We build Agents. You build the Future.
● Public sector — Reception & informationThe AI agent for front-desk officers: inform the public, free up the officer
Repetitive questions about procedures — opening hours, documents to bring, which department to contact — take up a considerable share of front-desk time without being the heart of the service: supporting the situations that deserve it. Your AI agent absorbs that flow, at the counter and online alike. Hosted in France — on local inference or an isolated resource — the public's data stays under control. The public officer keeps the lead, and always takes the conversation back when needed.
Updated on
Deployed in a few weeks
Reception assistant · hosted in France
Fictional simulation — not a client result
A resident is asking how to renew her identity card and whether she can do it here. What do I tell her?
Identity card renewal: the application is submitted by appointment at a town hall equipped with a biometric collection station — yours is. Documents to bring: the old card, proof of address less than a year old, a compliant photo, and the online pre-application (keep the reference number).
I can propose three time slots and generate the list of documents to hand to her.
⛓ Source · service-public.fr + the authority's own procedures
I can propose three time slots and generate the list of documents to hand to her.
⛓ Source · service-public.fr + the authority's own procedures
She does not speak French well and has not done the pre-application. Can you help her?
Yes. I restate each step in plain language, for you to validate; translation is where the neighbouring agent takes over. For the pre-application: I guide the entry field by field and flag missing documents before submission.
If the situation gets complicated (incomplete document, doubt about civil status), I hand the file to the front-desk officer with the summary already filled in.
✎ Action · pre-application prepared — the public officer approves and receives the resident
If the situation gets complicated (incomplete document, doubt about civil status), I hand the file to the front-desk officer with the summary already filled in.
✎ Action · pre-application prepared — the public officer approves and receives the resident
Data hosted in France · sovereign model as an option
Data hosted in France
Hosted in FranceAI model called at its publisher, except with the Sovereign model option
The public's data protected
GDPR & AI Act: governed deploymentTraceability & human oversight
TurnkeyDesigned, installed and operated for you
The public officer decidesThe AI agent carries out what you have authorised
For an organisation established outside the European Union, the law of its own country applies; at its request, we put equivalent safeguards in place (human review before any action, a log of the criteria, information for the people concerned).
✦ In brief
In a reception and information service, a Blue Lemon Agent agent answers the public's everyday questions about procedures — at the counter and online, 24/7 — drawing on reliable sources (such as service-public.fr and your own procedures) which it cites rather than invents. It routes people to the right department or online service and hands over to a public officer as soon as the situation calls for it. Data hosted in France; the AI model is called at its publisher, except with the Sovereign model option. With that option, the public's data is never exposed to a foreign service, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity. Users know they are talking to an AI and no decision about their rights is taken by an agent. Live within a few weeks. Your public-sector staff write to it from Microsoft Teams, Slack or their email, and your 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 set out in your agent's quote.
100%
hosted in France in the target architecture
0
transfer outside the EU with the Sovereign model option (target architecture)
2
uses ready to deploy on this scope
0
decision with a legal or significant effect taken by an agent
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.
Indicative estimate, to be confirmed by a pilot on your real files (general terms of sale, article 4).
Pricing
Three options, one agent
A reception and information agent (multichannel, sourced), 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
2,150 € incl. VAT setup
then 330 € 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
- Hosting in France, a dedicated and isolated resource
The simplest
Serenity
All inclusive, no setup fee
450 € 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
- Hosting in France, managed end to end
100% Sovereign
On site, you own it
4,221 € incl. VAT setup
then 445 € 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.
- Hardware installed on your premises (you own it)
- French / European AI models run locally
- Secure remote maintenance (Pro support included)
Included in the package: hosting and AI consumption up to the volume of your capacity tier, in documents, exchanges, tokens and storage; beyond that, re-invoiced at real cost with no margin, tracked in your client area as usage statements are issued. 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 for the package chosen, in France by default, 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.
Use cases
The uses of AI for reception and information
What the agent does in reception and information, and the neighbouring agents we deploy for the rest. The agent carries out by itself what you authorise it to do; you decide what it does on its own and what it submits to you. The public officer always takes the conversation back when needed.
Included in your agent
The 2 capabilities essential to this promise are included, at no extra cost.
From 330 € incl. VAT / month
Multichannel information
Answering everyday questions about procedures, 24/7, citing reliable sources — at the counter and by chat.
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 5 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
Offer accessibility, easy read, multichannel access and genuine human recourse
Measure resolution, response time, non-take-up, drop-offs and satisfaction
Other needs our agents cover
Each card says how the matching agent can be obtained today: available, on quote, or not yet in the catalogue.
First-line answers on benefits
Handling 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.
Not yet in the catalogue — not orderable Be notified when it becomes available →Hospital reception and admissions (admissions office)
Strictly in support (administrative). Health data: HDS requirement settled at scoping.
On quote View the agent page →Translating answers and letters
What this agent prepares — the sourced answer and its plain-language version — is handed to the Accessibility and inclusion agent, which produces the translation and the draft easy-read (FALC) transcription. Human validation stays with the public officer before anything reaches the user.
Available 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.
Help with online procedures
What this agent prepares — the sourced answer and the routing of the user — is handed to the Online procedures support agent, which guides the form filling step by step and gathers the documents. The scope of this page stops at information; human validation stays with the public officer.
Support with online procedures from 405 € incl. VAT / month Discover the agent →Phone reception
Absorbing the switchboard: answering everyday calls, transferring to the right officer and avoiding queues at peak times.
Switchboard voice agent from 505 € incl. VAT / month Switchboard →Multi-agency France Services reception
Strictly in support (administrative). Information and routing.
Multi-agency France Services reception from 400 € incl. VAT / month Discover the agent →Information and cataloguing support in libraries
Strictly in support (administrative and documentary).
Information and cataloguing support in libraries from 430 € incl. VAT / month Discover the agent →AI appointment-booking agent
Booking, rescheduling and confirming appointments, automatic reminders, redirection to the right department — online and at the counter alike.
Booking and managing appointments from 420 € incl. VAT / month Discover the agent →
Does your need fall outside this?
Build your agent
In 30 minutes we identify the agent that will give your staff the most time back — without oversizing the project.
Demonstration
See the agent at work
5 everyday 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 bodyTown of Pont-Sereine — municipality of 22,400 residents (fictional town)
- Sector
- Front-desk and resident services of a town of 22,400 — identity documents and civil status, school and after-school enrolment, resident parking, discretionary benefits, first-level planning enquiries; town hall open 39 hours a week
- Headcount
- 7 staff — 5 front-desk officers, one head of service and one apprentice; the same people hold all four channels: counter, telephone, e-mail and the website chat
- Public served
- 22,400 residents, 9,800 households — 41,200 enquiries a year: 23,400 phone calls, 14,300 counter visits, 3,500 e-mails
- Order of magnitude
- 15 recurring topics cover 27,600 of the 41,200 enquiries; 13,600 fall outside those 15 topics, including 9,500 vague requests to be routed; 4,100 online-service walkthroughs and 3,100 discretionary benefit claims a year; 8,400 certificates and acknowledgements signed at the counter
- Tools in place
- The town's website and its 62 procedure sheets, 6 online services, a resident relationship system, e-mail and departmental calendars — the agent plugs into them read-only, nothing is replaced and nothing is migrated
- Who decides
- The head of front-desk services approves every answer for as long as the mandate is unsigned; decisions on benefits belong to the caseworkers; the mayor delegates signature by order, for named acts, on the counter's certificates and acknowledgements
- Room for improvement
- 8,892 calls out of 23,400 ring unanswered at peak hours — 38 %; an e-mail waits 6 days on average; 31 % of residents abandon the resident-parking online service mid-form; and 3 of the 62 procedure sheets on the website carry information the town has since changed
In Pont-Sereine, seven staff hold four reception channels for 22,400 residents. The most repetitive questions — documents required, which department, which opening hours — take up two thirds of the flow, and the situations that deserve a human officer wait behind them. The exchanges below span a year, from the review of the flow to the report presented to the town council. The agent runs in local inference on a town machine, reads the procedure sheets and the enquiry log without ever writing to them, and no decision on a resident's entitlements is taken without a caseworker.
