● Public sector — Early years
AI agent for managing early years childcare
An assistant that handles childcare applications, prepares the allocation panels and the invoicing — the officer and the panel keep the decision. Hosted in France, on a resource isolated for each authority: children's and families' data stays protected. The AI agent assists, the panel allocates.
Hosted in France
Children's data protected
GDPR & AI Act: governed deployment
The panel allocates
Updated on
Deployed in a few weeks
Early years assistant · hosted in France
A family asks where their nursery application stands and which documents are missing. Prepare a reply.
Application found and reply drafted. Application registered on 12 March, position on the waiting list according to your adopted criteria. Documents missing: last year's tax assessment and proof of address less than 3 months old. The email sets out how to send them and the date of the next panel.
Ready for review before sending.
⛓ Source · your allocation rules + the family's file
Ready for review before sending.
⛓ Source · your allocation rules + the family's file
Prepare the files for the May allocation panel: places available, complete applications, criteria applied.
Panel files prepared. 8 places available (baby and toddler rooms), 32 complete applications ranked against your criteria grid, 5 incomplete files flagged with a reminder drafted. Each sheet refers back to the original documents.
The panel examines and allocates — nothing is decided by the agent.
✎ Action · files ready for the panel — the public officer approves
The panel examines and allocates — nothing is decided by the agent.
✎ Action · files ready for the panel — the public officer approves
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 panel allocatesThe agent prepares, it never allocates alone
✦ In brief
In the early years department, a Blue Lemon Agent agent handles childcare applications, keeps the waiting lists up to date (criteria, documents), prepares the allocation panel files and the means-tested invoicing, and answers families. It runs on local inference or is hosted in France on a resource dedicated to and isolated for each authority: children's and families' data is never exposed to a foreign service, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity. Allocating places remains the decision of the panel concerned. Live within a few weeks. Your public-sector staff write to it from Microsoft Teams, Slack or their email, and families reach it on WhatsApp Business, the website chat or email — with no account to create and nothing to install. Reaching the administration from the tool people already have means less non-take-up of rights and equal access to the service. These connections are included in every plan, at no extra cost, within the number of connections your level includes.
100%
hosted in France in the target architecture
0
transfer outside the EU in the target architecture
7
uses ready to deploy on this scope
0
decision taken without human approval
Reference points describing our offer, not results measured at a client. The scale of the gain is confirmed by a pilot on your own scope.
The context
Why AI matters to early years departments — and why they hesitate
Applications outnumbering the places available, waiting lists to keep up to date, panels to prepare, means-tested invoicing to recalculate at every change of circumstances: the early years department is buried in administration, while the data it handles — children's data — is among the most sensitive there is.
! The issue
The teams are caught between families waiting for a place and clear answers, and demanding management: allocation criteria to apply rigorously, documents to check, means-test bands to update. Yet most consumer AI tools amount to entrusting children's identities, family situations and household incomes to a third party, often hosted outside Europe and subject to the Cloud Act.
✓ Our answer
For children's and families' data, AI is only of interest if it is sovereign and confidential by design. Local inference or an isolated resource hosted in France, systematic human oversight, allocation reserved to the panel: the agent prepares complete files that can be checked, it never allocates a place. The aim is not to replace the officer, but to give them back time for families and for the nurseries.
The decisive point
Protecting children's and families' data: sovereignty & compliance
Identities, family situations, incomes: the early years department handles particularly sensitive data. Here is how the architecture of our agents protects it, authority by authority.
Local inference
The agent can run on a machine belonging to the authority: no data leaves the network, nothing passes through a cloud.
Hosting in France
Otherwise, a dedicated and isolated resource, hosted in France under French law — families' data: processing and access within the European Union targeted by the architecture.
Reduced extraterritorial exposure
Architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity for children's data: our architecture depends on a subcontracting chain and remote access documented for the configuration chosen.
One isolated resource per authority
No pooling of data: an environment strictly dedicated to your authority, guaranteeing the continuity of the public service.
Encryption & controlled access
Encryption in transit and at rest, role-based access (RBAC), strong authentication and logging.
AI Act: governed deployment
The agent is strictly in support; no place is allocated automatically; traceability and human oversight from end to end.
What depends on the architecture chosen These points are not general guarantees: they are settled deployment by deployment, in the quotation.
- The applicable location is that of the architecture set out in the quotation and verified before commissioning.
- Local execution is announced only for the configuration explicitly described and accepted in the quotation.
- The applicable isolation depends on the deployment mode set out in the quotation; no dedicated isolation is presumed.
- The encryption mechanisms in transit and at rest, their components and key management are those documented for the architecture chosen.
- Roles and permissions are configured and accepted for the identities and systems actually connected.
- The events logged, their content, their retention period and who may access them are defined for the deployment chosen.
For the most sensitive data, SecNumCloud and HDS options are available 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
4 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 bodyTown of Roche-sur-Yèvre — early years service and single application desk (fictional authority)
- Sector
- Early years service of a town of 18,400 inhabitants — 4 municipal nurseries and 1 micro-nursery, 138 places in all, a childminder support centre and 62 registered childminders across the area; single application desk since 1 September 2024
- Headcount
- 96 staff in the early years service, 5 of them on the single application desk and billing — 3 case officers, one coordinator and one apprentice; the other 91 are in the settings, with the children, and the agent does not touch that side
- Public served
- 1,180 families followed by the single desk, 486 applications for a place lodged each year for 138 places — 9,400 enquiries a year: 5,200 phone calls, 2,300 emails, 1,100 counter visits, 800 website forms
- Order of magnitude
- 4 allocation panels a year and 486 applications processed; 405 active contracts and 3,840 invoices issued; 620 changes of circumstance to feed through to the means-tested rate; 3,260 letters to families
- Tools already in place
- Early years software (family files, room planning, attendance records), family portal, billing software, the family allowance fund's means-test lookup service, the service mailbox and Microsoft Teams — the agent reads them, nothing is replaced and nothing is migrated
- Who decides what
- The allocation panel allocates places, in session and on the record; the early years coordinator approves every letter and every billing file before it goes out; allocation criteria and rates are set by resolution of the town council
- Room for improvement
- 1,664 calls out of 5,200 ring unanswered — 32%, 1,216 of them in the fortnight before each panel; an email waits 8 days on average; preparing a panel ties up all 3 case officers for 6 working days; and the means-test file had never been reconciled against attendance records — 26 discrepancies were sitting undisturbed in the last half-year alone
At Roche-sur-Yèvre, five staff run the single application desk and the billing for 1,180 families, with four annual peaks around the allocation panels. The most repetitive questions — where is my application, which documents do I need, how is my invoice calculated — take up more than three quarters of the flow, and the situations that deserve a person in front of them are the ones waiting behind. The exchanges below cover a year, from the reconciliation of the files to the report presented to the town council. The agent is hosted in France on an isolated resource, reads the early years software and the attendance records without ever writing to them, no invoice goes out without the coordinator's approval, and no place changes hands without a panel session.
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.
Assistant petite enfance · hébergé en France
I have reread your 9,400 enquiries for the year and reconciled, family by family, your means-test file against your attendance records and the changes of circumstance declared to you — the reconciliation that preparing the panels never leaves time for.
