare BEFORE starting — abandonment mid-filing goes from 41 % to 15 %, and that is where equal access to entitlements really plays out: a grant you cannot manage to apply for is a grant that does not exist.
The gain, measured: answering a recurring question goes from 10 % to 6 % of the time an enquiry takes — from 10 minutes to 6, and across 18,000 enquiries a year, that is 1,200 hours given back to reception and to the officers.
The next step I propose: that you reread the thirteen answers tomorrow, one by one — half an hour. As soon as they are approved, the department answers that same night, and I hand you every morning the one-page record of what went out. information_18000-questions-13-subjects.pdf13 answers written, approved easy-read, abandonment from 41 % to 15 %
⛓ Sourced · enquiry log over 12 months, rules of the 21 schemes, interpreter requests over 3 years, online filing log
The thirteen subjects, drawn from your own exchanges: how far a file has got, documents to provide, criteria of a scheme, ceiling and amount, assessment time, date of the next committee, payment date, documents for the balance, reason for a refusal, routes of appeal, online filing, combination with another grant, and who one is speaking to.
What I propose, and you keep the key: I answer at any hour, Saturdays included, and I say in my first sentence that I am the urban community's digital assistant, not an officer. This is not an option you could switch off: the European regulation on artificial intelligence requires that anyone interacting with an AI system be informed, and the applicant can ask for an officer at any moment — I take their details and leave a dated call-back.
On the status of a file, the answer is the file's own: stage reached, document awaited if one is missing, date of the last request for documents, committee it is listed for. I read your assessment software, not a reception promise.
What is accessible to everyone, and not only to those who can read administrative language:
· Every answer and every leaflet exists in an easy-read version — “easy to read and understand”: short sentences, one message per sentence, everyday words rather than administrative ones. It is the applicant who chooses their version.
· The draft easy-read version is submitted to the head of service, who dates and settles it — a plain-language version is approved, it is not inferred, and it is the same dated version that serves the portal, the phone, the front desk and the letter.
· On the filing platform, I guide screen by screen, with the documents to prepare BEFORE starting — abandonment mid-filing goes from 41 % to 15 %, and that is where equal access to entitlements really plays out: a grant you cannot manage to apply for is a grant that does not exist.
The gain, measured: answering a recurring question goes from 10 % to 6 % of the time an enquiry takes — from 10 minutes to 6, and across 18,000 enquiries a year, that is 1,200 hours given back to reception and to the officers.
The next step I propose: that you reread the thirteen answers tomorrow, one by one — half an hour. As soon as they are approved, the department answers that same night, and I hand you every morning the one-page record of what went out. information_18000-questions-13-subjects.pdf13 answers written, approved easy-read, abandonment from 41 % to 15 %
⛓ Sourced · enquiry log over 12 months, rules of the 21 schemes, interpreter requests over 3 years, online filing log
And those who never file? Our elected members tell us some grants do not find their public, and we do not know what to answer.
You can answer them with a figure: 890 applicants already known to the authority meet the conditions of a grant they have never claimed — and 431 have obtained it since.
Non-take-up is when a person or an association meets the conditions of an entitlement without ever claiming it — because they do not know, because the process discouraged them, or because they believed they were not entitled.
What I did, with your data alone: I set the applicants already known — those who have filed at least once in three years — against the conditions of the 21 schemes. 890 situations, of which 610 households eligible for the renovation grant and 280 associations eligible for the running grant.
What the campaign produced, once approved by your management: 890 letters in plain language, with the estimated amount, the three documents to enclose and the direct filing link; 520 files filed; 431 grants awarded after assessment by your officers and decision by the competent authority. That is 431 projects funded that your rules made eligible from the start.
And what that tells your elected members, which is the argument for the session: a scheme that does not find its public is not necessarily a bad scheme — it is often a scheme its public does not know about. On the renovation grant, the filing rate among eligible households went from 31 % to 58 % in a year, without a single criterion being changed.
What it costs in officer time, since the question will come: the 520 files of the campaign arrived complete in 91 % of cases — because the letter named the documents — and they took 12 minutes of pre-assessment each instead of 45.