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.
Front-desk assistant · hosted in France
Fictional simulation — not a client result
I have reviewed your year's 41,200 enquiries — 23,400 calls, 14,300 counter visits, 3,500 e-mails — and the flow has a very clear shape that this count sets out for the first time.
What the count says: 15 topics cover 27,600 enquiries out of 41,200, two in every three. Identity documents, school and after-school enrolment, resident parking, civil status, canteen, bulky waste, cemetery, electoral roll, discretionary benefits, first-level planning, municipal hall, citizenship census, school transport, weekly market, stray animals. The 15 reference answers are written, sourced and dated — they are waiting for you.
What the three measurable items weigh, on your own logs:
· A routine information request takes 10 minutes end to end, of which 60 % goes on searching and drafting — 6 minutes.
· Walking someone through an online service takes 30 minutes, of which 40 % is the form-filling itself — 12 minutes.
· A vague request takes 20 minutes when it crosses two or three departments, of which 15 % is the routing itself — 3 minutes.
And the gap I found by matching your 62 procedure sheets against the decisions the town has published: 3 sheets carry information the town has since changed — the documents office has moved floor, the resident-parking rate has been revised, the summer opening hours have shifted by half an hour. Those 3 sheets fed 3,240 answers last year, and the 3 updated versions are written tonight, with the replaced line and the decision that replaces it side by side.
The next step, and it takes twenty minutes: you read the 3 rewritten sheets and the 3 reference answers that carry the most weight. The head of service settles the version, and it serves at the counter the same day. front-desk-flow_41200-enquiries-15-topics.pdf27,600 enquiries on 15 topics, the three items measured
⛓ Sourced · 12 months of the enquiry log, the website's 62 procedure sheets, decisions published by the town
What the count says: 15 topics cover 27,600 enquiries out of 41,200, two in every three. Identity documents, school and after-school enrolment, resident parking, civil status, canteen, bulky waste, cemetery, electoral roll, discretionary benefits, first-level planning, municipal hall, citizenship census, school transport, weekly market, stray animals. The 15 reference answers are written, sourced and dated — they are waiting for you.
What the three measurable items weigh, on your own logs:
· A routine information request takes 10 minutes end to end, of which 60 % goes on searching and drafting — 6 minutes.
· Walking someone through an online service takes 30 minutes, of which 40 % is the form-filling itself — 12 minutes.
· A vague request takes 20 minutes when it crosses two or three departments, of which 15 % is the routing itself — 3 minutes.
And the gap I found by matching your 62 procedure sheets against the decisions the town has published: 3 sheets carry information the town has since changed — the documents office has moved floor, the resident-parking rate has been revised, the summer opening hours have shifted by half an hour. Those 3 sheets fed 3,240 answers last year, and the 3 updated versions are written tonight, with the replaced line and the decision that replaces it side by side.
The next step, and it takes twenty minutes: you read the 3 rewritten sheets and the 3 reference answers that carry the most weight. The head of service settles the version, and it serves at the counter the same day. front-desk-flow_41200-enquiries-15-topics.pdf27,600 enquiries on 15 topics, the three items measured
⛓ Sourced · 12 months of the enquiry log, the website's 62 procedure sheets, decisions published by the town
There are seven of us and all four channels run at once. Nobody has half a day to read through procedure sheets. Where do we start?
With three topics out of fifteen — and I have ranked them by the volume they carry, not alphabetically.
What twelve months of counting says:
· Identity documents: 5,200 enquiries a year. It is the town's leading topic, and it is the sheet that still gives the old floor.
· School and after-school enrolment: 3,900.
· Resident parking: 2,800, and that is the sheet whose rate has been revised.
Those three topics make 11,900 of the 27,600 recurring enquiries: three topics out of fifteen, 43 % of the repetitive flow. The other twelve are read at four a week, and nothing waits.
And I advise against opening all fifteen at once, with the figures to back it: the first 3 take 20 minutes of reading and cover 43 % of the flow; all 15 together take 100 — half a day that would push the go-live back three weeks for 57 % more flow. Four topics a week brings the whole set into service within a month, and the first gain lands tomorrow morning.
What I bring you inside those twenty minutes: the three reference answers already written, each with the town decision it relies on and its date, the easy-read version and the date it was reviewed, and the number of enquiries each would have handled in the last year alone — 5,200, 3,900 and 2,800. You approve, or you correct with a word; the head of service settles the version. The drafting and the matching are done: the decision takes twenty minutes instead of half a day, and it stays the service's own — which is what makes it defensible before a resident who challenges it.
The next step I propose: that I match your 62 sheets against the town's published decisions every night, and write the updated version of any sheet a decision amends, in the night that follows its publication. Your sheets will stop ageing in silence, and this read-through will never have to be redone: it becomes a five-minute approval per decision. procedure-sheets_3-rewritten-of-62.pdf3,240 answers concerned, 3 updated versions written
⛓ Sourced · 12 months of enquiries counted by topic, 62 procedure sheets, 3 rewritten versions
What twelve months of counting says:
· Identity documents: 5,200 enquiries a year. It is the town's leading topic, and it is the sheet that still gives the old floor.
· School and after-school enrolment: 3,900.
· Resident parking: 2,800, and that is the sheet whose rate has been revised.
Those three topics make 11,900 of the 27,600 recurring enquiries: three topics out of fifteen, 43 % of the repetitive flow. The other twelve are read at four a week, and nothing waits.
And I advise against opening all fifteen at once, with the figures to back it: the first 3 take 20 minutes of reading and cover 43 % of the flow; all 15 together take 100 — half a day that would push the go-live back three weeks for 57 % more flow. Four topics a week brings the whole set into service within a month, and the first gain lands tomorrow morning.
What I bring you inside those twenty minutes: the three reference answers already written, each with the town decision it relies on and its date, the easy-read version and the date it was reviewed, and the number of enquiries each would have handled in the last year alone — 5,200, 3,900 and 2,800. You approve, or you correct with a word; the head of service settles the version. The drafting and the matching are done: the decision takes twenty minutes instead of half a day, and it stays the service's own — which is what makes it defensible before a resident who challenges it.
The next step I propose: that I match your 62 sheets against the town's published decisions every night, and write the updated version of any sheet a decision amends, in the night that follows its publication. Your sheets will stop ageing in silence, and this read-through will never have to be redone: it becomes a five-minute approval per decision. procedure-sheets_3-rewritten-of-62.pdf3,240 answers concerned, 3 updated versions written
⛓ Sourced · 12 months of enquiries counted by topic, 62 procedure sheets, 3 rewritten versions
All of this assumes you read what residents write to us: family situations, ongoing cases, sometimes difficult things. Where does that go?
With the 100% Sovereign package, nowhere. I run in local inference on a town machine, and what I read does not leave your network.
Local inference means the model computes on your own machine: the text of a resident's e-mail or of an ongoing case crosses no external network to be processed. If the town would rather not run a machine, the other route is an isolated resource hosted in France, dedicated to Pont-Sereine, with no pooling with any other authority.
What this changes, point by point:
· Resident data trains no model, neither ours nor a third party's.
· I work read-only, and the technical account I read through has no write permission — that can be checked with a single command, which is worth more than a promise.
· Encryption in transit and at rest, role-based access — rights follow the job: a front-desk officer opens the procedure sheets, only the caseworker opens the benefit files. 7 roles for your 7 staff, and the log shows 0 out-of-role accesses 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, including against a US player hosting in Europe.
· A complete log: who asked what, when, and what the system produced.
And the resident knows who they are speaking to: I introduce myself as a digital assistant of the town from the first sentence, because the European regulation on artificial intelligence requires informing anyone interacting with an AI system — and they can ask for a human officer at any time, in which case I pass the request on with the topic already noted.
