+33 (0)1 87 66 00 65 · Monday to Friday, 9am–6pm Free audit (15 min)
● Public sector — Revenue office

AI agent for the revenue office

An assistant that prepares receipts, tracks collections and drafts reminders — the cashier remains solely responsible for handling the funds. The AI agent never touches the money: it prepares, tracks and flags; the cashier collects, checks and reports to the public accountant. Hosted in France, on a resource isolated for each authority.

Hosted in France The public's data protected GDPR & AI Act: governed deployment The cashier alone handles the funds

Updated on

Deployed in a few weeks
Revenue office assistant · hosted in France
Prepare the receipts for the school meals and after-school club payments entered this morning.
23 receipts prepared from the entries: identity of the payer, nature of the revenue, amount and means of payment, continuous numbering for the office. One possible duplicate flagged (same family, same amount, two entries).
For you to check and issue — I never touch the funds.
⛓ Source · entries made at the office + the rates adopted
Give me the position on the sums still to be collected and prepare this term's reminders.
Position established: 41 debts outstanding by nature of revenue, receipts reconciled against the entries — 2 discrepancies flagged for checking. 28 draft informal reminders prepared from your templates, with amounts taken from your statements.
All submitted for your check before transmission to the public accountant.
✎ Action · statement & reminders to check — the cashier 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 cashier handles the fundsThe agent prepares, it never collects
✦ In brief

In a revenue office, a Blue Lemon Agent agent assists the cashier with the repetitive tasks — preparing receipts and revenue orders, tracking collections by nature of revenue, reconciling entries, reminders and statements of outstanding sums — and prepares the schedules for the public accountant. It runs on local inference or is hosted in France on a resource isolated for each authority, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity. The AI agent never handles the funds: collection and the cashier's personal financial liability remain theirs. Live within a few weeks. Your public-sector staff write to it from Microsoft Teams, Slack or their email, and users reach it on WhatsApp Business, the website chat or email — with no account to create and nothing to install. Reaching the administration from the tool people already have means less non-take-up of rights and equal access to the service. These connections are included in every plan, at no extra cost, within the number of connections your level includes.

100%
hosted in France in the target architecture
0
transfer outside the EU in the target architecture
7
uses ready to deploy on this scope
0
decision taken without human approval

Reference points describing our offer, not results measured at a client. The scale of the gain is confirmed by a pilot on your own scope.

The context

Why AI matters to revenue offices — and why they hesitate

School meals, after-school clubs, swimming pools, market pitch fees: the cashier collects, issues receipts, reconciles and reports to the public accountant, often on top of other duties. Their personal financial liability leaves no room for approximation.

! The issue

The revenue cashier is caught between demanding formalities — receipt books, reconciliations, schedules, statements of outstanding sums — and limited time, since the revenue office is often only part of the job. Yet most consumer AI tools amount to entrusting payers' identities, family situations and the authority's financial data to a third party, often hosted outside Europe and subject to the Cloud Act.

Our answer

For a revenue office, AI is only of interest if it is sovereign, confidential and strictly bounded. Local inference or an isolated resource hosted in France, systematic human oversight, a scope limited to preparation and tracking: the AI agent never handles the funds, collects nothing and decides nothing — the cashier's liability stays whole, but relieved of the paperwork. The time won back goes to checking and to serving the public.

The decisive point

Protecting the revenue office's data: sovereignty & compliance

A revenue office handles personal and financial data: payers, means-tested rates, collections. 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 — payers' 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 payers' 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 and its revenue offices.

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 collection and no handling of funds; 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

5 real situations, taken from those that come up most often. Pick one: the exchange unfolds as it would in your organisation.

A scripted demonstration. These exchanges show how the agent behaves — its sources, its refusals, what it leaves to your teams. Nothing is sent from this page, no model is queried here, and the matters named are fictional. That is precisely what we promise your data.
The behaviours shown here — monitoring, automation rules, routing and reminders — are configured with you during deployment, from your tools, your rules and your thresholds.
The architecture points named in these exchanges — location, local execution, isolation, encryption, role-based access, logging — are not a guarantee attached to the demonstration: they are those of the architecture set out in your quotation, and verified before commissioning.