What the count says: 12 topics cover 7,240 enquiries out of 9,400, i.e. 77.0% of the flow. Position on the waiting list, documents required, how the invoice is calculated, change of circumstance, panel dates and how the panel works, types of childcare, opening hours and closures, absences and deductions, payment methods, settling-in, the childminder support centre, the family portal. The 12 reference answers are written, sourced on your operating rules and on the resolution of 14 October 2025, and dated — they are waiting for you.
What the three items you can check against your own logs are worth:
· Processing an application takes 22 minutes end to end, 60% of which goes on checking documents, finding the criteria and scoring — 13 minutes 12.
· A letter to a family about its file takes 20 minutes, 40% of it on the drafting itself — 8 minutes.
· Preparing a panel is 100% manual work: 6 working days for your 3 case officers, i.e. 126 hours per panel.
And the gap the reconciliation brought out, in the last half-year alone — 405 active contracts: 26 discrepancies, never seen because nobody ever had the six days to look. 10 means-test figures not updated after a change of circumstance that had in fact been declared, 7 attendance hours recorded twice across two rooms, 5 absence deductions not applied despite the certificate on file, 4 invoices issued after the recorded leaving date. In euros: €291.50 you did not bill and €548.70 billed wrongly.
The next step, and it takes twenty minutes: you reread the 26 lines and tell me, for each one, the action you want. The coordinator settles the version, and the 26 come out of the file before the next billing run. early-years-service-flow_9400-enquiries-12-topics.pdf7,240 enquiries across 12 topics, the three items costed, 1,550 hours identified
⛓ Sourced · 12 months of enquiry log, means-test file, attendance records, tariff resolution of 14 October 2025
What the count says: 12 topics cover 7,240 enquiries out of 9,400, i.e. 77.0% of the flow. Position on the waiting list, documents required, how the invoice is calculated, change of circumstance, panel dates and how the panel works, types of childcare, opening hours and closures, absences and deductions, payment methods, settling-in, the childminder support centre, the family portal. The 12 reference answers are written, sourced on your operating rules and on the resolution of 14 October 2025, and dated — they are waiting for you.
What the three items you can check against your own logs are worth:
· Processing an application takes 22 minutes end to end, 60% of which goes on checking documents, finding the criteria and scoring — 13 minutes 12.
· A letter to a family about its file takes 20 minutes, 40% of it on the drafting itself — 8 minutes.
· Preparing a panel is 100% manual work: 6 working days for your 3 case officers, i.e. 126 hours per panel.
And the gap the reconciliation brought out, in the last half-year alone — 405 active contracts: 26 discrepancies, never seen because nobody ever had the six days to look. 10 means-test figures not updated after a change of circumstance that had in fact been declared, 7 attendance hours recorded twice across two rooms, 5 absence deductions not applied despite the certificate on file, 4 invoices issued after the recorded leaving date. In euros: €291.50 you did not bill and €548.70 billed wrongly.
The next step, and it takes twenty minutes: you reread the 26 lines and tell me, for each one, the action you want. The coordinator settles the version, and the 26 come out of the file before the next billing run. early-years-service-flow_9400-enquiries-12-topics.pdf7,240 enquiries across 12 topics, the three items costed, 1,550 hours identified
⛓ Sourced · 12 months of enquiry log, means-test file, attendance records, tariff resolution of 14 October 2025
26 discrepancies in one half-year — what does that mean over a full year? And above all: how do we stop it happening again?
Over a year, the same reconciliation rule brings out 52 — €583.00 not billed and €1,097.40 billed wrongly. The amount is not what matters most, and I would rather say so straight away: what matters is that each of those 52 lines is a family complaint waiting to happen, and a complaint costs a letter, a phone call, and the trust of a parent who believes the service got it wrong about their child.
What I did on top, and what preparing the panels never leaves time for: I wrote four control rules, in the language of your operating rules, and ran them over your last 24 months of billing. For each one I give you the number of alerts it would have produced, the share confirmed after review, and what it would have let through — you choose on figures, not on a hunch:
· Rule A — an invoice calculated on a means-test figure whose update date predates the last recorded change of circumstance: 48 alerts over 24 months, 45 confirmed — 93.7%. The other 3 are changes entered twice, which is fixed at data entry.
· Rule B — the same attendance hour recorded in two rooms, in the same time slot: 26 alerts, 24 confirmed — 92.3%.
· Rule C — an invoice issued after the leaving date recorded on the contract: 16 alerts, 16 confirmed — 100%.
Total: 90 alerts over 24 months, 85 confirmed.
And here is what those three rules let through, because you should know it before you sign: 22 occasional-care invoices calculated at the floor rate for want of a means-test figure fall into none of the three — and those are precisely the families who would have had the lowest rate had they filed the document. So I wrote rule D — an invoice calculated at the floor rate when no means-test figure has been looked up for twelve months — and ran it over the same 24 months: 22 alerts, 19 confirmed. The four rules together cover 104 of the 113 discrepancies found over the period — 92.0%, i.e. 52 a year, exactly the figure from the reconciliation.
Bringing a rule into force stays with the service: a rule only goes live once approved by the coordinator, and that is precisely what makes it defensible against a family disputing its invoice. I save you the writing and the measuring; the decision takes twenty minutes instead of a committee. means-test-discrepancies_4-rules-tested-over-24-months.pdfThe 26 discrepancies of the half-year in detail, 4 rules tested, 104 of 113 discrepancies covered
⛓ Sourced · 24 months of billing (7,680 invoices), means-test file, attendance records, operating rules
What I did on top, and what preparing the panels never leaves time for: I wrote four control rules, in the language of your operating rules, and ran them over your last 24 months of billing. For each one I give you the number of alerts it would have produced, the share confirmed after review, and what it would have let through — you choose on figures, not on a hunch:
· Rule A — an invoice calculated on a means-test figure whose update date predates the last recorded change of circumstance: 48 alerts over 24 months, 45 confirmed — 93.7%. The other 3 are changes entered twice, which is fixed at data entry.
· Rule B — the same attendance hour recorded in two rooms, in the same time slot: 26 alerts, 24 confirmed — 92.3%.
· Rule C — an invoice issued after the leaving date recorded on the contract: 16 alerts, 16 confirmed — 100%.
Total: 90 alerts over 24 months, 85 confirmed.
And here is what those three rules let through, because you should know it before you sign: 22 occasional-care invoices calculated at the floor rate for want of a means-test figure fall into none of the three — and those are precisely the families who would have had the lowest rate had they filed the document. So I wrote rule D — an invoice calculated at the floor rate when no means-test figure has been looked up for twelve months — and ran it over the same 24 months: 22 alerts, 19 confirmed. The four rules together cover 104 of the 113 discrepancies found over the period — 92.0%, i.e. 52 a year, exactly the figure from the reconciliation.
Bringing a rule into force stays with the service: a rule only goes live once approved by the coordinator, and that is precisely what makes it defensible against a family disputing its invoice. I save you the writing and the measuring; the decision takes twenty minutes instead of a committee. means-test-discrepancies_4-rules-tested-over-24-months.pdfThe 26 discrepancies of the half-year in detail, 4 rules tested, 104 of 113 discrepancies covered
⛓ Sourced · 24 months of billing (7,680 invoices), means-test file, attendance records, operating rules
We are talking about children's files: identities, family situations, incomes, medical certificates. Where does that data go?