Decisions on a resident's entitlements belong to the caseworker — and I hand them over in four minutes, complete file, documents checked and the calculation set out in plain language.
The figure that sums all of this up: 0 resident data out of the town's network across 41,200 enquiries, and processing in the EU targeted.
What I propose: that I keep up to date the record your data protection officer will ask for — hosting, data processed, retention periods, who accesses what. The first version is written and attached; it updates itself every time a new topic is opened. technical-framework_where-resident-data-lives.pdf100% Sovereign package: local 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 text of a resident's e-mail or of an ongoing case crosses no external network to be processed. If the town would rather not run a machine, the other route is an isolated resource hosted in France, dedicated to Pont-Sereine, with no pooling with any other authority.
What this changes, point by point:
· Resident data trains no model, neither ours nor a third party's.
· I work read-only, and the technical account I read through has no write permission — that can be checked with a single command, which is worth more than a promise.
· Encryption in transit and at rest, role-based access — rights follow the job: a front-desk officer opens the procedure sheets, only the caseworker opens the benefit files. 7 roles for your 7 staff, and the log shows 0 out-of-role accesses 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, including against a US player hosting in Europe.
· A complete log: who asked what, when, and what the system produced.
And the resident knows who they are speaking to: I introduce myself as a digital assistant of the town from the first sentence, because the European regulation on artificial intelligence requires informing anyone interacting with an AI system — and they can ask for a human officer at any time, in which case I pass the request on with the topic already noted.
Decisions on a resident's entitlements belong to the caseworker — and I hand them over in four minutes, complete file, documents checked and the calculation set out in plain language.
The figure that sums all of this up: 0 resident data out of the town's network across 41,200 enquiries, and processing in the EU targeted.
What I propose: that I keep up to date the record your data protection officer will ask for — hosting, data processed, retention periods, who accesses what. The first version is written and attached; it updates itself every time a new topic is opened. technical-framework_where-resident-data-lives.pdf100% Sovereign package: local inference, read-only, processing in the EU targeted
✎ Framework · deployment architecture, technical account permissions, first version of the processing record
Fictional simulation — not a client result
At peak hours the switchboard saturates. Residents call three times, or end up coming to the counter for a question that takes one sentence.
23,400 calls last year, 8,892 unanswered — 38 %, and all of them concentrated in four hours of the day. This is not a matter of reception quality: it is seven staff holding four channels at the same time.
What I have done with those 8,892 calls, after reading your call-backs and your e-mails: 16,100 of the 23,400 calls fall within the 15 topics, and the 15 answers are written, sourced and dated. The answer already existed somewhere: what was missing was availability, and that is exactly what I bring.
What I do, and you keep the key: I pick up at any hour, including Saturdays and the lunch break, I introduce myself as a digital assistant of the town from the first sentence, and the resident can ask for a human officer whenever they wish — I then take their number and leave you a dated call-back, topic already noted.
The rule I hold most firmly, and it is the one that protects the town: I answer within the municipality's own field of competence, and anything belonging to another counter leaves with that counter's name, address, opening hours and processing time. The resident leaves with a complete route.
The gain, measured on your volumes: answering a routine information request goes from 60 % to 8 % of the request's time — from 6 minutes to 48 seconds. Across 27,600 recurring enquiries a year, that is 2,392 hours going back to the situations that deserve a human officer. And unanswered calls go from 8,892 to 512: what remains are those who ask for a human officer, and they reach you with the topic already noted.
The next step I propose: that I start picking up tonight on the first three topics — documents, school enrolment, resident parking — that is 11,900 enquiries a year. You read the first night's log tomorrow morning, and you open the other twelve topics at whatever pace suits you. switchboard_23400-calls-8892-missed.pdf38 % → 2 %, 16,100 calls on the 15 topics
⛓ Sourced · 12 months of the switchboard log, 15 reference answers drafted, the municipality's field of competence
What I have done with those 8,892 calls, after reading your call-backs and your e-mails: 16,100 of the 23,400 calls fall within the 15 topics, and the 15 answers are written, sourced and dated. The answer already existed somewhere: what was missing was availability, and that is exactly what I bring.
What I do, and you keep the key: I pick up at any hour, including Saturdays and the lunch break, I introduce myself as a digital assistant of the town from the first sentence, and the resident can ask for a human officer whenever they wish — I then take their number and leave you a dated call-back, topic already noted.
The rule I hold most firmly, and it is the one that protects the town: I answer within the municipality's own field of competence, and anything belonging to another counter leaves with that counter's name, address, opening hours and processing time. The resident leaves with a complete route.
The gain, measured on your volumes: answering a routine information request goes from 60 % to 8 % of the request's time — from 6 minutes to 48 seconds. Across 27,600 recurring enquiries a year, that is 2,392 hours going back to the situations that deserve a human officer. And unanswered calls go from 8,892 to 512: what remains are those who ask for a human officer, and they reach you with the topic already noted.
The next step I propose: that I start picking up tonight on the first three topics — documents, school enrolment, resident parking — that is 11,900 enquiries a year. You read the first night's log tomorrow morning, and you open the other twelve topics at whatever pace suits you. switchboard_23400-calls-8892-missed.pdf38 % → 2 %, 16,100 calls on the 15 topics
⛓ Sourced · 12 months of the switchboard log, 15 reference answers drafted, the municipality's field of competence
Answering directly is quite another matter from preparing an answer I read first. I will not sign a blank cheque, and I cannot read 27,600 answers either.
You do not have to choose between the two: I propose a written mandate, capped, dated and withdrawable with a word — and it is already drafted and attached.
What it delivers first, because that is what decides: the average reply time on an e-mail goes from 6 days to 1, a request arriving on Friday evening stops waiting until Tuesday, and the 27,600 recurring enquiries are handled at any hour.
What the mandate says, and it fits in six lines:
· Exact scope: the 15 topics, listed by name, and nothing else. Any request outside that list reaches you with a drafted, sourced answer, ready to go out from you.
· Ongoing individual cases sit outside the mandate — a benefit claim under assessment, a civil-status situation: there, I hand you the answer already written, with the file's documents alongside, and you are the one who sends it. During the trial quarter those answers went out within 2 hours instead of 6 days — and they went out from an officer, which is what makes them reasoned and open to challenge.
· Every answer carries its source — the procedure sheet or the town decision, with its date — and a statement that it was prepared by a digital assistant of the town.
· You receive each morning the one-page log of the answers that went out the day before. An answer to be corrected is caught within the hour.
· Term: reviewed after three months, with the record of what has changed. Without an explicit decision at the review, the mandate stops — it is renewal that needs a signature.
· Withdrawal: a word from the head of service, and direct sending stops within the minute, answers reverting to drafts for approval.
The decision belongs to the town, and it is taken on a text already written, with one signature. The mandate and the public information note — the one to be displayed at the counter and published on the website — are both drafted. You sign, the service is live the next morning, and the review is already in your calendar on the 15th of the third month. direct-answer-mandate_15-topics-capped.pdf15 topics, individual cases handed over drafted, 3-month review
✎ Framework · drafted mandate, list of the 15 topics, public information note, log of outgoing answers
What it delivers first, because that is what decides: the average reply time on an e-mail goes from 6 days to 1, a request arriving on Friday evening stops waiting until Tuesday, and the 27,600 recurring enquiries are handled at any hour.
What the mandate says, and it fits in six lines:
· Exact scope: the 15 topics, listed by name, and nothing else. Any request outside that list reaches you with a drafted, sourced answer, ready to go out from you.
· Ongoing individual cases sit outside the mandate — a benefit claim under assessment, a civil-status situation: there, I hand you the answer already written, with the file's documents alongside, and you are the one who sends it. During the trial quarter those answers went out within 2 hours instead of 6 days — and they went out from an officer, which is what makes them reasoned and open to challenge.
· Every answer carries its source — the procedure sheet or the town decision, with its date — and a statement that it was prepared by a digital assistant of the town.