The public body in this demonstration

Fictional public body

Municipality of Marnoy — four revenue offices, 9,800 residents

Sector
Municipality of 9,800 residents — four revenue offices: school catering, after-school care, the municipal swimming pool and market pitch fees
Headcount
1 appointed revenue manager and 2 deputies — the revenue office is only part of their post; the manager also staffs the town hall reception two days a week
Public served
9,800 residents, 1,240 families enrolled in the canteen or after-school care, 46 market traders — 2,600 user enquiries a year
Order of magnitude
18,400 payments taken a year for €412,000 — 14,900 receipts issued, 210 operating days, 12 monthly remittance slips to the public accountant, 980 amicable reminders
Tools in place
School billing software, card terminal, family portal and five years of archived receipt books — the agent plugs into them read-only, nothing is replaced and nothing is migrated
Who decides
The revenue manager alone handles the money and issues the receipts; the public accountant receives the remittance slips and checks them; the authorising officer applies the rates adopted by the council
Room for improvement
Preparing the receipts takes 1 h 36 out of the 2 h 40 the office consumes each day; the statement of outstanding revenue and the month's remittance slips take a whole day; 68 reconciliation discrepancies were found only at year-end close; and €39,600 remained outstanding over three financial years, 61 % of it on families that had never been sent a reminder

Fictional public body, invented for the demonstration. At Marnoy, one person runs four revenue offices on top of the town hall reception: 18,400 payments a year, 14,900 receipts, twelve remittance slips to the public accountant. The revenue manager is personally answerable for the money he handles, and it is that responsibility which rules out any approximation in the receipt book as in the reconciliation. The agent runs on local inference on a machine at the town hall, reads the billing software and five years of receipt books without ever writing to them, and touches no money. The exchanges below span a year, from the overhaul of the discrepancy causes to the review presented to the council.

This public body, its figures and the exchanges that follow were invented for the demonstration. They illustrate a common situation; they describe no real service.

Revenue office assistant · hosted in France
I reread your last five financial years over the weekend — 92,000 payments taken, 18,400 a year, across four revenue offices — and I have drawn out a finding your receipt books were carrying without anyone having the time to establish it.
A revenue office allows an officer to take certain payments directly on the municipality's behalf, and requires him to remit them to the public accountant. The revenue manager is personally answerable for the money he handles: a cash shortfall is charged to him until he is discharged of it. It is that responsibility which makes an exhaustive, traced check more protective than a spot check — and that is exactly what I bring you.
The gap measured, and it decides everything else: your five financial years carry 1,240 reconciliation discrepancies a year, and four causes alone carry 1,004 of them — 81 %.
· Payment entered against the wrong revenue office: 412 a year, almost all between canteen and after-school care, the two services the same families pay for on the same day.
· Receipt issued with no matchable invoice reference: 268.
· Apparent duplicate payment: 248.
· Gap between the receipt book total and the amount remitted to the public accountant: 76, always below €15 and always regularised — but each one cost a search.
What that costs today:
· 68 of those discrepancies were found only at the year-end close, when the original document can no longer be consulted and only the adjusting entry remains.
· Preparing a day's receipts accounts for 60 % of the office's time: 1 h 36 out of the 2 h 40 the day consumes, from the first payment to the filing of the receipt book.
· And €39,600 remained outstanding over three financial years, 61 % of it — €24,156 — on families that never received a single reminder.
What I propose, and it is not a promise — it is already done: I ran the four checks over the 18,400 payments of the past year, as a dry run. They would have found 1,004 of the 1,240 discrepancies on the day itself, against 68 discovered at the closesame work, with the original document still to hand.
The time this shifts: preparing the receipts goes from 60 % to 10 % of the operating day — from 1 h 36 to 16 minutes. Over 210 operating days a year, that is 280 hours returning to the service.
The next step, and it takes half an hour: you read the four written checks, cause by cause, with the number of discrepancies each would have found. You settle them, and they run from Monday's operating daybringing them into service is the only act I leave to you. causes-of-discrepancy_1004-out-of-1240.pdf4 causes behind 81 % of discrepancies, each measured over the year
⛓ Sourced · 5 financial years of receipt books, 92,000 payments, reconciliation log, statements of arrears
I run four revenue offices and the town hall reception. I am not going to open four projects: where do we start?
With two checks, not four — and I have ranked them by the number of discrepancies they spare you, not by difficulty.
The ranking, with the count behind it:
· The allocation check: 412 discrepancies a year, that is a third of all your discrepancies on its own. It matches the payment, the enrolled child, the service used and the adopted rate. A family whose two children have lunch and go to after-school care pays the same amount twice on the same day; that is where the entry goes to the wrong office.
· The invoice-matching check: 268 discrepancies a year. A receipt with no matchable reference is a receipt that will not be found at reconciliation, and it is the one that costs an evening at month-end.
· The duplicate check: 248 discrepancies, and that is the one I suggest opening last — I will tell you why once it has proved itself.
· The remittance gap: 76 a year, all below €15 and all regularised.
These first two carry 680 of the 1,240 discrepancies: two checks out of four, 55 % of the search work. The other two can wait until the end of the first quarter without changing anything to your workload.
What I do on top, and that nobody has time for: on the allocation check, I do not merely flag the gap — I reread your 412 cases and wrote the rule that covers them all, in the form of your office instructions: the service is deduced from the child's attendance slot, not from the amount. Run over the year, it files 396 of the 412 payments correctly and sends 16 back to your check, all of them days when the child was recorded at both services.
You judge on a written and already tested ruleand you correct it in one word, since it is your own material that I have set out clearly.
The next step I propose: let the two checks run for a fortnight alongside your usual verification. You compare the two outputs line by line; if mine add nothing, they stop at a word. On the trial day already run, they found 6 discrepancies out of 94 payments, and all 6 would have been discovered at the month-end close. control-plan_two-checks-for-55-percent.pdf4 causes measured, 2 checks to open, the allocation rule written
⛓ Sourced · count of discrepancies by cause over 5 financial years, written allocation rule, trial day of 94 payments
All of that assumes you read families' names, their means-tested bands and what they pay. Where exactly does that go?
It goes nowhere. I run on local inference on a machine at the town hall, and nothing I read leaves the municipality's network.
Local inference means the model computes on your own machine: a family's name or their means-tested band crosses no external network to be processed. If the municipality would rather not host a machine, the other route is an isolated resource hosted in France, dedicated to Marnoyno pooling with another authority, which is the condition for the continuity of your revenue offices.
What that changes, point by point:
· Family data trains no model, neither ours nor a third party's.
· I work read-only, and the technical account I read through has no right to writeneither in the receipt book nor in the billing software. This is not a promise, it is a permission that can be checked with one command.
· Encryption in transit and at rest, role-based accessrights follow the job: a deputy opens the office he runs, not the other three. 3 roles for your 3 authorised 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.
And the point that decides, for someone personally answerable for his own cash: I never touch the money, and I have no technical means of doing so. Taking the payment, issuing the receipt and remitting to the public accountant are your acts, and they stay yoursI prepare them, I reconcile them and I trace them. What that brings you is not lighter responsibility, it is what makes it bearable: an exhaustive, dated and enforceable check across all 18,400 payments, instead of a spot check made at the end of the day. The day the public accountant or the regional audit chamber asks how you ensure that the totals agree, you produce the log, not your good faith.
The figure that sums all this up: 0 family data out of the municipality's network across the 18,400 payments of the year, and processing in the EU targeted.
What I propose: that I keep up to date the record your data protection officer and your public accountant will both ask for — hosting, data processed, retention periods, who accesses what, and the exact permissions of the technical account. It is asked for once a year and takes two days to rebuild; the first version is written and you have it attached. technical-framework_where-revenue-office-data-lives.pdfLocal inference, read-only, no access to the money
✎ Framework · deployment architecture, technical account permissions, access log, first version of the processing record
Local inference · no data outside the EU