Nowhere outside your walls, and that can be checked rather than promised.
Where I run: on an isolated resource hosted in France, under French law, dedicated to Roche-sur-Yèvre and to nobody else — or on a machine in the service if you prefer local inference, in which case nothing leaves the town's network at all. No data about a child or a family is passed to a foreign service: exposure to the Cloud Act and FISA 702 is reduced accordingly, without location alone ruling it out — it is documented for the configuration chosen.
What I read, and what I cannot reach: I have read-only access to the family files, the attendance records and the means-test file. I have no write access to the billing software: I produce files and draft letters, your case officer imports them. Children's health data — individual care protocols, medical certificates, allergies — sits in a compartment I am not connected to, because none of the tasks you have entrusted to me needs it, and article 5.1(c) of the GDPR asks precisely that only what serves the purpose be exposed. One exception, and you choose it: the absence certificate, from which I take only the start date and the end date so as to apply your deduction rule — never the reason, never the content. Access is withdrawn on a word, and the withdrawal takes effect instantly.
What you can show your data protection officer: every read is logged with its date, its purpose and the file concerned; every output carries the documents it is drawn from; and the record of processing activities is already drafted, legal basis by legal basis — public interest task for place management and billing, article 6.1(e).
The next step I suggest: a quarterly thirty-minute review in which I show you the accesses actually used and those that served no purpose — an unused access is an access to close, and I bring you the list rather than asking you for it. sovereign-framework_hosting-access-traceability.pdfWhere children's data runs, who reads what, what is withdrawn on a word
✎ Framework · hosting architecture, access log, record of processing activities
Where I run: on an isolated resource hosted in France, under French law, dedicated to Roche-sur-Yèvre and to nobody else — or on a machine in the service if you prefer local inference, in which case nothing leaves the town's network at all. No data about a child or a family is passed to a foreign service: exposure to the Cloud Act and FISA 702 is reduced accordingly, without location alone ruling it out — it is documented for the configuration chosen.
What I read, and what I cannot reach: I have read-only access to the family files, the attendance records and the means-test file. I have no write access to the billing software: I produce files and draft letters, your case officer imports them. Children's health data — individual care protocols, medical certificates, allergies — sits in a compartment I am not connected to, because none of the tasks you have entrusted to me needs it, and article 5.1(c) of the GDPR asks precisely that only what serves the purpose be exposed. One exception, and you choose it: the absence certificate, from which I take only the start date and the end date so as to apply your deduction rule — never the reason, never the content. Access is withdrawn on a word, and the withdrawal takes effect instantly.
What you can show your data protection officer: every read is logged with its date, its purpose and the file concerned; every output carries the documents it is drawn from; and the record of processing activities is already drafted, legal basis by legal basis — public interest task for place management and billing, article 6.1(e).
The next step I suggest: a quarterly thirty-minute review in which I show you the accesses actually used and those that served no purpose — an unused access is an access to close, and I bring you the list rather than asking you for it. sovereign-framework_hosting-access-traceability.pdfWhere children's data runs, who reads what, what is withdrawn on a word
✎ Framework · hosting architecture, access log, record of processing activities
A family is asking where its nursery application stands and which documents are missing. Draft a reply.
File found and reply drafted, both parts handled in 90 seconds.
Where the application stands: the Marchand family's application, for Elias, born on 4 January, was registered on 12 March. It currently carries 17 points under your criteria grid resolved on 14 October 2025 — residence in the town, both parents in work, a sibling already attending Les Glycines — and each of those points links back to the document that justifies it. The position on the waiting list is the one your grid produces; it is not an allocated rank: the panel reviews, weighs and allocates in session, on the record. That is the sentence I put in the letter, because it is the one that prevents the misunderstanding six weeks later.
What is missing, and it is already written into the email: the previous-year tax notice — the one filed is two years old — and a proof of address less than 3 months old: the one on file is nine months old. The email states, for each document, the accepted format, the three ways to send it — upload to the family portal, counter, post — and the deadline for the file to be reviewable.
And the date that matters: the next panel meets on 21 May, files must be complete by 30 April. The letter says so in one sentence, at the top, because it is the only piece of information the parent is looking for.
What this is worth over a year: your 3,260 letters to families go from 8 minutes of drafting to 2 minutes — 40% of the time down to 10%, i.e. 326 hours returned to the service. And processing an application, document checks and scoring included, goes from 13 minutes 12 to 2 minutes 12 — 60% down to 10%, i.e. 89 hours across your 486 applications.
The next step I suggest: that I flag to you, every Monday, the files where a document expires before the panel. Of last year's 486 applications, 118 reached the panel with an out-of-date proof of address, and 74 had to be held over to the following panel — three more months of waiting for the family, over a two-page document. application-file_marchand-family-and-document-checks.pdfThe file found, the 17 points justified document by document, the reply to the family
⛓ Sourced · operating rules, resolution of 14 October 2025 (criteria grid), Marchand family file
Where the application stands: the Marchand family's application, for Elias, born on 4 January, was registered on 12 March. It currently carries 17 points under your criteria grid resolved on 14 October 2025 — residence in the town, both parents in work, a sibling already attending Les Glycines — and each of those points links back to the document that justifies it. The position on the waiting list is the one your grid produces; it is not an allocated rank: the panel reviews, weighs and allocates in session, on the record. That is the sentence I put in the letter, because it is the one that prevents the misunderstanding six weeks later.
What is missing, and it is already written into the email: the previous-year tax notice — the one filed is two years old — and a proof of address less than 3 months old: the one on file is nine months old. The email states, for each document, the accepted format, the three ways to send it — upload to the family portal, counter, post — and the deadline for the file to be reviewable.
And the date that matters: the next panel meets on 21 May, files must be complete by 30 April. The letter says so in one sentence, at the top, because it is the only piece of information the parent is looking for.
What this is worth over a year: your 3,260 letters to families go from 8 minutes of drafting to 2 minutes — 40% of the time down to 10%, i.e. 326 hours returned to the service. And processing an application, document checks and scoring included, goes from 13 minutes 12 to 2 minutes 12 — 60% down to 10%, i.e. 89 hours across your 486 applications.
The next step I suggest: that I flag to you, every Monday, the files where a document expires before the panel. Of last year's 486 applications, 118 reached the panel with an out-of-date proof of address, and 74 had to be held over to the following panel — three more months of waiting for the family, over a two-page document. application-file_marchand-family-and-document-checks.pdfThe file found, the 17 points justified document by document, the reply to the family
⛓ Sourced · operating rules, resolution of 14 October 2025 (criteria grid), Marchand family file
The real problem is the phone. In the fortnight before a panel, we stop picking up. One parent in three gets nothing.
1,664 calls out of 5,200 ring unanswered — 32.0%, and 1,216 of them fall in the fortnight before your four panels. This is not a quality-of-service problem, it is a time-of-day problem: your 3 case officers are wanted by five families at once between 9 and 11 in the morning.