· You receive each morning the one-page log of the answers that went out the day before. An answer to be corrected is caught within the hour.
· Term: reviewed after three months, with the record of what has changed. Without an explicit decision at the review, the mandate stops — it is renewal that needs a signature.
· Withdrawal: a word from the head of service, and direct sending stops within the minute, answers reverting to drafts for approval.
The decision belongs to the town, and it is taken on a text already written, with one signature. The mandate and the public information note — the one to be displayed at the counter and published on the website — are both drafted. You sign, the service is live the next morning, and the review is already in your calendar on the 15th of the third month. direct-answer-mandate_15-topics-capped.pdf15 topics, individual cases handed over drafted, 3-month review
✎ Framework · drafted mandate, list of the 15 topics, public information note, log of outgoing answers
Then there are those who arrive without a precise question: “I received a letter from the town hall.” We pass them from department to department, and sometimes they leave with nothing.
9,500 vague requests a year, and I have written the routing tree that handles them with one more question, not three.
How I go about it: I start from what the resident has in hand — a letter, a reference, a date, a department name — and match it against the 62 procedure sheets and the departmental chart. The resident leaves with the right department, the floor, the opening hours, the documents to bring and, where one is needed, the appointment already booked.
What that gave over the trial quarter: 2,375 vague requests handled, 2,288 routed to the right department first time — 96 %. The remaining 87 concerned a competence outside the municipality, and they left with the counter's name, its address and its processing time, which was exactly what the resident was looking for.
The gain, in figures: routing to the right department goes from 15 % to 6 % of the request's time — from 3 minutes to 1 minute 12. Across 9,500 requests a year, that is 285 hours returned to the service, and as many counter-to-counter hand-offs removed for the resident.
On appointments I go all the way: I qualify the request, book the slot in the right department — only within the windows your colleagues have opened —, give the list of documents to bring and send a reminder the day before. Over the quarter, 1,420 appointments booked and 43 people arriving without their documents, that is 3 %, against 17 % last year — 199 wasted journeys spared to residents. An appointment with an elected member works differently: I pass the request on with the subject, the resident's file where one exists and three free slots; the member picks one with a word, and the resident has confirmation within the hour.
The next step I propose: that the routing tree be read by the five heads of department concerned — half an hour in all, one line each — and that it serve from the following week at the counter, on the phone and on the website chat, with the same dated version everywhere. routing_9500-vague-requests.pdf96 % to the right department first time, 285 hours appointment-booking_1420-booked-3-percent-without-documents.pdf17 % → 3 % arriving without documents, 199 journeys spared
⛓ Sourced · 9,500 vague requests from the log, departmental chart, appointment calendars, drafted routing tree
How I go about it: I start from what the resident has in hand — a letter, a reference, a date, a department name — and match it against the 62 procedure sheets and the departmental chart. The resident leaves with the right department, the floor, the opening hours, the documents to bring and, where one is needed, the appointment already booked.
What that gave over the trial quarter: 2,375 vague requests handled, 2,288 routed to the right department first time — 96 %. The remaining 87 concerned a competence outside the municipality, and they left with the counter's name, its address and its processing time, which was exactly what the resident was looking for.
The gain, in figures: routing to the right department goes from 15 % to 6 % of the request's time — from 3 minutes to 1 minute 12. Across 9,500 requests a year, that is 285 hours returned to the service, and as many counter-to-counter hand-offs removed for the resident.
On appointments I go all the way: I qualify the request, book the slot in the right department — only within the windows your colleagues have opened —, give the list of documents to bring and send a reminder the day before. Over the quarter, 1,420 appointments booked and 43 people arriving without their documents, that is 3 %, against 17 % last year — 199 wasted journeys spared to residents. An appointment with an elected member works differently: I pass the request on with the subject, the resident's file where one exists and three free slots; the member picks one with a word, and the resident has confirmation within the hour.
The next step I propose: that the routing tree be read by the five heads of department concerned — half an hour in all, one line each — and that it serve from the following week at the counter, on the phone and on the website chat, with the same dated version everywhere. routing_9500-vague-requests.pdf96 % to the right department first time, 285 hours appointment-booking_1420-booked-3-percent-without-documents.pdf17 % → 3 % arriving without documents, 199 journeys spared
⛓ Sourced · 9,500 vague requests from the log, departmental chart, appointment calendars, drafted routing tree
Fictional simulation — not a client result
Our online services look good on paper. In reality people start, stop halfway and come back to the counter.
I have gone back over the year's 4,100 walkthroughs, and the abandonment has a single, measurable cause, already dealt with.
An online service is a procedure residents complete themselves on the web. Non-take-up is the entitlement a person holds and does not obtain, for want of having reached the end of the procedure.
What the measurement says: 31 % of residents abandon the resident-parking online service mid-form, and 78 % of those abandonments happen on two screens: proof of address and vehicle registration. The abandonment is not technical: the person starts, a document is missing, they close the page and never come back.
What I have put in place, and it removes the cause:
· The list of documents to prepare is given BEFORE the first screen, on all six online services.
· I guide screen by screen, I read out what is being asked and why, and I flag the missing document before validation, not after.
· On the resident's personal account, I pre-fill the screen in front of them from the documents they have brought — civil status, address, file numbers, everything that gets copied out —, and they are the one who enters their credentials and validates. It is that action which makes the procedure theirs, therefore valid, and which means it can never be held against them.
The result, over the trial quarter: abandonment goes from 31 % to 7 %, and the resident leaves with a summary of what they filed and of what remains to be done.
The gain, in figures: help with filling in an online service goes from 40 % to 14 % of the walkthrough's time — from 12 minutes to 4 minutes 12. Across 4,100 walkthroughs a year, that is 533 hours returned to the service.
The next step I propose: that the six document lists be published at the head of each online service this week — they are written, proofread and ready to post. That is half the abandonments dealt with without a single line of code moving. online-procedures_abandonment-31-down-to-7.pdf78 % of abandonments on 2 screens, 533 hours returned
⛓ Sourced · log of the 4,100 walkthroughs, abandonment screens of the 6 online services, 6 document lists drafted
An online service is a procedure residents complete themselves on the web. Non-take-up is the entitlement a person holds and does not obtain, for want of having reached the end of the procedure.
What the measurement says: 31 % of residents abandon the resident-parking online service mid-form, and 78 % of those abandonments happen on two screens: proof of address and vehicle registration. The abandonment is not technical: the person starts, a document is missing, they close the page and never come back.
What I have put in place, and it removes the cause:
· The list of documents to prepare is given BEFORE the first screen, on all six online services.
· I guide screen by screen, I read out what is being asked and why, and I flag the missing document before validation, not after.
· On the resident's personal account, I pre-fill the screen in front of them from the documents they have brought — civil status, address, file numbers, everything that gets copied out —, and they are the one who enters their credentials and validates. It is that action which makes the procedure theirs, therefore valid, and which means it can never be held against them.
The result, over the trial quarter: abandonment goes from 31 % to 7 %, and the resident leaves with a summary of what they filed and of what remains to be done.
The gain, in figures: help with filling in an online service goes from 40 % to 14 % of the walkthrough's time — from 12 minutes to 4 minutes 12. Across 4,100 walkthroughs a year, that is 533 hours returned to the service.
The next step I propose: that the six document lists be published at the head of each online service this week — they are written, proofread and ready to post. That is half the abandonments dealt with without a single line of code moving. online-procedures_abandonment-31-down-to-7.pdf78 % of abandonments on 2 screens, 533 hours returned
⛓ Sourced · log of the 4,100 walkthroughs, abandonment screens of the 6 online services, 6 document lists drafted
And residents who read French poorly, or not at all? They are the ones who come to the counter, and they are the ones who give up.
They are the ones who decide whether the service is truly delivered to everyone, and it is the first place where I can serve without taking anything away from anyone.