The next step I propose: that detection run every quarter and that letters go out in waves of 200, so that your twelve officers absorb the filings it generates. The schedule is written, the first wave is ready — give me your approval and it goes out on Monday. non-take-up_890-situations-431-grants-awarded.pdf890 detected, 520 filings, 431 grants awarded
⛓ Sourced · applicants known over 3 years, conditions of the 21 schemes, campaign and filing log
Non-take-up is when a person or an association meets the conditions of an entitlement without ever claiming it — because they do not know, because the process discouraged them, or because they believed they were not entitled.
What I did, with your data alone: I set the applicants already known — those who have filed at least once in three years — against the conditions of the 21 schemes. 890 situations, of which 610 households eligible for the renovation grant and 280 associations eligible for the running grant.
What the campaign produced, once approved by your management: 890 letters in plain language, with the estimated amount, the three documents to enclose and the direct filing link; 520 files filed; 431 grants awarded after assessment by your officers and decision by the competent authority. That is 431 projects funded that your rules made eligible from the start.
And what that tells your elected members, which is the argument for the session: a scheme that does not find its public is not necessarily a bad scheme — it is often a scheme its public does not know about. On the renovation grant, the filing rate among eligible households went from 31 % to 58 % in a year, without a single criterion being changed.
What it costs in officer time, since the question will come: the 520 files of the campaign arrived complete in 91 % of cases — because the letter named the documents — and they took 12 minutes of pre-assessment each instead of 45.
The next step I propose: that detection run every quarter and that letters go out in waves of 200, so that your twelve officers absorb the filings it generates. The schedule is written, the first wave is ready — give me your approval and it goes out on Monday. non-take-up_890-situations-431-grants-awarded.pdf890 detected, 520 filings, 431 grants awarded
⛓ Sourced · applicants known over 3 years, conditions of the 21 schemes, campaign and filing log
Last set of questions, and they are the community council's: what has this given us, where do you get it wrong, what do you do alone, and what if we want to stop?
6,533 officer hours given back to the department, an error rate I publish myself, three automatic acts withdrawable with one word, and an exit that can be tested before the vote.
The calculation, item by item, so you can redo it:
· Pre-assessment: 6,400 files, 45 minutes down to 12 — 45 % → 12 % — that is 3,520 hours.
· Completeness check: 6,400 files, 25 minutes down to 8 — 25 % → 8 % — that is 1,813 hours.
· Recurring questions: 18,000 a year, 10 minutes down to 6 — 10 % → 6 % — that is 1,200 hours.
What those hours are, and it is what defends best in front of elected members: officer time given back to the department, at unchanged headcount — no post cut, no post created. This is not a staffing saving: it is public spending that assesses more files, faster and more equally, for the same running budget.
What they became, according to your own records: average assessment time 71 days → 38 · the rules' two-month deadline met 58 times out of 100 → 96 · files complete at filing 48 % → 86 % · notification of a decision 24 days → 3 · informal appeals 14 % → 5 % · interpretation gap between officers 11 % → 3 % · €128,400 of public spending funding 47 more projects · and 431 grants awarded to applicants who had never claimed them.
The figure that does not flatter me, and I publish it with the rest: 118 pre-assessment notes corrected on substance out of 1,460 in the first quarter — 8.1 %. The cause is measured: 84 of the 118 concerned three schemes whose rules had been amended by resolution without the version in force being carried into my sources. What I did with it: every rulebook now carries its effective date and the number of the resolution that sets it, the updated version is written overnight after the session, and a note produced on a version earlier than the file is blocked and comes back to a named officer. The following quarter: 17 corrections out of 1,520 — 1.1 %, and the 17 are situations without precedent, no expired rulebook left.
The three things I do without being asked: I check the completeness of every file filed, overnight after filing — and a document removed from a file disappears from my indexes at the same hour; I write the updated version of any rulebook a resolution amends, the day after the session; I hand the head of department the weekly record every Monday. Everything else waits for a request.
And the decisions that stay with the public official: awarding, refusing, withdrawing a grant, asking for repayment — four decisions with effects on an applicant, reasoned, traced and open to challenge. Across 6,400 files, they were taken 6,400 times.