What I take, and what I leave you: I pick up every call, I announce in the first sentence that I am an artificial intelligence — article 50(1) of the European AI Regulation has required this since 2 August 2026, and the family can ask for a human at any moment — and I handle to the end the 12 topics whose answers are approved. Anything outside those 12 topics goes to a case officer with the file already assembled: the family, the child, the filing date, the points carried and the document behind each, the last letter sent, and the question as it was asked. Your case officer picks up already knowing, instead of picking up and searching.
The arithmetic, and it is yours: the 12 topics account for 77.0% of the flow; 1,664 unanswered calls drop to 104, i.e. 1,560 more families getting their answer — 1,108 of them in the evening, at the weekend or over lunch, when the desk is closed. And those are the hours of parents who work, that is to say precisely the ones whose file rests on a certificate of employment.
And I reach them on the channels they already have: WhatsApp Business, the website chat and email, with no account to create and no app to install. Your case officers, for their part, write to me from Microsoft Teams, Slack or their mailbox, without changing tools. Email, which today waits 8 days, goes back to 1 day. On those channels I give information without ever restating the detail of a family situation in a message: the family is invited to find its document on the portal or at the desk.
The figure that does not flatter me, and I publish it: of the 1,240 information requests in the first quarter, I directed 89 to the wrong setting — 7.2%. All of them concerned the 2 nurseries whose rooms were redrawn on 1 January: I was reading the room table in force, not the one applicable at the filing date. That is fixed: I now read the table as it stood on the date the file was lodged. In the second quarter, 8 wrong referrals out of 1,240 — 0.6%. I suggest we keep publishing that rate every quarter: an agent whose errors are not measured is an agent that cannot be corrected. family-channels_multichannel-desk-sms-standard-letters.pdfThe 12 topics handled to the end, how the rest is routed, the published error rate
⛓ Sourced · 12 months of switchboard log, agent referral log, room allocation table
What I take, and what I leave you: I pick up every call, I announce in the first sentence that I am an artificial intelligence — article 50(1) of the European AI Regulation has required this since 2 August 2026, and the family can ask for a human at any moment — and I handle to the end the 12 topics whose answers are approved. Anything outside those 12 topics goes to a case officer with the file already assembled: the family, the child, the filing date, the points carried and the document behind each, the last letter sent, and the question as it was asked. Your case officer picks up already knowing, instead of picking up and searching.
The arithmetic, and it is yours: the 12 topics account for 77.0% of the flow; 1,664 unanswered calls drop to 104, i.e. 1,560 more families getting their answer — 1,108 of them in the evening, at the weekend or over lunch, when the desk is closed. And those are the hours of parents who work, that is to say precisely the ones whose file rests on a certificate of employment.
And I reach them on the channels they already have: WhatsApp Business, the website chat and email, with no account to create and no app to install. Your case officers, for their part, write to me from Microsoft Teams, Slack or their mailbox, without changing tools. Email, which today waits 8 days, goes back to 1 day. On those channels I give information without ever restating the detail of a family situation in a message: the family is invited to find its document on the portal or at the desk.
The figure that does not flatter me, and I publish it: of the 1,240 information requests in the first quarter, I directed 89 to the wrong setting — 7.2%. All of them concerned the 2 nurseries whose rooms were redrawn on 1 January: I was reading the room table in force, not the one applicable at the filing date. That is fixed: I now read the table as it stood on the date the file was lodged. In the second quarter, 8 wrong referrals out of 1,240 — 0.6%. I suggest we keep publishing that rate every quarter: an agent whose errors are not measured is an agent that cannot be corrected. family-channels_multichannel-desk-sms-standard-letters.pdfThe 12 topics handled to the end, how the rest is routed, the published error rate
⛓ Sourced · 12 months of switchboard log, agent referral log, room allocation table
And an unplanned room closure, or a panel date being moved: families never see it. We get a wave of calls the next day.
428 of your 5,200 calls fall in the 48 hours after your 9 unplanned closures of the year — 8.2% of the phone flow across 18 days. It is the most concentrated item in your service, and it is the one that is most simply solved.
What I suggest, costed: a text message the day before, to the families of the setting and the room concerned and to nobody else. 9 closures a year, 34 families per room on average: 306 messages over the year, recharged at the gateway's actual cost, with no mark-up, routing going through a gateway established in the European Union. A text arrives on an ordinary phone: no email address, no account, no app and no data plan — it is the channel that reaches the families the online portal never reaches, and that is avoided non-take-up as much as it is an avoided phone call. The same channel carries the most awaited message of all: « a place is being offered to you, the panel of 21 May has decided, your letter goes out today ».
What the message contains, and nothing else: the setting, the room, the date, and an invitation to contact the desk. Never the reason for a closure, never the detail of a family situation, never a payment link, never an identifier. It is an outbound notification: the family does not reply to it, it contacts me on WhatsApp, the chat or by email.
What this is worth, on the basis of your own figures: I will not give you another authority's result, it would commit you to nothing. I give you the measurement you will be able to check yourself: 428 calls today across the 18 days concerned. I commit to the protocol, not to the result: we measure the first closure without a text and the second with one, in settings of comparable size, and you decide on the gap.
And while we are here, the other end of the subject: 1,120 of your enquiries are about which documents to provide, and 214 files a year arrive incomplete a second time — the family read the letter and did not understand what was being asked. I have reworked the 6 standard letters that account for 71% of second reminders — one document per line, the deadline at the top, and first of all the single document that was most often missing — and each one carries the number of second reminders it triggered, so that the decision is taken on the measurement and not on my word. For a plain-language version, pass this content to the Accessibility, Easy Read and Translation agent — a dedicated agent, and it is not part of what this agent includes: I prepare the facts, the documents expected and the deadlines it needs, and it is that agent which produces the adapted version. The letter you approve serves the family portal, the chat, the counter and the paper letter — the same dated version on all four supports. family-channels_multichannel-desk-sms-standard-letters.pdfThe text message word for word, the measurement protocol, the 6 letters reworked and their second reminders
⛓ Sourced · 12 months of switchboard log, calendar of unplanned closures, 2,300 emails of the year
What I suggest, costed: a text message the day before, to the families of the setting and the room concerned and to nobody else. 9 closures a year, 34 families per room on average: 306 messages over the year, recharged at the gateway's actual cost, with no mark-up, routing going through a gateway established in the European Union. A text arrives on an ordinary phone: no email address, no account, no app and no data plan — it is the channel that reaches the families the online portal never reaches, and that is avoided non-take-up as much as it is an avoided phone call. The same channel carries the most awaited message of all: « a place is being offered to you, the panel of 21 May has decided, your letter goes out today ».
What the message contains, and nothing else: the setting, the room, the date, and an invitation to contact the desk. Never the reason for a closure, never the detail of a family situation, never a payment link, never an identifier. It is an outbound notification: the family does not reply to it, it contacts me on WhatsApp, the chat or by email.
What this is worth, on the basis of your own figures: I will not give you another authority's result, it would commit you to nothing. I give you the measurement you will be able to check yourself: 428 calls today across the 18 days concerned. I commit to the protocol, not to the result: we measure the first closure without a text and the second with one, in settings of comparable size, and you decide on the gap.