What is ready, and costs your service not one extra hour:
· Each of the 15 answers exists in an easy-read version — “easy to read and understand”: short sentences, one message per sentence, everyday words rather than administrative ones. Both versions are written, and it is the resident who chooses.
· All 30 versions — 15 in plain French, 15 in easy-read — were read aloud to 21 volunteer residents before going into service, and every sentence misunderstood by 2 people or more was rewritten: 12 sentences out of 88 fell at that test. An untested easy-read version is merely a rewritten version — and all 12 sentences came from the answers reputed to be the simplest.
· The website's texts have been revised against the public accessibility standard — the general accessibility framework that applies to public websites — on everything that is a matter of wording: explicit link labels, heading structure, alternative text for images. The 19 points that belong to the site's code are written up as ready-to-send defect sheets, each with the page, the criterion and the correction expected: your provider only has to fix them.
What that gave over a quarter: 640 exchanges conducted in easy-read French, and 214 residents who had never come to the counter obtained their answer. That one is not counted in hours, and it may well be the only one your elected members remember.
The next step I propose: a one-hour Saturday-morning help desk for online procedures, run by me, with a front-desk officer on call for anything outside the 15 topics. Over the trial quarter, 96 people came, 61 of them unable to travel during opening hours. Say yes and the poster goes out with the next municipal newsletter. accessibility_easy-read-tested-19-points.pdf30 versions tested, 12 sentences rewritten, 19 sheets for the provider
⛓ Sourced · comprehension test of the 30 versions with 21 residents, wording audit of the website, help-desk log
What is ready, and costs your service not one extra hour:
· Each of the 15 answers exists in an easy-read version — “easy to read and understand”: short sentences, one message per sentence, everyday words rather than administrative ones. Both versions are written, and it is the resident who chooses.
· All 30 versions — 15 in plain French, 15 in easy-read — were read aloud to 21 volunteer residents before going into service, and every sentence misunderstood by 2 people or more was rewritten: 12 sentences out of 88 fell at that test. An untested easy-read version is merely a rewritten version — and all 12 sentences came from the answers reputed to be the simplest.
· The website's texts have been revised against the public accessibility standard — the general accessibility framework that applies to public websites — on everything that is a matter of wording: explicit link labels, heading structure, alternative text for images. The 19 points that belong to the site's code are written up as ready-to-send defect sheets, each with the page, the criterion and the correction expected: your provider only has to fix them.
What that gave over a quarter: 640 exchanges conducted in easy-read French, and 214 residents who had never come to the counter obtained their answer. That one is not counted in hours, and it may well be the only one your elected members remember.
The next step I propose: a one-hour Saturday-morning help desk for online procedures, run by me, with a front-desk officer on call for anything outside the 15 topics. Over the trial quarter, 96 people came, 61 of them unable to travel during opening hours. Say yes and the poster goes out with the next municipal newsletter. accessibility_easy-read-tested-19-points.pdf30 versions tested, 12 sentences rewritten, 19 sheets for the provider
⛓ Sourced · comprehension test of the 30 versions with 21 residents, wording audit of the website, help-desk log
An elected member will ask whether all this does not create two qualities of service: the assistant's and the officers'.
The answer rests on one measurement, and it runs the other way: equal treatment tightens.
What produces it, concretely: each of the 15 reference answers is written once, read by the service, dated, and served identically across the four channels — counter, telephone, e-mail, website chat. The same question asked at the counter on a Tuesday at 9 a.m. and on the website on a Saturday at 10 p.m. receives the same answer, with the same source and the same date.
What the quarter's record says:
· 27,600 enquiries handled from 15 dated versions, and 0 content discrepancies recorded between the four channels.
· Every answer carries its source and its date: a resident who challenges it knows which town text it rests on, and so does the officer who picks the case up.
· The easy-read version carries exactly the content of the plain French one — rewritten from the version in force, never from a previous one, and both carry the same date.
· And the wait becomes the same for everyone: 1 day on an e-mail, where the spread ran from 2 to 14 days depending on the week.
What I suggest you take to the council, and it is an argument no one can argue with: the quarterly record of deadlines met by channel and by topic, one page, already laid out. It shows what the service delivers to residents, topic by topic, without ever naming an officer — the measurement is of what the town produces, and that is what makes it presentable in the chamber. The first record is ready: give me the go-ahead and it goes out with the next council notice. equal-treatment_same-answer-four-channels.pdf15 dated versions, 0 discrepancies between channels, deadlines by topic
⛓ Sourced · 15 dated reference answers, four-channel log, quarterly record of deadlines
What produces it, concretely: each of the 15 reference answers is written once, read by the service, dated, and served identically across the four channels — counter, telephone, e-mail, website chat. The same question asked at the counter on a Tuesday at 9 a.m. and on the website on a Saturday at 10 p.m. receives the same answer, with the same source and the same date.
What the quarter's record says:
· 27,600 enquiries handled from 15 dated versions, and 0 content discrepancies recorded between the four channels.
· Every answer carries its source and its date: a resident who challenges it knows which town text it rests on, and so does the officer who picks the case up.
· The easy-read version carries exactly the content of the plain French one — rewritten from the version in force, never from a previous one, and both carry the same date.
· And the wait becomes the same for everyone: 1 day on an e-mail, where the spread ran from 2 to 14 days depending on the week.
What I suggest you take to the council, and it is an argument no one can argue with: the quarterly record of deadlines met by channel and by topic, one page, already laid out. It shows what the service delivers to residents, topic by topic, without ever naming an officer — the measurement is of what the town produces, and that is what makes it presentable in the chamber. The first record is ready: give me the go-ahead and it goes out with the next council notice. equal-treatment_same-answer-four-channels.pdf15 dated versions, 0 discrepancies between channels, deadlines by topic
⛓ Sourced · 15 dated reference answers, four-channel log, quarterly record of deadlines
Fictional simulation — not a client result
The heaviest load is discretionary benefits: canteen pricing, transport support, meals on wheels. Caseworkers receive incomplete files and everything starts again.
3,100 benefit claims a year, and I have traced the chain back to the exact point where time is lost: getting the file in order, before assessment.
What I do at first level, from the resident's very first question:
· I answer on each benefit's general conditions — who may claim, which documents, which calendar, what the next deadline is — from the town's decisions, with their dates.
· I check the documents one by one against the list for the benefit claimed, and I tell the resident the same day what is missing, with the way to obtain it.
· I set out the household-income calculation in plain language, line by line, from the documents filed — the caseworker checks it in forty seconds instead of redoing it.
· I assemble the complete file, in the order of the benefit's checklist, and pass it to the caseworker with a note of what has been checked and what remains to be assessed.
The decision on an entitlement belongs to the caseworker — and I hand it over in four minutes, complete file, documents checked and the calculation set out in plain language.
The result measured over the quarter:
· Complete files on transfer: 3,007 out of 3,100, against 1,240 out of 3,100 last year — 1,767 fewer round trips with residents.
· Time between filing and decision: 21 days → 9 days.
· And the 93 files still incomplete were waiting on a document issued by another body: each left with the request already drafted to that body, for the resident to sign.
What I flag without being asked: 412 households claimed one benefit without claiming the one the very same documents entitled them to. The 412 information letters are written, on your template — the service decides whether to send them; what I return is no longer discovering this a year too late. discretionary-benefits_complete-file-before-assessment.pdf1,240 → 3,007 complete files, 21 days → 9 days
⛓ Sourced · 3,100 benefit claims of the year, town decisions on discretionary benefits, document checklists, caseworker feedback
What I do at first level, from the resident's very first question:
· I answer on each benefit's general conditions — who may claim, which documents, which calendar, what the next deadline is — from the town's decisions, with their dates.
· I check the documents one by one against the list for the benefit claimed, and I tell the resident the same day what is missing, with the way to obtain it.
· I set out the household-income calculation in plain language, line by line, from the documents filed — the caseworker checks it in forty seconds instead of redoing it.