On leaving: my indexes are deleted and contained none of your files; the log is handed to you in an open format or destroyed, as you choose; the 21 restored and dated rulebooks, the 13 answers, the easy-read versions and the rewritten checks stay with the urban community, readable without us; and since there was no migration on the way in, there is none on the way out. I propose a dry-run exit at the end of the first quarter — half a day: we switch off, we check that the department assesses exactly as before, we switch back on. The protocol is written and the least costly date is the first Friday of August: 11 filings that day against 26 on average. year-results_6533-hours-given-back.pdf45→12, 25→8, 10→6, and the calculation redoable on one side substantive-corrections_118-then-17.pdf8.1 % → 1.1 %, measured cause, dated rulebooks what-the-agent-does-alone_and-reversibility.pdf3 reversible acts, 4 decisions that stay with the public official
⛓ Sourced · log of assessed files and their corrections, times and appeals, non-take-up campaign, dry-run exit protocol
The calculation, item by item, so you can redo it:
· Pre-assessment: 6,400 files, 45 minutes down to 12 — 45 % → 12 % — that is 3,520 hours.
· Completeness check: 6,400 files, 25 minutes down to 8 — 25 % → 8 % — that is 1,813 hours.
· Recurring questions: 18,000 a year, 10 minutes down to 6 — 10 % → 6 % — that is 1,200 hours.
What those hours are, and it is what defends best in front of elected members: officer time given back to the department, at unchanged headcount — no post cut, no post created. This is not a staffing saving: it is public spending that assesses more files, faster and more equally, for the same running budget.
What they became, according to your own records: average assessment time 71 days → 38 · the rules' two-month deadline met 58 times out of 100 → 96 · files complete at filing 48 % → 86 % · notification of a decision 24 days → 3 · informal appeals 14 % → 5 % · interpretation gap between officers 11 % → 3 % · €128,400 of public spending funding 47 more projects · and 431 grants awarded to applicants who had never claimed them.
The figure that does not flatter me, and I publish it with the rest: 118 pre-assessment notes corrected on substance out of 1,460 in the first quarter — 8.1 %. The cause is measured: 84 of the 118 concerned three schemes whose rules had been amended by resolution without the version in force being carried into my sources. What I did with it: every rulebook now carries its effective date and the number of the resolution that sets it, the updated version is written overnight after the session, and a note produced on a version earlier than the file is blocked and comes back to a named officer. The following quarter: 17 corrections out of 1,520 — 1.1 %, and the 17 are situations without precedent, no expired rulebook left.
The three things I do without being asked: I check the completeness of every file filed, overnight after filing — and a document removed from a file disappears from my indexes at the same hour; I write the updated version of any rulebook a resolution amends, the day after the session; I hand the head of department the weekly record every Monday. Everything else waits for a request.
And the decisions that stay with the public official: awarding, refusing, withdrawing a grant, asking for repayment — four decisions with effects on an applicant, reasoned, traced and open to challenge. Across 6,400 files, they were taken 6,400 times.
On leaving: my indexes are deleted and contained none of your files; the log is handed to you in an open format or destroyed, as you choose; the 21 restored and dated rulebooks, the 13 answers, the easy-read versions and the rewritten checks stay with the urban community, readable without us; and since there was no migration on the way in, there is none on the way out. I propose a dry-run exit at the end of the first quarter — half a day: we switch off, we check that the department assesses exactly as before, we switch back on. The protocol is written and the least costly date is the first Friday of August: 11 filings that day against 26 on average. year-results_6533-hours-given-back.pdf45→12, 25→8, 10→6, and the calculation redoable on one side substantive-corrections_118-then-17.pdf8.1 % → 1.1 %, measured cause, dated rulebooks what-the-agent-does-alone_and-reversibility.pdf3 reversible acts, 4 decisions that stay with the public official
⛓ Sourced · log of assessed files and their corrections, times and appeals, non-take-up campaign, dry-run exit protocol
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 an assessment department
Each use corresponds to an agent we deploy. All of them work in support, subject to your approval — the AI agent assists, the public officer decides.
Included in your agent
The 6 capabilities essential to this promise are included, at no extra cost.
From 1,005 € incl. VAT / month
Check the documents required by the grant scheme's rules
OCR and extraction of a file's documents (statutes, bank details, notices, budgets), completeness check against the required list.
Completeness & consistency checks
Cross-checking amounts, dates, identifiers and ceilings; flagging inconsistencies before assessment.
Preliminary assessment against your rules
Testing every application against the eligibility criteria of the funding rules, with a reasoned preliminary assessment note for approval.
