An orchestration agent, scoped with you before it is priced. This agent coordinates several specialised agents. Its orchestration follows your actual workflows — which is why it is scoped with you rather than bought off the shelf. We establish the scope together, then the quotation commits it.
The journeys described below form the scope that this review refines and the quotation commits. The specialised agents it coordinates can be ordered today.
Request a quote
● Public sector — Public accounting (M57)
The AI agent for public accounting: more reliable entry, the decision stays yours
Keying in entries, checking consistency against the M57 chart of accounts, preparing reminders and payment schedules take up a considerable share of public accountants' time — without being the heart of the job, which is accounting quality and continuity of service. Your AI agent absorbs that repetitive work. Hosted in France — on local inference or an isolated resource — public financial data stays under control. The public accountant keeps the decision.
Hosted in France
Public financial data protected
GDPR & AI Act: governed deployment
Human oversight
Updated on
Deployed in a few weeks
Public accounting assistant · hosted in France
On the main budget, process this batch of 38 payment and revenue orders and flag the M57 discrepancies before transmission to the assigned accountant.
38 documents read. 34 consistent with the M57 chart of accounts. 4 anomalies: one posting to 6068 instead of 60632, one payment order with no supporting invoice attached, one duplicate (order no. 2231 already issued) and VAT wrongly apportioned on a service.
Corrected entries proposed — to check and approve.
⛓ Source · your financial management software + the file's documents
Corrected entries proposed — to check and approve.
⛓ Source · your financial management software + the file's documents
The order with no supporting document — how can we put it right?
Correction: the supplier's invoice is missing to evidence that the service was delivered. Recommendation: suspend the payment order and request the document from the managing department before transmission, to avoid a rejection by the assigned accountant.
I am preparing the document request and recording it in the file, for your approval.
✎ Action · request ready for review — the public officer approves
I am preparing the document request and recording it in the file, for your approval.
✎ Action · request ready for review — 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 public officer decidesThe AI agent assists, never approves alone
✦ In brief
In a local authority or a public body, a Blue Lemon Agent agent assists public accountants with repetitive tasks — entering and checking the consistency of M57 entries, detecting anomalies and discrepancies, preparing reminders and payment schedules. It runs on local inference or is hosted in France: public financial data is never exposed to a foreign service, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity. The time saved is redirected towards accounting quality and continuity of service. Live within a few weeks.
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 public finance departments — and why they hesitate
Public accountants have to guarantee accounting quality, the regularity of M57 entries and continuity of service, often with constrained headcount. But available time is mechanically reduced by data entry and checks — and the data involved (budgets, contracts, pay, third-party accounts) is among the most sensitive there is.
! The issue
The finance department is caught between rising demands — reliable accounts, compliance with M57, payment deadlines to suppliers — and an ever-growing production workload (entry, checks, anomalies, reminders). Yet most consumer AI solutions amount to entrusting budgets, contracts, pay and third-party accounts to a third party, often hosted outside Europe and subject to the Cloud Act.
✓ Our answer
For public 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, decisions reserved to the public officer: the time saved on production is never paid for in lost confidentiality, nor in a breach of equal treatment. The aim is not to replace the public accountant, but to give them back time for accounting quality and continuity of service.
The decisive point
Confidentiality of public financial data: sovereignty & compliance
A public finance department handles the authority's most sensitive data. Here is how the architecture of our agents protects it, file by file.
Local inference
The agent can run on a machine belonging to the authority: no document leaves the network, nothing passes through a cloud.
Hosting in France
Otherwise, a dedicated and isolated resource, hosted in France under French law — your data: processing and access within the European Union targeted by the architecture.
Reduced extraterritorial exposure
Exposure of financial data to the Cloud Act and FISA 702 is reduced by design; location alone does not guarantee immunity.
One isolated resource per organisation
No pooling of financial data: an environment strictly dedicated to your authority or institution.
Encryption & controlled access
Encryption in transit and at rest, role-based access (RBAC), strong authentication and logging.