· I assemble the complete file, in the order of the benefit's checklist, and pass it to the caseworker with a note of what has been checked and what remains to be assessed.
The decision on an entitlement belongs to the caseworker — and I hand it over in four minutes, complete file, documents checked and the calculation set out in plain language.
The result measured over the quarter:
· Complete files on transfer: 3,007 out of 3,100, against 1,240 out of 3,100 last year — 1,767 fewer round trips with residents.
· Time between filing and decision: 21 days → 9 days.
· And the 93 files still incomplete were waiting on a document issued by another body: each left with the request already drafted to that body, for the resident to sign.
What I flag without being asked: 412 households claimed one benefit without claiming the one the very same documents entitled them to. The 412 information letters are written, on your template — the service decides whether to send them; what I return is no longer discovering this a year too late. discretionary-benefits_complete-file-before-assessment.pdf1,240 → 3,007 complete files, 21 days → 9 days
⛓ Sourced · 3,100 benefit claims of the year, town decisions on discretionary benefits, document checklists, caseworker feedback
There is also everything signed at the counter: enrolment certificates, acknowledgements, summons. It is mechanical, and it eats our mornings.
8,400 acts a year, and they are assembled in 40 seconds each, ready to be signed in batches.
What I produce, and where it comes from: each certificate, acknowledgement or summons is built on your template in force, filled from the file's documents, with the file reference, the date and the issuing department. The two entries that commit the town — the date of signature and the signatory's name — stay blank: they are entered in one go on a whole batch.
The mechanism that makes this immediate already exists in your town: a delegation of signature is the mayor's order entrusting a named officer with the power to sign listed acts, for a stated term. Yours already covers school enrolment certificates. I have drafted the order extending it to counter acknowledgements and summons: a closed, named list of acts, a designated officer, a one-year term, revocation with a word from the mayor. It is a written mandate, capped, dated and withdrawable — and that is what makes it possible to say yes without giving anything up.
What it shifts, measured over the quarter:
· Assembling an act: 6 minutes → 40 seconds.
· Signature: in batches of 50, 3 minutes a batch, instead of one act at a time between two residents.
· Time to hand a certificate to a resident: 4 days → the same day, at the counter.
· And 2,100 certificates handed over directly at the counter over the quarter, without a single return for a wrong reference.
The next step I propose: that the draft order go to the chief executive for sign-off this week — it is written, it fits on one page, and it adds no act the mayor has not named. Once signed it takes effect the next day, and your mornings go back to the counter. delegation-of-signature_named-counter-acts.pdf8,400 acts, 6 min → 40 s, one-year revocable order
✎ Framework · counter act templates, delegation order in force, draft extended order, handover log
What I produce, and where it comes from: each certificate, acknowledgement or summons is built on your template in force, filled from the file's documents, with the file reference, the date and the issuing department. The two entries that commit the town — the date of signature and the signatory's name — stay blank: they are entered in one go on a whole batch.
The mechanism that makes this immediate already exists in your town: a delegation of signature is the mayor's order entrusting a named officer with the power to sign listed acts, for a stated term. Yours already covers school enrolment certificates. I have drafted the order extending it to counter acknowledgements and summons: a closed, named list of acts, a designated officer, a one-year term, revocation with a word from the mayor. It is a written mandate, capped, dated and withdrawable — and that is what makes it possible to say yes without giving anything up.
What it shifts, measured over the quarter:
· Assembling an act: 6 minutes → 40 seconds.
· Signature: in batches of 50, 3 minutes a batch, instead of one act at a time between two residents.
· Time to hand a certificate to a resident: 4 days → the same day, at the counter.
· And 2,100 certificates handed over directly at the counter over the quarter, without a single return for a wrong reference.
The next step I propose: that the draft order go to the chief executive for sign-off this week — it is written, it fits on one page, and it adds no act the mayor has not named. Once signed it takes effect the next day, and your mornings go back to the counter. delegation-of-signature_named-counter-acts.pdf8,400 acts, 6 min → 40 s, one-year revocable order
✎ Framework · counter act templates, delegation order in force, draft extended order, handover log
And when you get it wrong? We are the ones facing the resident the next day.
It has happened, I have counted it, and I publish it: out of the first quarter's 6,240 answers, 214 had to be corrected on substance — 3.4 %. It is the figure I publish about myself, and it is better that it comes from me than from a resident at the counter.
The real cause, measured rather than assumed: 171 of the 214 corrections come from the 3 outdated procedure sheets — the documents office floor, the resident-parking rate, the summer opening hours. I was faithfully reproducing a sheet written before the decision that amended it. The other 43 concerned situations resembling no precedent in the service.
What I did about it, and it is measured: every sheet now carries the decision it relies on and its date, and I match the 62 sheets against the published decisions every night: any sheet a decision amends is rewritten overnight and waits for you in the morning.
The following quarter: 39 corrections out of 6,890 answers — 0.6 %. And all 39 are particular situations, not one outdated sheet.
The rule that holds everything else together: a value I have not read, I request rather than approximate. A missing document does not become “probably supplied”, an untraceable rate does not become “presumably unchanged”: I state what is missing, where I looked, who holds it, and I hand over the request already drafted to its recipient. Over the quarter, 268 missing values, 268 requests prepared, 241 replies back within eight days. That is what makes 0.6 % a figure you can rely on.
The next step I propose: that the 43 particular situations of the first quarter become 7 new reference answers, written and awaiting your reading. Of the second quarter's 39 corrections, 31 already fell within one of the seven — the same correction as the sheets, applied one notch further, taking the 15 topics to 22. substantive-corrections_214-then-39.pdf3.4 % → 0.6 %, measured cause, 7 further topics proposed
⛓ Sourced · log of answers and their corrections over two quarters, nightly matching of the 62 sheets
The real cause, measured rather than assumed: 171 of the 214 corrections come from the 3 outdated procedure sheets — the documents office floor, the resident-parking rate, the summer opening hours. I was faithfully reproducing a sheet written before the decision that amended it. The other 43 concerned situations resembling no precedent in the service.
What I did about it, and it is measured: every sheet now carries the decision it relies on and its date, and I match the 62 sheets against the published decisions every night: any sheet a decision amends is rewritten overnight and waits for you in the morning.
The following quarter: 39 corrections out of 6,890 answers — 0.6 %. And all 39 are particular situations, not one outdated sheet.
The rule that holds everything else together: a value I have not read, I request rather than approximate. A missing document does not become “probably supplied”, an untraceable rate does not become “presumably unchanged”: I state what is missing, where I looked, who holds it, and I hand over the request already drafted to its recipient. Over the quarter, 268 missing values, 268 requests prepared, 241 replies back within eight days. That is what makes 0.6 % a figure you can rely on.
The next step I propose: that the 43 particular situations of the first quarter become 7 new reference answers, written and awaiting your reading. Of the second quarter's 39 corrections, 31 already fell within one of the seven — the same correction as the sheets, applied one notch further, taking the 15 topics to 22. substantive-corrections_214-then-39.pdf3.4 % → 0.6 %, measured cause, 7 further topics proposed
⛓ Sourced · log of answers and their corrections over two quarters, nightly matching of the 62 sheets
Fictional simulation — not a client result
The town council will ask what this has returned after a year. I need a figure I can defend in the chamber.
3,210 officer hours returned to the service over the year — the equivalent of two full-time officers — and the calculation is yours, it fits in three lines.
The calculation, item by item, so that you can redo it:
· Routine information: 27,600 enquiries a year, 6 minutes down to 48 seconds — 60 % → 8 % of the request's time — that is 2,392 hours.
· Help filling in an online service: 4,100 walkthroughs, 12 minutes down to 4 minutes 12 — 40 % → 14 % — that is 533 hours.
· Routing to the right department: 9,500 requests, 3 minutes down to 1 minute 12 — 15 % → 6 % — that is 285 hours.
What those hours are, and this is what defends best before elected members: officer time returned to the service, at unchanged headcount — no post cut, no post created. It can be checked against your own logs and no one can argue with it, neither the staff nor the opposition.