Fraud & anomaly detection
Spotting duplicates, inconsistent declarations and atypical signals to make grant awards more reliable.
Decision letters & document requests
Preparing draft letters (approval, reasoned refusal, request for additional documents) from your templates, for approval.
Plain language & easy-read draft
Preparation of a plain-language version and of a draft easy-read (FALC) transcription, submitted for human validation.
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.
Assessing benefits & entitlements
Assessing benefits and entitlements belongs to another agent — the AI agent for benefits caseworkers — and is scoped separately with you. It is therefore part of neither this agent's scope nor its price.
Being architected — not orderable Be notified when it becomes available →Accessibility, easy-read & translation
Part of the public remains cut off from public information: people who do not speak French, people with a cognitive disability, people uncomfortable with the written word.
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.
Administrative information
What this agent prepares — the status of the file and the list of documents expected — is handed to the Multichannel administrative information agent, which answers the applicant on their own channels. The scope of this page stops at pre-assessment; human validation stays with the public officer.
Multichannel administrative information from 700 € incl. VAT / month Discover the agent →Help with online procedures
What this agent prepares — the list of missing documents — is handed to the Online procedures support agent, which guides the applicant through filing and completing the application. The scope of this page stops at the filed application; human validation stays with the public officer.
Support with online procedures from 780 € incl. VAT / month Discover the agent →Pre-assessment of social assistance (local welfare centre)
Strictly in support (administrative). Award decision by a human.
Pre-assessment of social assistance (local welfare centre) from 1,050 € incl. VAT / month Discover the agent →
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 an assessment department win back?
By automating document reading, completeness checks and the first test against the criteria, a department can aim for a reduction by half in preliminary assessment time on standardised files — reinvested in complex cases and in the quality of the reasoning.
Preliminary assessment of a grant application
Checking the completeness of a file
Answering a recurring question from a member of the public
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
An assessment agent (completeness, consistency, preliminary assessment, letters), installed and operated for you. Choose according to how you work. Prices exclude VAT — annual subscription, the time it takes for the gains to settle in.
Agility
Setup + controlled subscription
11,270 € incl. VAT setup
then 1,005 € 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
1,635 € 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
16,191 € incl. VAT setup
then 1,270 € 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 an assessment department
Applicants' data never leaves the departmentLocal inference or an isolated resource hosted in France; no document from a file entrusted to a foreign third party.
Data in France, under French lawFor applicants' data: hosting under French law, native minimisation, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The public officer keeps the decisionThe agent produces verifiable checks, preliminary assessments and draft letters; no decision on entitlement is automated.
Human oversight & traceabilityReading supporting documents: monitoring, updates and logging, compliant with the AI Act and with equal treatment of applicants.
Frequently asked questions
Your questions, our answers
Does the agent award the grant?
No: it prepares. The preliminary assessment, the checks and the draft letters are produced by the agent, but the decision rests with the case officer and the competent authority. No decision on entitlement is automated.
Does it adapt to our local schemes?
Yes: the agent is configured on your funding rules, your eligibility criteria, the documents you require and your letter templates. Each scheme has its own checking logic.
What is the framing under the AI Act?
Strictly in support, with reinforced human oversight and no automated decision about people's entitlements. The agent assists the assessment and records its analyses; the public officer decides and gives the reasons.
Is applicants' data protected?
Yes. The documents in the files contain personal, social and financial data: our agents run locally within the authority or are hosted in France, with the deployment objective of processing and access operated within the European Union and an architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
Does the agent help guarantee equal treatment?
Yes. By applying the same checks and the same criteria to every file, and by recording its analyses, the agent contributes to consistent and reasoned assessment. The decision and the judgement behind it remain human.
Do we have to change our assessment tools?
No. The agent connects to your existing tools (submission platform, document management, assessment software) and complements them, without imposing a migration. We adapt the integration to your environment.
How long does it take to deploy an agent?
A few weeks as a rule, after a free audit that identifies the scheme with the best return to equip, then a phase of configuring it on your rules, integrating and testing before going live and training the case officers.
Related roles
Other roles in assessment and public administration
Let's talk
Let us estimate the potential in your assessment department
15 minutes to identify the scheme with the best return to equip — hosted in France, supervised, with no commitment.