Your case is not here? That is exactly what a 15-minute conversation is for. Book the free audit

Use cases

The uses of AI in a revenue office

Each use corresponds to an agent we deploy. All work in support, under the cashier's control — and the cashier alone handles the funds.

Included in your agent The 7 capabilities essential to this promise are included, at no extra cost.
From 775 € incl. VAT / month

Receipts & revenue orders

Prepare the receipts and revenue orders from the entries, with continuous numbering, to be checked and issued.

Tracking payments received

Tracks incoming payments by revenue category, reconciles entries and flags discrepancies to the responsible revenue officer.

Reminders & letters to the public

Draft informal reminders and answers to residents about their position, for approval before sending.

Statements of outstanding sums

Prepare the statement of sums still to be collected and the schedules for the public accountant, ready to check.

Flagging variances

Detect duplicates, reconciliation gaps and entry anomalies, and flag them for a human check.

Answers to the public

Inform residents about rates, invoices and how to pay, around the clock and across channels.

Help with paying online

Guide residents towards online payment and reduce collections at the counter.

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 Reconcile supporting documents, rules and accounting entries with an audit trail Handle discrepancies and validations before any accounting entry or payment
What the agent must be connected to This connection is required for the agent to work. It concerns your information system and is scoped during the audit.
Integrate with accounting and banking software and with approved platforms (French e-invoicing)
The gain

How much time can a revenue office win back?

By automating the preparation of receipts, the reconciliation of entries and the reminders, a cashier can aim for a clear reduction in the office's administrative time — reinvested in checking and in their other duties.

Preparing a day's receipts
Today · done by hand
Prepared by the agent, to approve
An informal reminder to a resident
Today · done by hand
Near-instant
Statement of outstanding sums and the month's schedules
Today · done by hand
Prepared by the agent, to approve
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 agent supporting the revenue office (receipts, tracking collections, reminders, statements), installed and operated for you. Choose according to how you are organised and how demanding your security requirements are.