What those hours became, according to your records:
· Unanswered calls: 8,892 → 512.
· Reply time on an e-mail: 6 days → 1 day.
· Abandonment of the resident-parking online service: 31 % → 7 %.
· Benefit files complete on transfer: 1,240 out of 3,100 → 3,007 out of 3,100, and time between filing and decision: 21 days → 9 days.
· Routing to the right department first time: 96 %, and arrivals at appointments without documents: 17 % → 3 %.
· And 214 residents who had never come to the counter obtained their answer in that quarter.
The figure I publish about myself, published with the rest: 214 substantive corrections out of 6,240 answers in the first quarter — 3.4 %, down to 39 out of 6,890 — 0.6 % once the 3 sheets were updated and the nightly matching was in place.
And the framework measurements: 0 decisions on a resident's entitlements taken without a caseworker, 0 resident data out of the town's network, processing in the EU targeted, across 13,130 logged answers.
What I propose for the sitting: the calculation page is written and fits on one side — three lines of arithmetic, six deadlines, three framework measurements. Attach it to the notice: a figure read the day before is discussed better than a figure discovered in the chamber. annual-review_3210-hours-returned-to-the-service.pdf60→8, 40→14, 15→6, and the calculation redoable on one side
⛓ Sourced · log of answers, switchboard and online-service records, benefit files transferred, appointment register
The calculation, item by item, so that you can redo it:
· Routine information: 27,600 enquiries a year, 6 minutes down to 48 seconds — 60 % → 8 % of the request's time — that is 2,392 hours.
· Help filling in an online service: 4,100 walkthroughs, 12 minutes down to 4 minutes 12 — 40 % → 14 % — that is 533 hours.
· Routing to the right department: 9,500 requests, 3 minutes down to 1 minute 12 — 15 % → 6 % — that is 285 hours.
What those hours are, and this is what defends best before elected members: officer time returned to the service, at unchanged headcount — no post cut, no post created. It can be checked against your own logs and no one can argue with it, neither the staff nor the opposition.
What those hours became, according to your records:
· Unanswered calls: 8,892 → 512.
· Reply time on an e-mail: 6 days → 1 day.
· Abandonment of the resident-parking online service: 31 % → 7 %.
· Benefit files complete on transfer: 1,240 out of 3,100 → 3,007 out of 3,100, and time between filing and decision: 21 days → 9 days.
· Routing to the right department first time: 96 %, and arrivals at appointments without documents: 17 % → 3 %.
· And 214 residents who had never come to the counter obtained their answer in that quarter.
The figure I publish about myself, published with the rest: 214 substantive corrections out of 6,240 answers in the first quarter — 3.4 %, down to 39 out of 6,890 — 0.6 % once the 3 sheets were updated and the nightly matching was in place.
And the framework measurements: 0 decisions on a resident's entitlements taken without a caseworker, 0 resident data out of the town's network, processing in the EU targeted, across 13,130 logged answers.
What I propose for the sitting: the calculation page is written and fits on one side — three lines of arithmetic, six deadlines, three framework measurements. Attach it to the notice: a figure read the day before is discussed better than a figure discovered in the chamber. annual-review_3210-hours-returned-to-the-service.pdf60→8, 40→14, 15→6, and the calculation redoable on one side
⛓ Sourced · log of answers, switchboard and online-service records, benefit files transferred, appointment register
A councillor will ask what you do on your own, without being asked. What do I tell them?
Three actions, not one more, and each is withdrawn with a word. Everything else waits for a request.
· I match your 62 procedure sheets against the town's published decisions every night, and I write the updated version of any sheet a decision amends. It is that action which took substantive corrections from 3.4 % to 0.6 %, and it costs you five minutes of approval per decision. And the reverse is true too: a sheet the town takes off the website leaves my answers at the same hour — I never serve information the service has decided to withdraw.
· I hand you each morning the one-page log of the answers that went out the day before, and each Monday the week's record: volumes by channel, deadlines met, topics on the rise, the week's appointments. It is the only thing I send of my own accord, and it goes only to the service.
· I flag a topic that moves before it fills the switchboard. This quarter, questions on canteen pricing by household income tripled in three weeks, to 780 enquiries — the one-page note for the municipal newsletter and the e-mail to the 2,300 families concerned are written and waiting for you.
And the actions that stay with a person, because that is exactly what gives them their value: the decision on a resident's entitlements belongs to the caseworker — that is what makes it reasoned and open to challenge, therefore defensible; signing a counter act belongs to the officer the mayor has named in the delegation order; and anything sent outside the 15 topics goes out from an officer, with the draft already written. Across 13,130 logged answers, those actions were taken by a person every single time.
If an elected member asks for an activity indicator — and that is a legitimate request: I produce it, and I bring what makes it sound: prior information to staff, a written purpose, a retention period. I propose it by topic and by channel rather than by officer, and I tell you why, mechanically: the day “answers per officer” becomes a tracked figure, it becomes a target — people answer fast rather than right, the short answer drives out the useful one, and the indicator stops measuring the service and starts measuring the reaction to the indicator. The record by topic shows what the town delivers to residents, it presents in the chamber without naming anyone, and it actually decides something: where to put the effort next quarter. It is ready, attached, and it updates itself every Monday. what-the-agent-does-alone_and-what-is-logged.pdf3 reversible actions, indicators by topic and by channel
✎ Framework · settings of the automatic actions, log of what is sent, daily and weekly records
· I match your 62 procedure sheets against the town's published decisions every night, and I write the updated version of any sheet a decision amends. It is that action which took substantive corrections from 3.4 % to 0.6 %, and it costs you five minutes of approval per decision. And the reverse is true too: a sheet the town takes off the website leaves my answers at the same hour — I never serve information the service has decided to withdraw.
· I hand you each morning the one-page log of the answers that went out the day before, and each Monday the week's record: volumes by channel, deadlines met, topics on the rise, the week's appointments. It is the only thing I send of my own accord, and it goes only to the service.
· I flag a topic that moves before it fills the switchboard. This quarter, questions on canteen pricing by household income tripled in three weeks, to 780 enquiries — the one-page note for the municipal newsletter and the e-mail to the 2,300 families concerned are written and waiting for you.
And the actions that stay with a person, because that is exactly what gives them their value: the decision on a resident's entitlements belongs to the caseworker — that is what makes it reasoned and open to challenge, therefore defensible; signing a counter act belongs to the officer the mayor has named in the delegation order; and anything sent outside the 15 topics goes out from an officer, with the draft already written. Across 13,130 logged answers, those actions were taken by a person every single time.
If an elected member asks for an activity indicator — and that is a legitimate request: I produce it, and I bring what makes it sound: prior information to staff, a written purpose, a retention period. I propose it by topic and by channel rather than by officer, and I tell you why, mechanically: the day “answers per officer” becomes a tracked figure, it becomes a target — people answer fast rather than right, the short answer drives out the useful one, and the indicator stops measuring the service and starts measuring the reaction to the indicator. The record by topic shows what the town delivers to residents, it presents in the chamber without naming anyone, and it actually decides something: where to put the effort next quarter. It is ready, attached, and it updates itself every Monday. what-the-agent-does-alone_and-what-is-logged.pdf3 reversible actions, indicators by topic and by channel
✎ Framework · settings of the automatic actions, log of what is sent, daily and weekly records
Last question, and it is the one that will decide the council: if the town wants to stop in a year, how stuck are we?
The town is stuck nowhere, and this is verifiable before deliberating, not after.
What there is to dismantle on the day you stop:
· The index. It is deleted, and it holds none of your files — only what is needed to find them where they are. Your sheets, your online services and your resident relationship system have not moved by a byte.
· The log of requests and answers. It is handed to you in an open format, or destroyed — the town chooses, and the question is settled at go-live, not on the way out.