And while we are here, the other end of the subject: 1,120 of your enquiries are about which documents to provide, and 214 files a year arrive incomplete a second time — the family read the letter and did not understand what was being asked. I have reworked the 6 standard letters that account for 71% of second reminders — one document per line, the deadline at the top, and first of all the single document that was most often missing — and each one carries the number of second reminders it triggered, so that the decision is taken on the measurement and not on my word. For a plain-language version, pass this content to the Accessibility, Easy Read and Translation agent — a dedicated agent, and it is not part of what this agent includes: I prepare the facts, the documents expected and the deadlines it needs, and it is that agent which produces the adapted version. The letter you approve serves the family portal, the chat, the counter and the paper letter — the same dated version on all four supports. family-channels_multichannel-desk-sms-standard-letters.pdfThe text message word for word, the measurement protocol, the 6 letters reworked and their second reminders
⛓ Sourced · 12 months of switchboard log, calendar of unplanned closures, 2,300 emails of the year
Prepare the files for the May allocation panel: available places, complete applications, criteria applied.
Panel files prepared. 8 places available — 3 in the baby room and 5 in the toddler room, split between Les Glycines and Le Parc nurseries, each with the date it frees up and the reason it does. 32 complete applications scored against your criteria grid resolved on 14 October 2025, and 5 incomplete files flagged, with a reminder drafted for each.
What each sheet carries, and this is the rule I suggest we hold to: no file goes before the panel without its source document and the scoring criterion that justifies it. For the 32 files you have the points line by line — residence, parents in work, single-parent situation, sibling attending, filing date, referral by social services — and, against each point, the document that grounds it and its date. That is what lets you produce, to a parent asking why, the article of the resolution and the document, rather than a rank.
What the scoring does not settle, and I flag it rather than smoothing it over: 6 files come out level on points for the 3 places in the baby room. The position the grid produces is not an allocated place — and on these 6 files there is not even a position: this is exactly where the panel does something no scoring grid replaces, and I have prepared for each of them a sourced situation note so that it can deliberate with the facts in hand.
The 5 incomplete files: for each one, the missing document, the date it was first requested, and the reminder ready to go — 2 previous-year tax notices, 2 proofs of address less than 3 months old, 1 family allowance fund certificate. Sent tonight, the 5 files are reviewable on 21 May.
What this is worth: a panel used to tie up your 3 case officers for 6 working days — 126 hours. Now: 25 hours 12 of review and approval, i.e. 20% of the current time, one day for three people. 100 hours 48 returned per panel, 403 hours across your 4 annual panels.
Nothing is decided: the 32 sheets, the 6 situation notes and the 5 reminders are waiting for the coordinator, and allocation is taken in session, on the record.
The next step I suggest: keeping the history of every waiting-list movement — entry, change of points, hold-over, exit — with its date, its reason and its author. Of last year's 486 applications, 61 changed position without anyone being able to say today why. Those are 61 answers you cannot give a parent who asks for them — and they will all be available from the next panel onwards. allocation-panel_may_8-places-32-applications.pdfThe 8 places, the 32 sheets justified document by document, the 6 ties, the 5 reminders
✎ Action · files ready for the panel — the public official approves, the panel allocates · resolution of 14 October 2025
What each sheet carries, and this is the rule I suggest we hold to: no file goes before the panel without its source document and the scoring criterion that justifies it. For the 32 files you have the points line by line — residence, parents in work, single-parent situation, sibling attending, filing date, referral by social services — and, against each point, the document that grounds it and its date. That is what lets you produce, to a parent asking why, the article of the resolution and the document, rather than a rank.
What the scoring does not settle, and I flag it rather than smoothing it over: 6 files come out level on points for the 3 places in the baby room. The position the grid produces is not an allocated place — and on these 6 files there is not even a position: this is exactly where the panel does something no scoring grid replaces, and I have prepared for each of them a sourced situation note so that it can deliberate with the facts in hand.
The 5 incomplete files: for each one, the missing document, the date it was first requested, and the reminder ready to go — 2 previous-year tax notices, 2 proofs of address less than 3 months old, 1 family allowance fund certificate. Sent tonight, the 5 files are reviewable on 21 May.
What this is worth: a panel used to tie up your 3 case officers for 6 working days — 126 hours. Now: 25 hours 12 of review and approval, i.e. 20% of the current time, one day for three people. 100 hours 48 returned per panel, 403 hours across your 4 annual panels.
Nothing is decided: the 32 sheets, the 6 situation notes and the 5 reminders are waiting for the coordinator, and allocation is taken in session, on the record.
The next step I suggest: keeping the history of every waiting-list movement — entry, change of points, hold-over, exit — with its date, its reason and its author. Of last year's 486 applications, 61 changed position without anyone being able to say today why. Those are 61 answers you cannot give a parent who asks for them — and they will all be available from the next panel onwards. allocation-panel_may_8-places-32-applications.pdfThe 8 places, the 32 sheets justified document by document, the 6 ties, the 5 reminders
✎ Action · files ready for the panel — the public official approves, the panel allocates · resolution of 14 October 2025
The documents, precisely: we chase them for three weeks, and we only notice at the assessment stage.
1,106 document sets a year — 486 applications and 620 changes of circumstance — and the check is done today at assessment, that is to say three weeks too late.
What I do at filing, within the minute: I read the proof of address, the tax notice and the family allowance fund certificate attached to the request, and I say what is missing before the file enters the queue. Across last year's 1,106 files run back through this rule: 214 were incomplete — 19.3%. Document absent in 108 cases, proof of address more than three months old in 62 cases, tax notice for the wrong year — two years back instead of one — in 44 cases. Those 44 are worth their weight: an out-of-date tax notice makes an out-of-date means-test figure, so a wrong invoice, so an adjustment six months later on a family budget that was not expecting it.
What I read and what I keep: I record the name, the address, the date of the document, the number of people in the household, the reference taxable income and the claimant number. The rest of the tax notice is of no interest to me and is not kept — a tax notice carries far more than a means test needs, and article 5.1(c) of the GDPR asks precisely that only the useful part be retained. That is an argument your data protection officer can reuse as it stands.
What it delivers: the time to complete a file goes from 24 days to 5, because today's three weeks are three weeks of documents going back and forth. And of the 74 files held over to the next panel last year for want of an up-to-date document, almost all of them came down to that check — three more months of waiting for the family, over two pages.
And here is where you need to give me a mandate, because it commits you: sending out an invoice, notifying an offer of a place, raising an adjustment are acts that commit the town. These are not acts I rule out — they are acts I carry out as soon as the mandate is written: named scope, stated ceiling, review date, and withdrawal on a word. The model mandate is drafted, with three scopes to choose from — from the most cautious (nothing goes without line-by-line approval) to the widest (document reminders and acknowledgements go on their own, everything else waits). Allocating a place appears in none of the three: it belongs to the panel, and that is a boundary the slider does not cross. application-file_marchand-family-and-document-checks.pdf214 incomplete files caught at filing, what is read and what is kept, the three-setting mandate
⛓ Sourced · 1,106 files of the year run back through the control rule, operating rules, resolution of 14 October 2025
What I do at filing, within the minute: I read the proof of address, the tax notice and the family allowance fund certificate attached to the request, and I say what is missing before the file enters the queue. Across last year's 1,106 files run back through this rule: 214 were incomplete — 19.3%. Document absent in 108 cases, proof of address more than three months old in 62 cases, tax notice for the wrong year — two years back instead of one — in 44 cases. Those 44 are worth their weight: an out-of-date tax notice makes an out-of-date means-test figure, so a wrong invoice, so an adjustment six months later on a family budget that was not expecting it.