Agility

Setup + controlled subscription

9,355 € incl. VAT setup
then 775 € 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
Order →
The simplest Serenity

All inclusive, no setup fee

1,295 € 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
Order →
100% Sovereign

On site, you own it

13,801 € incl. VAT setup
then 1,015 € 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)
Order →
Not included in the packages: AI consumption (model tokens), re-invoiced at real cost with no margin, and tracked in real time in your client area. Maintenance and supervision subscription for an initial term of 12 months for the Agility package, 24 months for the Serenity package and 36 months for the 100% Sovereign package, renewable; support levels (SLA 72 h / 24 h / 4 h) optional. Bespoke development, additional integrations or exceptional volumes are quoted separately. Support Monday to Friday, 9am to 6pm. Prices include VAT at 20%: as a public body that is not VAT-registered, you cannot reclaim it.
AI model: none of the AI models offered currently carries a fixed surcharge. When the selected model carries a cost, that cost is shown when you choose it, before you order, and re-invoiced at the cost incurred, with no mark-up; usage is billed at the publisher's price. Publishers' prices are published in US dollars: the amount re-invoiced is the amount in euros actually borne by Blue Lemon Agent on the publisher's invoice, at that invoice's exchange rate, with no commission or mark-up.
Included components and additional components Components included in the base offer: the Blue Lemon Agent software foundation, the AI models listed in the order journey, the standard channels (Microsoft Teams, Slack, WhatsApp Business, email, website chat, calendars, Microsoft 365 / Google Workspace, file storage, market VoIP telephony, professional social-media pages and accounts, Google Business Profile), hosting in France for the package chosen, backups, supervision, updates and support. If adapting the AI agent to your constraints, your needs or your requests requires other paid components — a third-party publisher's software licence, paid API access to one of your applications, hosting of health data, for which French law requires an HDS-certified host (art. L. 1111-8 of the French Public Health Code), SecNumCloud-qualified hosting, a speech synthesis service, particular hardware —, they are offered to you as an option or on quotation and re-invoiced at the cost incurred; nothing is committed without your written agreement. Where the artificial intelligence model you choose entails an additional cost, that cost is shown to you before you order and re-invoiced to you at the cost incurred, with no margin.
What to expect
Go-live 2 to 3 weeks
Agent designed, channels connected, team trained.
Steady state 4 to 7 weeks
After a few weeks of real use, once the agent's behaviour matches what you expect. Indicative estimate, adjusted to the options you keep. It is not a delivery commitment.
Our commitment

Four guarantees that matter to a revenue office

Payers' 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 lawPayers' data: minimisation and location in France, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The cashier alone handles the fundsThe AI agent prepares receipts, statements and reminders that can be checked; it collects nothing and decides nothing.
Human oversight & traceabilityFor receipts & revenue orders: systematic updates and logging, compliant with the AI Act.
Frequently asked questions

Your questions, our answers

Does the agent handle the funds?
No, never: it prepares and tracks administratively; collection and the handling of funds remain the cashier's responsibility.
Is it compatible with our revenue offices and M57?
Yes: secure integration with your revenue office and public accounting tools, defined at the design stage.
Does it integrate with the revenue office and the public accountant?
Yes: through secure exchanges defined at the design stage (revenue office, treasury, financial software).
Does it produce the supporting documents for the accountant?
Yes: it prepares the statements and schedules, which the cashier checks and passes to the public accountant.
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 free audit that identifies the most useful use case, then a phase of design, integration and testing before going live.
Do we need a technical team to run it?
No: the agent is designed, installed and operated by us. The cashier uses it from their usual environment, after a few hours of familiarisation.
Where is the revenue office's data hosted?
In France, on a resource dedicated to and isolated for each authority. GDPR: governed deployment. Depending on the need, the agent can also run on local inference, with no data leaving the network.
Which tools can users use to reach the agent?
The ones they already have. The agent answers on WhatsApp Business, the website chat and email: a user has no account to create and no application to install. The agent gives information on a price, an invoice or the ways to pay, but no money is taken in the conversation: handling funds stays with the collection officer, and users pay through the means you have opened. This is a lever for access to the service — WhatsApp and the telephone reach people an online form never does, which reduces the non-take-up of rights and serves equal access. Internally, your public-sector staff talk to the agent from Microsoft Teams, Slack or their email, without switching tools. Oversight runs from a web dashboard. These connectors rely on open standards, including the MCP protocol; they are included in every plan, at no extra cost, within the number of connections your level includes. Only the fees charged by the platforms themselves — WhatsApp Business bills per conversation — are passed on at actual cost, with no margin, outside the subscription.
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.