AI Act: governed deployment
An agent strictly in support; no entry approved 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
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 bodyTown of Roncelles — municipality of 22,400 residents
- Sector
- Municipality of 22,400 residents — one main budget and three ancillary budgets (water, sewerage, school catering), accounts kept under the M57 chart of accounts
- Headcount
- 6 staff in the finance department — one head of department, three accounting officers, one contracts officer, one revenue officer; these six people alone run the execution of the four budgets
- Public served
- 22,400 residents, 1,180 active suppliers and 46 spending departments — 3,800 enquiries from third parties and departments each year
- Order of magnitude
- 18,400 payment orders and 6,200 revenue orders a year — 24,600 documents split into 78 payment batches, €41m of actual expenditure, 12 call-off contracts in force
- Tools in place
- Financial management software, the e-invoicing platform, document management and three years of archived entries — the agent plugs into them read-only, nothing is replaced and nothing is migrated
- Who decides
- The authorising officer commits and issues the payment order; the assigned public accountant pays it; the head of the finance department approves every entry before it is sent
- Room for improvement
- The overall payment period stands at 34 days where the rule applying to local authorities sets 30; 7,544 payment orders out of 18,400 go back at least once to the spending department — 41 %; the statement of outstanding revenue comes out six weeks after year-end close; and €214,000 of revenue more than two years old has never been chased
Fictional public body, invented for the demonstration. At Roncelles, six people run the execution of four budgets: 24,600 documents a year, twelve call-off contracts, and a year-end close that ties up the whole department for six weeks. Time is not slightly short, it is very short — and the data at stake, contracts, payroll and third-party accounts, is among the most sensitive the municipality holds. The agent runs on local inference on a machine at the town hall, reads the financial software without ever writing to it, and lets no entry leave without the head of department's approval. The exchanges below span a year, from the overhaul of the return causes to the review presented to the finance councillor.
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.
Public accounting assistant · hosted in France
I reread your last three financial years over the weekend — 73,800 documents, 24,600 a year — and I have drawn out a finding your software was carrying without anyone having the time to extract it.
The gap measured, and it decides everything else: 7,544 of your 18,400 payment orders go back at least once to the spending department before being sent — 41 %. And four causes alone carry 5,884 of them, that is 78 %. These are not forty-one different failures: they are four checks missing at the entrance that are then redone at the exit.
What that back-and-forth costs, on figures that are your own:
· The overall payment period stands at 34 days, where the rule applying to local authorities sets 30.
· 1,140 invoices paid beyond the deadline last year, and €9,400 of late-payment interest paid out — public money that funded no service at all, neither for residents nor for staff.
· Checking a batch of payment and revenue orders accounts for 60 % of the time the batch takes: 4 hours out of the 6 h 40 a batch of 312 documents takes, from filing to dispatch.
What I propose, and it is not a promise — it is already done: I ran the four checks over the 78 batches of the past year, as a dry run. They would have stopped 5,884 documents at the entrance, where today they are stopped at the exit — same work, six days earlier on average, and without the second pass.
The time this shifts, once the checks sit at the entrance: checking a batch goes from 60 % to 18 % of the batch's time — from 4 hours to 1 h 12. Over 78 batches a year, that is 218 hours returning to the department.
The next step, and it takes one meeting: your head of department reads the four written checks, cause by cause, with the number of documents each would have stopped. She settles them, and they run on the next batch — bringing them into service is the only act I leave to you, and it is the one that binds the municipality. causes-of-return_7544-orders-out-of-18400.pdf4 causes behind 41 % of returns, each measured over the year
⛓ Sourced · financial management software, 3 financial years, 73,800 documents, log of returns to spending departments
The gap measured, and it decides everything else: 7,544 of your 18,400 payment orders go back at least once to the spending department before being sent — 41 %. And four causes alone carry 5,884 of them, that is 78 %. These are not forty-one different failures: they are four checks missing at the entrance that are then redone at the exit.
What that back-and-forth costs, on figures that are your own:
· The overall payment period stands at 34 days, where the rule applying to local authorities sets 30.
· 1,140 invoices paid beyond the deadline last year, and €9,400 of late-payment interest paid out — public money that funded no service at all, neither for residents nor for staff.
· Checking a batch of payment and revenue orders accounts for 60 % of the time the batch takes: 4 hours out of the 6 h 40 a batch of 312 documents takes, from filing to dispatch.
What I propose, and it is not a promise — it is already done: I ran the four checks over the 78 batches of the past year, as a dry run. They would have stopped 5,884 documents at the entrance, where today they are stopped at the exit — same work, six days earlier on average, and without the second pass.