What I read and what I keep: I record the name, the address, the date of the document, the number of people in the household, the reference taxable income and the claimant number. The rest of the tax notice is of no interest to me and is not kept — a tax notice carries far more than a means test needs, and article 5.1(c) of the GDPR asks precisely that only the useful part be retained. That is an argument your data protection officer can reuse as it stands.
What it delivers: the time to complete a file goes from 24 days to 5, because today's three weeks are three weeks of documents going back and forth. And of the 74 files held over to the next panel last year for want of an up-to-date document, almost all of them came down to that check — three more months of waiting for the family, over two pages.
And here is where you need to give me a mandate, because it commits you: sending out an invoice, notifying an offer of a place, raising an adjustment are acts that commit the town. These are not acts I rule out — they are acts I carry out as soon as the mandate is written: named scope, stated ceiling, review date, and withdrawal on a word. The model mandate is drafted, with three scopes to choose from — from the most cautious (nothing goes without line-by-line approval) to the widest (document reminders and acknowledgements go on their own, everything else waits). Allocating a place appears in none of the three: it belongs to the panel, and that is a boundary the slider does not cross. application-file_marchand-family-and-document-checks.pdf214 incomplete files caught at filing, what is read and what is kept, the three-setting mandate
⛓ Sourced · 1,106 files of the year run back through the control rule, operating rules, resolution of 14 October 2025
And the billing? Changes of circumstance during the year are what put us most behind.
Preparation complete for the month of May. 405 active contracts, 320 invoices to issue: hourly rate calculated from the means-test figure and the effort rate in your resolution of 14 October 2025, according to the number of dependent children, applied to the hours booked on the contract and the hours actually recorded, with your absence deduction rules. 311 invoices ready, 9 lines set aside for checking — they do not go out as invoices, they go to review.
The detail you will not have to look for: total billed €41,286.40, a 2.1% variance against April, fully explained by 6 arrivals, 4 departures and 11 means-test changes — I give you the 21 movements, line by line, with the document behind each.
The year's 620 changes of circumstance are where the delay is created: a birth, a separation, a return to work, a job loss all change the means-test figure, and therefore the rate. I take them as they come: as soon as a change is declared and its document filed, I recalculate the applicable rate, date its effect according to your rules, and prepare the letter that explains it to the family — before the invoice, not after. Of last year's 620 changes, 187 were fed through more than a month late, and 61 gave rise to an adjustment of more than €40 — those 61 letters are what generates the following month's phone calls.
Nothing is issued: the billing file and the 311 letters are waiting for the coordinator's approval, and I can take it all back on a word for as long as nothing has gone out.
The next step I suggest: bring rule C in first — 16 alerts over 24 months, 100% confirmed, so no review time wasted — then rule A, which carries the largest volume. I advise against switching them all on in the same month, and I cost that advice: all four together would produce 56 alerts in the first run, i.e. about 14 hours of review for your 3 case officers, in the middle of panel preparation. One rule per run is 3 hours 30 each time, and you keep control of what you look at.
And if a rule starts flagging noise, you will hear it from me: a rule that falls below 80% confirmation is a rule that has families reviewed for nothing, and I warn you in the very run concerned, with the tightened version already written and already run over the history — as I did for rule B: its first drafting, which flagged any attendance recorded twice in a day, produced 71 alerts for 24 confirmed (33.8%); the version tightened to the same time slot produces 26 for the same 24 confirmed — 45 families spared being looked at for nothing. It is the tightened version I am putting to you, not the first. means-tested-billing_may_320-invoices-and-21-movements.pdf311 invoices ready, 9 lines under review, the 620 changes of circumstance taken as they come
✎ Action · billing prepared — the coordinator approves before issue · resolution of 14 October 2025, May attendance records
The detail you will not have to look for: total billed €41,286.40, a 2.1% variance against April, fully explained by 6 arrivals, 4 departures and 11 means-test changes — I give you the 21 movements, line by line, with the document behind each.
The year's 620 changes of circumstance are where the delay is created: a birth, a separation, a return to work, a job loss all change the means-test figure, and therefore the rate. I take them as they come: as soon as a change is declared and its document filed, I recalculate the applicable rate, date its effect according to your rules, and prepare the letter that explains it to the family — before the invoice, not after. Of last year's 620 changes, 187 were fed through more than a month late, and 61 gave rise to an adjustment of more than €40 — those 61 letters are what generates the following month's phone calls.
Nothing is issued: the billing file and the 311 letters are waiting for the coordinator's approval, and I can take it all back on a word for as long as nothing has gone out.
The next step I suggest: bring rule C in first — 16 alerts over 24 months, 100% confirmed, so no review time wasted — then rule A, which carries the largest volume. I advise against switching them all on in the same month, and I cost that advice: all four together would produce 56 alerts in the first run, i.e. about 14 hours of review for your 3 case officers, in the middle of panel preparation. One rule per run is 3 hours 30 each time, and you keep control of what you look at.
And if a rule starts flagging noise, you will hear it from me: a rule that falls below 80% confirmation is a rule that has families reviewed for nothing, and I warn you in the very run concerned, with the tightened version already written and already run over the history — as I did for rule B: its first drafting, which flagged any attendance recorded twice in a day, produced 71 alerts for 24 confirmed (33.8%); the version tightened to the same time slot produces 26 for the same 24 confirmed — 45 families spared being looked at for nothing. It is the tightened version I am putting to you, not the first. means-tested-billing_may_320-invoices-and-21-movements.pdf311 invoices ready, 9 lines under review, the 620 changes of circumstance taken as they come
✎ Action · billing prepared — the coordinator approves before issue · resolution of 14 October 2025, May attendance records
The town council is asking me for a twelve-month report. What do I present to it?
Four figures, and each one can be checked against your own logs — it is the only way to stand up before a council.
· 1,550 hours returned to the service over the year, item by item: 89 h on processing applications, 326 h on letters to families, 403 h on the four panels, 732 h on day-to-day information. At 1,607 hours to a working year, that is more than ten months, or more than forty-four weeks, returned to your five staff. No post cut, no post created: it is time given back to families, to the settings and to the situations that need a professional in front of them.
· 1,664 unanswered calls down to 104, and email from 8 days to 1 day.
· 52 billing discrepancies taken out of the file over the year, where the reconciliation was never done — €583.00 of income restored and €1,097.40 that did not go out wrongly, and above all 52 complaint letters that never arrived.
· 0 places allocated outside a panel, 0 positions changed after a session, 0 invoices issued without a named approval.
What I bring you on top, and what waits for you every year: your early years annual activity report and the activity data expected under your funding agreement draw in part on data I already keep up to date. The 11 activity and family-relations indicators are extracted, dated, with the query that produces them, so that a councillor who asks where a figure comes from gets the answer in the session. The report is still written and presented by you: it is an act of the authority, not a machine output.