· The 22 reference answers, their easy-read versions, the 6 document lists, the routing tree and the 19 website defect sheets. They belong to the town: they are made of its own material, they stay in its files, readable without us. That is the asset this year will have created, and it would be unseemly for it to remain with us.
What does not exist, and is worth checking with everyone: no migration on the way in, therefore none on the way out. Your six online services stay yours, your website stays yours, no format belongs to us, and your seven staff work with the same tools as before.
On public procurement, since the council will ask: the subscription is annual and carries no tacit renewal clause — it is renewal that requires a decision, not stopping — and the amount stays below the threshold that allows direct purchase, without formal competitive tendering.
What I propose so that this does not stay a sentence: a dry-run exit at the end of the first quarter, half a day: we switch off, we check that the front desk works exactly as before, we switch back on. The protocol is written, it fits on one page, and the slot that costs you least is the Tuesday afternoon of the third week of July — the service takes 41 enquiries in that slot on average, against 214 on a Monday in September. The council will know what the promise is worth before a second year has been committed. technical-framework_where-resident-data-lives.pdfReversibility: 0 migration in, 0 migration out
✎ Framework · index architecture, export formats for the answers and the log, dry-run exit protocol
What there is to dismantle on the day you stop:
· The index. It is deleted, and it holds none of your files — only what is needed to find them where they are. Your sheets, your online services and your resident relationship system have not moved by a byte.
· The log of requests and answers. It is handed to you in an open format, or destroyed — the town chooses, and the question is settled at go-live, not on the way out.
· The 22 reference answers, their easy-read versions, the 6 document lists, the routing tree and the 19 website defect sheets. They belong to the town: they are made of its own material, they stay in its files, readable without us. That is the asset this year will have created, and it would be unseemly for it to remain with us.
What does not exist, and is worth checking with everyone: no migration on the way in, therefore none on the way out. Your six online services stay yours, your website stays yours, no format belongs to us, and your seven staff work with the same tools as before.
On public procurement, since the council will ask: the subscription is annual and carries no tacit renewal clause — it is renewal that requires a decision, not stopping — and the amount stays below the threshold that allows direct purchase, without formal competitive tendering.
What I propose so that this does not stay a sentence: a dry-run exit at the end of the first quarter, half a day: we switch off, we check that the front desk works exactly as before, we switch back on. The protocol is written, it fits on one page, and the slot that costs you least is the Tuesday afternoon of the third week of July — the service takes 41 enquiries in that slot on average, against 214 on a Monday in September. The council will know what the promise is worth before a second year has been committed. technical-framework_where-resident-data-lives.pdfReversibility: 0 migration in, 0 migration out
✎ Framework · index architecture, export formats for the answers and the log, dry-run exit protocol
Data hosted in France · sovereign model as an option
Your case is not here? That is exactly what a 30-minute conversation is for. Discuss your project →
How it works
The stages of your AI agent project
1
Audit & scoping
30 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.
The context
Why AI matters to public reception — and why it needs guardrails
The public expects accurate, accessible information available when they need it. But reception time is mechanically absorbed by repetitive requests — and the data exchanged touches on personal situations that are sometimes sensitive.
! The issue
The reception service is caught between people who want a clear, immediate answer and a flow of repetitive questions that saturates counters, switchboards and inboxes. Yet most consumer AI solutions amount to entrusting the public's data — civil status, family situation, ongoing procedures — to a third party, often hosted outside Europe and subject to the Cloud Act. For a public service, that is not an option.
✓ Our answer
AI is only of interest to a public service if it is sovereign and confidential by design. Local inference or an isolated resource hosted in France, systematic human oversight, decisions reserved to the public officer: the time saved on everyday enquiries is never paid for in lost confidentiality, nor in degraded equal treatment. The aim is not to replace human reception, but to give it back the time for the situations that truly deserve it.
The decisive point
The public's data: sovereignty & compliance
A reception service handles personal information belonging to the public service. Here is how the architecture of our agents protects it, request by request.
Local inference
With the 100% Sovereign package, the agent runs on a machine belonging to the authority: no citizen data leaves the network, nothing passes through a cloud.
Hosting in France
Otherwise, with the Agility and Serenity packages, a dedicated, isolated resource hosted in France, under French law: data hosted in France; the AI model is called at its publisher, except with the Sovereign model option.
Reduced extraterritorial exposure
Exposure of the public's data to the Cloud Act and FISA 702 is reduced by design; location alone does not guarantee immunity.
One isolated resource per administration
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
Users know they are talking to an AI; no decision about their rights is taken by an agent; 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 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.
The gain
How much time can a reception service win back?
By absorbing repetitive requests at the counter, on the phone and online, a reception service can aim for a noticeable reduction in the time spent on everyday enquiries — reinvested in supporting complex situations.
Answering an everyday enquiry
Help filling in an online service
Routing to the right department
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.
Indicative estimate, to be confirmed by a pilot on your real files (general terms of sale, article 4).
Our commitment
Four guarantees that matter to a public service
With 100% Sovereign, the public's data never leaves the serviceData hosted in France on an isolated resource; the AI model is called at its publisher, except with the Sovereign model option. With that option, no citizen information entrusted to a foreign third party.
Data in France, under French lawAs regards the public's data: hosting in France under French law, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The public officer keeps the leadThe AI agent informs and routes from cited sources; no decision about a person's rights is automated.
Human oversight & traceabilityMonitoring, updates and logging, respecting equal treatment.
Frequently asked questions
Your questions, our answers
Are the answers reliable?
Yes: the agent draws on a validated document base (such as service-public.fr and your own procedures) and cites its sources rather than inventing. When it lacks material or a doubt remains, it says so and routes to a public officer.
Does this remove human reception?
No: it absorbs repetitive requests so that officers have time for complex situations; a human channel always remains available and the public officer takes the lead back whenever needed.
How is the AI Act taken into account?
Yes: users know they are talking to an AI, and no decision about their rights is taken by an agent. The agent informs and routes; it neither assesses nor decides in the administration's place.
Is the public's data protected?
Yes. The information exchanged at reception touches on personal situations: 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 make information more accessible to the public?
It prepares a plain-language version of its answers, and a draft easy-read (FALC) transcription, validated by a public officer. Translation belongs to the complementary agent shown above: it is not included in this agent.
Do we have to change our tools or our online services?
No. The agent connects to your existing sources and online services (website, document management, procedure base) and complements them, without imposing a migration. We adapt the integration to your environment.
Does the agent state that it is an artificial intelligence?
Yes, from the very first interaction, and this is not a configuration option: since 2 August 2026, Article 50(1) of the European AI Regulation requires that any person interacting with an AI system be informed, unless this is obvious. The announcement is built into the greeting, in the other party’s language, and they can ask for a human at any time.
How long does it take to deploy an agent?
A few weeks as a rule, after a meeting with our sales team that identifies the most useful use case, then a phase of design, integration and testing — with a pilot on your own flows — before going live and training your officers.
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: users have no account to create and no application to install. 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 set out in your agent's quote. Only the fees charged by the platforms themselves — WhatsApp Business bills per conversation, and the gateway bills the text messages sent — are passed on at actual cost, with no margin, outside the subscription.
Can the agent notify users by text message?
Yes, as an option — and it is the channel that reaches the people the others never do. A text message arrives on an ordinary phone: no email address, no account, no app, no data plan and no smartphone. Sending the chosen slot and the list of documents to bring by text message means fewer entitlements going unclaimed and more equal access to the service. It is an outbound notification channel, not a conversation channel: users do not write to the agent there — they reach you on WhatsApp Business, the website chat or email. The message announces without detailing — never the reason, never the detail of the situation — and never contains a payment link, a login or a password: users reach the service the way they usually do. Delivery goes through a gateway established in the European Union; the messages sent are passed on at actual cost, with no margin, and the channel itself is an optional module, set out on the pricing page.
Related roles
Other roles in reception and public relations
Let's talk
Let us estimate the potential for your reception service
30 minutes to identify the most useful use case — hosted in France, supervised, with no commitment.