The time this shifts, once the checks sit at the entrance: checking a batch goes from 60 % to 18 % of the batch's time — from 4 hours to 1 h 12. Over 78 batches a year, that is 218 hours returning to the department.
The next step, and it takes one meeting: your head of department reads the four written checks, cause by cause, with the number of documents each would have stopped. She settles them, and they run on the next batch — bringing them into service is the only act I leave to you, and it is the one that binds the municipality. causes-of-return_7544-orders-out-of-18400.pdf4 causes behind 41 % of returns, each measured over the year
⛓ Sourced · financial management software, 3 financial years, 73,800 documents, log of returns to spending departments
There are six of us for four budgets. I am not going to open four projects at once: where do we start?
With three checks, not four — and I have ranked them by the number of documents they spare you, not by difficulty.
M57 is the chart of accounts for local authorities: it states which account every item of expenditure and revenue must be posted to. Certified service is the confirmation that the work has actually been delivered before the expense is paid — the check that prevents paying for what has not been received. The assigned public accountant is the one who pays the expense the authorising officer has ordered, and who refuses it if a document is missing.
The four causes, counted over the year:
· Certified-service document missing or insufficient: 2,960 payment orders. It is the leading cause of rejection by the assigned public accountant, by a wide margin.
· Posting outside the M57 chart of accounts: 1,610 payment orders — of which 612 on a single pair of accounts, cleaning services posted as supplies.
· Duplicate invoice: 780 payment orders, 71 % of them on the twelve call-off contracts.
· Incorrect VAT split: 534 payment orders, almost all on the water and sewerage ancillary budgets.
These first three carry 5,350 of the 7,544 returns: three checks out of four, 71 % of the back-and-forth. The fourth 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 posting cause, I do not merely flag the gap — I reread your 612 cleaning payment orders and wrote the posting rule that covers them all, in the form of your internal guidance, with the account used and the account retained side by side. You judge on a written, measured rule, not on 612 cases to settle one by one.
The next step I propose: let the three checks run on the next batch, alongside your usual check, for a fortnight. You compare the two outputs line by line: if mine add nothing, they stop at a word. On the trial batch I have already run, they stopped 25 documents out of 312, and 22 of the 25 would have come back from the spending department ten days later. control-plan_three-checks-for-71-percent.pdf4 causes measured, 3 checks to open, the posting rule written
⛓ Sourced · count of returns by cause over 3 financial years, written posting rule, trial batch of 312 documents
M57 is the chart of accounts for local authorities: it states which account every item of expenditure and revenue must be posted to. Certified service is the confirmation that the work has actually been delivered before the expense is paid — the check that prevents paying for what has not been received. The assigned public accountant is the one who pays the expense the authorising officer has ordered, and who refuses it if a document is missing.
The four causes, counted over the year:
· Certified-service document missing or insufficient: 2,960 payment orders. It is the leading cause of rejection by the assigned public accountant, by a wide margin.
· Posting outside the M57 chart of accounts: 1,610 payment orders — of which 612 on a single pair of accounts, cleaning services posted as supplies.
· Duplicate invoice: 780 payment orders, 71 % of them on the twelve call-off contracts.
· Incorrect VAT split: 534 payment orders, almost all on the water and sewerage ancillary budgets.
These first three carry 5,350 of the 7,544 returns: three checks out of four, 71 % of the back-and-forth. The fourth 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 posting cause, I do not merely flag the gap — I reread your 612 cleaning payment orders and wrote the posting rule that covers them all, in the form of your internal guidance, with the account used and the account retained side by side. You judge on a written, measured rule, not on 612 cases to settle one by one.
The next step I propose: let the three checks run on the next batch, alongside your usual check, for a fortnight. You compare the two outputs line by line: if mine add nothing, they stop at a word. On the trial batch I have already run, they stopped 25 documents out of 312, and 22 of the 25 would have come back from the spending department ten days later. control-plan_three-checks-for-71-percent.pdf4 causes measured, 3 checks to open, the posting rule written
⛓ Sourced · count of returns by cause over 3 financial years, written posting rule, trial batch of 312 documents
All of that assumes you read our contracts, our payroll and our third-party accounts. Where exactly do they go?
They go 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: the amount of a contract or a member of staff's pay line crosses no external network to be processed. If the authority would rather not host a machine, the other route is an isolated resource hosted in France, dedicated to Roncelles — no pooling with another authority, which is the condition for the continuity of your finance department.