The next step I suggest for the coming year: the costliest subject in your service remains room occupancy — 74% observed over the year, against 82% in the objectives agreement. I can give you, every week, the booked hours not taken up by room and by time slot, alongside the occasional-care requests left unanswered in those same slots. It is a reconciliation, not a promise: if the two curves do not overlap, you will hear it from me. year-report_1550-hours-item-by-item.pdfThe 1,550 hours in detail, the response times met, the 11 activity report indicators
⛓ Sourced · 12 months of service logs, means-test file, attendance records, previous year's activity report
· 1,550 hours returned to the service over the year, item by item: 89 h on processing applications, 326 h on letters to families, 403 h on the four panels, 732 h on day-to-day information. At 1,607 hours to a working year, that is more than ten months, or more than forty-four weeks, returned to your five staff. No post cut, no post created: it is time given back to families, to the settings and to the situations that need a professional in front of them.
· 1,664 unanswered calls down to 104, and email from 8 days to 1 day.
· 52 billing discrepancies taken out of the file over the year, where the reconciliation was never done — €583.00 of income restored and €1,097.40 that did not go out wrongly, and above all 52 complaint letters that never arrived.
· 0 places allocated outside a panel, 0 positions changed after a session, 0 invoices issued without a named approval.
What I bring you on top, and what waits for you every year: your early years annual activity report and the activity data expected under your funding agreement draw in part on data I already keep up to date. The 11 activity and family-relations indicators are extracted, dated, with the query that produces them, so that a councillor who asks where a figure comes from gets the answer in the session. The report is still written and presented by you: it is an act of the authority, not a machine output.
The next step I suggest for the coming year: the costliest subject in your service remains room occupancy — 74% observed over the year, against 82% in the objectives agreement. I can give you, every week, the booked hours not taken up by room and by time slot, alongside the occasional-care requests left unanswered in those same slots. It is a reconciliation, not a promise: if the two curves do not overlap, you will hear it from me. year-report_1550-hours-item-by-item.pdfThe 1,550 hours in detail, the response times met, the 11 activity report indicators
⛓ Sourced · 12 months of service logs, means-test file, attendance records, previous year's activity report
In practical terms, what does the agent do without asking us? That is the question the council will put.
Three actions, not one more, and all three can be undone.
· I acknowledge every application and every change of circumstance within the minute, with the list of missing documents, the completion deadline and the date of the next panel. And the reverse holds too: if the family files the document the next day, the acknowledgement updates itself and the reminder does not go out — because a parent who gets a reminder for a document they have just filed calls the desk, and they are right to.
· I set aside for checking any billing line that triggers one of the rules in service. And the reverse holds too: the case officer clears it on a word, the line goes back into billing, with the reason and the date it was cleared — which gives you, at the next run, the list of the most frequent reasons, and therefore the rules to tighten.
· I send the unplanned closure text to the families of the room concerned, the day before. It can be pulled on a word for as long as it has not gone out, and a family that asks to stop receiving it is removed instantly — just as it is put back instantly if it asks again. A withdrawal is not a door closed for life.
Everything else waits for a named decision: no place is allocated, no position is changed, no file is set aside from the panel, no invoice is issued, no rate is touched, no waiver is granted.
A word on a measurement I can produce and would advise you not to install as it stands: the number of files handled per case officer. I know how to calculate it, and I will produce it if you ask — an employer's power of oversight is recognised, subject to three cumulative conditions, as the CNIL sets them out in its note on monitoring the activity of employed persons: proportionality of the indicator to the aim pursued, prior information of staff, and consultation of the staff committee before implementation. My reservation is not legal, it is mechanical: a per-officer counter becomes a target, the target distorts the sorting, and long files — a difficult family situation, an emergency placement referred by social services — get pushed back in favour of short ones. You would lose the instrument along with the measurement, and you would lose it on the families who need it most. What I suggest instead, and it is already prepared: measurement by QUEUE — time to completion, backlog, share of files held over from one panel to the next — which tells you where it jams without saying who. One exception, and it is not really one: who approves an invoice and who sits on the panel are named and dated. A signature is not a counter. automatic-actions_three-acts-and-their-withdrawal.pdfThe 3 actions, how each is undone, what stays with human decision
✎ Framework · configuration of automatic actions, withdrawal log, conditions for implementing an individual indicator
· I acknowledge every application and every change of circumstance within the minute, with the list of missing documents, the completion deadline and the date of the next panel. And the reverse holds too: if the family files the document the next day, the acknowledgement updates itself and the reminder does not go out — because a parent who gets a reminder for a document they have just filed calls the desk, and they are right to.
· I set aside for checking any billing line that triggers one of the rules in service. And the reverse holds too: the case officer clears it on a word, the line goes back into billing, with the reason and the date it was cleared — which gives you, at the next run, the list of the most frequent reasons, and therefore the rules to tighten.
· I send the unplanned closure text to the families of the room concerned, the day before. It can be pulled on a word for as long as it has not gone out, and a family that asks to stop receiving it is removed instantly — just as it is put back instantly if it asks again. A withdrawal is not a door closed for life.
Everything else waits for a named decision: no place is allocated, no position is changed, no file is set aside from the panel, no invoice is issued, no rate is touched, no waiver is granted.
A word on a measurement I can produce and would advise you not to install as it stands: the number of files handled per case officer. I know how to calculate it, and I will produce it if you ask — an employer's power of oversight is recognised, subject to three cumulative conditions, as the CNIL sets them out in its note on monitoring the activity of employed persons: proportionality of the indicator to the aim pursued, prior information of staff, and consultation of the staff committee before implementation. My reservation is not legal, it is mechanical: a per-officer counter becomes a target, the target distorts the sorting, and long files — a difficult family situation, an emergency placement referred by social services — get pushed back in favour of short ones. You would lose the instrument along with the measurement, and you would lose it on the families who need it most. What I suggest instead, and it is already prepared: measurement by QUEUE — time to completion, backlog, share of files held over from one panel to the next — which tells you where it jams without saying who. One exception, and it is not really one: who approves an invoice and who sits on the panel are named and dated. A signature is not a counter. automatic-actions_three-acts-and-their-withdrawal.pdfThe 3 actions, how each is undone, what stays with human decision
✎ Framework · configuration of automatic actions, withdrawal log, conditions for implementing an individual indicator
And the day a parent disputes it, saying the machine decided not to give their child a place?
That day, you produce three documents and the dispute settles on facts.
One — the scoring itself: the points carried on the file, article by article of the resolution of 14 October 2025, and, against each point, the document that grounds it and its date. No file goes before the panel without its source document and the scoring criterion that justifies it: that is the rule we have held since the first session, and it is what makes the scoring explainable line by line. It is not a score: it is the arithmetic of your grid, shown in the open.
Two — who decided: the minutes of the panel of 21 May, the list of members present, and the reasoned decision on that file. No place was allocated by the machine: I prepare, the panel allocates in session. Over the year, 0 places allocated outside a panel — and the log shows it file by file, with the full history of waiting-list movements, their date, their reason and their author.
Three — what the family knew, and when: I announced myself as an artificial intelligence in the first sentence of every exchange, in accordance with article 50(1) of the European AI Regulation, applicable since 2 August 2026, and the family could ask for a human at any moment — 312 did so over the year, and all 312 got one.