What that changes, point by point:
· Your financial data trains no model, neither ours nor a third party's.
· I work read-only on the software, and the technical account I read through has no right to write — this is not a promise, it is a permission that can be checked with one command, and your IT department will run it in front of you.
· Encryption in transit and at rest, role-based access — rights follow the job: the contracts officer opens commitments, not payroll statements. 6 roles for your 6 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 is worth more than all the rest for a finance department: I serve the separation between the authorising officer and the public accountant, I never bypass it. The one who decides the expense is not the one who pays it: two people, two checks, and that is what has protected public money for two centuries. Everything leading up to the payment order, I have already done — the entry built on your chart of accounts, the certified-service document matched, the control note stating what is compliant and what is missing. Issuing the payment order stays with the authorising officer, payment with the public accountant — and it is precisely because these two acts stay with two different people that an exhaustive check weakens no one: it arrives before them, it takes neither one's place.
The figure that sums all this up: 0 financial data out of the municipality's network across the 24,600 documents 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-financial-data-lives.pdfLocal inference, read-only, processing in the EU targeted
✎ Framework · deployment architecture, technical account permissions, access log, first version of the processing record
Local inference means the model computes on your own machine: the amount of a contract or a member of staff's pay line crosses no external network to be processed. If the authority would rather not host a machine, the other route is an isolated resource hosted in France, dedicated to Roncelles — no pooling with another authority, which is the condition for the continuity of your finance department.
What that changes, point by point:
· Your financial data trains no model, neither ours nor a third party's.
· I work read-only on the software, and the technical account I read through has no right to write — this is not a promise, it is a permission that can be checked with one command, and your IT department will run it in front of you.
· Encryption in transit and at rest, role-based access — rights follow the job: the contracts officer opens commitments, not payroll statements. 6 roles for your 6 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 is worth more than all the rest for a finance department: I serve the separation between the authorising officer and the public accountant, I never bypass it. The one who decides the expense is not the one who pays it: two people, two checks, and that is what has protected public money for two centuries. Everything leading up to the payment order, I have already done — the entry built on your chart of accounts, the certified-service document matched, the control note stating what is compliant and what is missing. Issuing the payment order stays with the authorising officer, payment with the public accountant — and it is precisely because these two acts stay with two different people that an exhaustive check weakens no one: it arrives before them, it takes neither one's place.
The figure that sums all this up: 0 financial data out of the municipality's network across the 24,600 documents 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-financial-data-lives.pdfLocal inference, read-only, processing in the EU targeted
✎ Framework · deployment architecture, technical account permissions, access log, first version of the processing record
Here is the fortnight's batch: 312 payment and revenue orders, to be sent on Friday. We will not have time to go through it all by hand.
All 312 have been read, 287 are consistent with your chart of accounts and your documents, and the remaining 25 come back with their corrected entry already proposed.
What I checked on each one: the existence and legibility of the supporting document, the match of the amount between invoice, commitment and payment order, the posting against M57 and your internal guidance, the VAT split, the presence of the certified-service confirmation, and the uniqueness of the document within the financial year.
The 25, by cause:
· 11 — certified-service document missing or insufficient. 7 are waiting on a certificate, 4 carry an unsigned delivery note.
· 6 — posting outside the chart of accounts. 4 fall under the cleaning rule you settled last month, and are corrected automatically by that rule, with the rule cited on the entry; 2 are new cases and are waiting for you to read them.
· 4 — apparent duplicates. 3 are confirmed (same supplier, same amount, same call-off order); 1 is not: it is a monthly instalment under a call-off contract, and I tell you why I still raised it.
· 3 — VAT split on the sewerage ancillary budget.
· 1 — revenue order issued against a payer whose account has been settled since February.
What each anomaly carries, and it is what saves you the review: the document alongside, the exact line at fault, the corrected entry proposed, and the reason written in plain language — the very one that will go to the spending department if you approve it.
The time this shifts: checking the batch goes from 60 % to 18 % of the batch's time — from 4 hours to 1 h 12. Over 78 batches a year, that is 218 hours returned to the department. And what remains inside that 1 h 12 is your reading of the 25: the only work where the municipality commits its signature.