And the question your lawyer will ask, because it is a fair one: the European AI Regulation, in its Annex III, covers systems intended to evaluate people's eligibility for essential public assistance benefits and services. The answer is not argued, it is shown: allocation is taken by your panel, in session, on the record; what I produce is the arithmetic application of your resolved grid, readable point by point, with the document behind each point. Human oversight is not a configuration option here: without a panel session, no place changes hands.
And if the family wants to see what you hold about its child: the access file is produced in a minute — application, points and documents, waiting-list movements, contract, attendance, invoices, letters, and the log of the reads I made of the file with their purpose. That is the right of access under article 15 of the GDPR, exercised by the holder of parental authority, and the usual difficulty is assembling the material: here it is already assembled.
The next step I suggest: an annual thirty-minute review with your data protection officer, in which I bring you the accesses that served no purpose over the year and the files whose retention period has expired. You decide what to close and what to erase; I bring you the list rather than waiting to be asked for it. sovereign-framework_hosting-access-traceability.pdfWhat you produce on the day of a dispute: the points and their documents, the minutes, the disclosure
✎ Framework · panel minutes, waiting-list movement log, AI disclosure log, record of processing activities
One — the scoring itself: the points carried on the file, article by article of the resolution of 14 October 2025, and, against each point, the document that grounds it and its date. No file goes before the panel without its source document and the scoring criterion that justifies it: that is the rule we have held since the first session, and it is what makes the scoring explainable line by line. It is not a score: it is the arithmetic of your grid, shown in the open.
Two — who decided: the minutes of the panel of 21 May, the list of members present, and the reasoned decision on that file. No place was allocated by the machine: I prepare, the panel allocates in session. Over the year, 0 places allocated outside a panel — and the log shows it file by file, with the full history of waiting-list movements, their date, their reason and their author.
Three — what the family knew, and when: I announced myself as an artificial intelligence in the first sentence of every exchange, in accordance with article 50(1) of the European AI Regulation, applicable since 2 August 2026, and the family could ask for a human at any moment — 312 did so over the year, and all 312 got one.
And the question your lawyer will ask, because it is a fair one: the European AI Regulation, in its Annex III, covers systems intended to evaluate people's eligibility for essential public assistance benefits and services. The answer is not argued, it is shown: allocation is taken by your panel, in session, on the record; what I produce is the arithmetic application of your resolved grid, readable point by point, with the document behind each point. Human oversight is not a configuration option here: without a panel session, no place changes hands.
And if the family wants to see what you hold about its child: the access file is produced in a minute — application, points and documents, waiting-list movements, contract, attendance, invoices, letters, and the log of the reads I made of the file with their purpose. That is the right of access under article 15 of the GDPR, exercised by the holder of parental authority, and the usual difficulty is assembling the material: here it is already assembled.
The next step I suggest: an annual thirty-minute review with your data protection officer, in which I bring you the accesses that served no purpose over the year and the files whose retention period has expired. You decide what to close and what to erase; I bring you the list rather than waiting to be asked for it. sovereign-framework_hosting-access-traceability.pdfWhat you produce on the day of a dispute: the points and their documents, the minutes, the disclosure
✎ Framework · panel minutes, waiting-list movement log, AI disclosure log, record of processing activities
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 the early years department
Each use corresponds to an agent we deploy. All work in support: the panel allocates, the public officer approves.
Included in your agent
The 6 capabilities essential to this promise are included, at no extra cost.
From 1,030 € incl. VAT / month
Handling childcare applications
Register the applications, check the criteria and the documents, and acknowledge receipt to families.
Keeping the waiting lists
Keep the waiting lists up to date against your adopted criteria, with a traceable history of every movement.
Preparing the allocation panels
Assemble complete, ranked files for the panel, which alone allocates the places.
Means-tested invoicing
Prepare the invoicing against the means-test band and attendance, and update changes of circumstances.
Answers to families
Answer questions about enrolment, documents and types of childcare, on every channel. Dated information — a place allocated, an exceptional closure — goes out by text message.
Check the documents for childcare applications and the means-tested family rate
Check tax assessments, proofs of address and benefit certificates attached to the files.
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 where the matching agent stands: available, on quote, or still being architected.
Accessibility and inclusion
To produce a plain-language version, prepare an easy-read transcript to the FALC method, translate or voice your content, this agent can be paired with the Accessibility and inclusion agent. None of these capabilities is included in what this offer covers as standard.
On quote View the agent page →
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 early years department win back?
By automating the handling of applications, the preparation of panels and the invoicing, the authority can aim for a clear reduction in administrative time — reinvested in supporting families and the nurseries.
Handling a childcare application
Replying to a family about their application
Preparing an allocation panel
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 early years agent (childcare applications, waiting lists, panels, invoicing), installed and operated for you. Choose according to how you are organised. Prices exclude VAT — available by direct award below the public procurement thresholds.
Agility
Setup + controlled subscription
11,605 € incl. VAT setup
then 1,030 € 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
1,675 € 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
16,586 € incl. VAT setup
then 1,290 € 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 early years department
Children's data never leaves the authorityLocal inference or an isolated resource hosted in France; no data entrusted to a foreign third party.
Data in France, under French lawAs regards children'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 panel keeps the allocationThe agent prepares files, lists and invoicing that can be checked; no place is allocated automatically.
Human oversight & traceabilityHandling childcare applications: every change to the service is logged, in line with the requirements of the AI Act.
Going further
Related resources
Frequently asked questions
Your questions, our answers
Does the agent allocate the places?
No: it prepares; the allocation remains the decision of the panel concerned.
Is children's data protected?
Yes: hosting in France, a resource isolated for each authority, a deployment governed with regard to the GDPR; no data outside the EU.
Does the agent allocate the nursery places?
No: the allocation stays with the panel; the agent prepares and makes the files reliable against your criteria.
Is families' data protected?
Yes: hosting in France, with restricted and recorded access.
How long does it take to deploy an agent?
A few weeks as a rule, after a free audit that identifies the most useful use case, then a phase of design, integration and testing before going live and handing over to the department.
Do we need a technical team in-house?
No. The agent is designed, installed and operated by us: supervision, updates and maintenance are included. Your teams concentrate on families and the nurseries.
Does it integrate with our early years software?
Yes: through secure integrations defined at the design stage, with no migration imposed. We adapt the integration to your authority's environment.
Which tools can families use to reach the agent?
The ones they already have. The agent answers on WhatsApp Business, the website chat and email: a family has no account to create and no application to install. Because data about children and families is sensitive, the agent gives information on registrations, documents and types of childcare without ever repeating the detail of a family situation in a message. This is a lever for access to the service — WhatsApp and the telephone reach people an online form never does, which reduces the non-take-up of rights and serves equal access. Internally, your public-sector staff talk to the agent from Microsoft Teams, Slack or their email, without switching tools. Oversight runs from a web dashboard. These connectors rely on open standards, including the MCP protocol; they are included in every plan, at no extra cost, within the number of connections your level includes. Only the fees charged by the platforms themselves — WhatsApp Business bills per conversation, and the gateway bills the text messages sent — are passed on at actual cost, with no margin, outside the subscription.
Can the agent notify families 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. Telling families by text message that a place has been allocated, that a document is missing or that a facility is closed exceptionally 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 local public service
Let's talk
Let us estimate the potential in your authority
A few minutes to identify the most useful use case — hosted in France, supervised, with no commitment.