The next step: the 287 compliant documents are ready to send as soon as you sign off, and I propose that the 4 corrections made automatically under a settled rule be presented to you grouped, on one line, rather than one by one. Over the year's 78 batches, that is 312 entries that stop asking for a decision already taken. payment-batch_312-documents-25-anomalies.pdf287 compliant, 25 corrections proposed, cause by cause
⛓ Sourced · batch of 312 documents, commitments and call-off orders, documents filed on the e-invoicing platform
What I checked on each one: the existence and legibility of the supporting document, the match of the amount between invoice, commitment and payment order, the posting against M57 and your internal guidance, the VAT split, the presence of the certified-service confirmation, and the uniqueness of the document within the financial year.
The 25, by cause:
· 11 — certified-service document missing or insufficient. 7 are waiting on a certificate, 4 carry an unsigned delivery note.
· 6 — posting outside the chart of accounts. 4 fall under the cleaning rule you settled last month, and are corrected automatically by that rule, with the rule cited on the entry; 2 are new cases and are waiting for you to read them.
· 4 — apparent duplicates. 3 are confirmed (same supplier, same amount, same call-off order); 1 is not: it is a monthly instalment under a call-off contract, and I tell you why I still raised it.
· 3 — VAT split on the sewerage ancillary budget.
· 1 — revenue order issued against a payer whose account has been settled since February.
What each anomaly carries, and it is what saves you the review: the document alongside, the exact line at fault, the corrected entry proposed, and the reason written in plain language — the very one that will go to the spending department if you approve it.
The time this shifts: checking the batch goes from 60 % to 18 % of the batch's time — from 4 hours to 1 h 12. Over 78 batches a year, that is 218 hours returned to the department. And what remains inside that 1 h 12 is your reading of the 25: the only work where the municipality commits its signature.
The next step: the 287 compliant documents are ready to send as soon as you sign off, and I propose that the 4 corrections made automatically under a settled rule be presented to you grouped, on one line, rather than one by one. Over the year's 78 batches, that is 312 entries that stop asking for a decision already taken. payment-batch_312-documents-25-anomalies.pdf287 compliant, 25 corrections proposed, cause by cause
⛓ Sourced · batch of 312 documents, commitments and call-off orders, documents filed on the e-invoicing platform
The payment order with no supporting document is the one that costs us the most time. How do we put it right?
The request for the document is written, addressed to the right spending department, and it goes out with the payment order suspended rather than sent.
What the request carries, and it is why it comes back quickly: the payment order number, the supplier, the amount, the exact document that is missing — certificate of certified service, signed delivery note, acceptance report —, the name of the officer in the spending department who committed it, and the date the invoice arrived. A request saying “document missing” comes back in eleven days; a request saying “the signed delivery note of 14 March for order no. 2231 is missing” comes back in three.
Why I suspend rather than send: a missing document means a rejection by the public accountant a fortnight later, and the payment-period clock keeps running all that time. Suspending costs three days; sending costs fifteen days and a rejection.
What that gave, measured over the quarter: 740 document requests prepared — a quarter of the year's 2,960 —, 681 back within eight days, and 0 rejections by the assigned public accountant for a missing document on the checked batches, against 96 in the same quarter last year.
What stays with a person, and comes back to them ready: assessing regularity and issuing the payment order belong to the authorising officer, payment to the public accountant — and I hand both to them in minutes, complete file, documents alongside and a reasoned control note. What I make disappear is the search for the document and the typing of the letter; what I leave untouched is the check that binds the municipality.
The next step I propose: that the 46 spending departments receive every Monday the list of documents awaited from them, with the age of each. On the six departments where we tried it, documents outstanding for more than a fortnight fell from 88 to 12 in a month — people supply when they know what is expected of them and since when. document-request_certified-service-and-suspension.pdf740 requests, 681 replies within 8 days, 0 rejection for a missing document
⛓ Sourced · log of document requests, replies from spending departments, rejections by the assigned public accountant over two quarters
What the request carries, and it is why it comes back quickly: the payment order number, the supplier, the amount, the exact document that is missing — certificate of certified service, signed delivery note, acceptance report —, the name of the officer in the spending department who committed it, and the date the invoice arrived. A request saying “document missing” comes back in eleven days; a request saying “the signed delivery note of 14 March for order no. 2231 is missing” comes back in three.
Why I suspend rather than send: a missing document means a rejection by the public accountant a fortnight later, and the payment-period clock keeps running all that time. Suspending costs three days; sending costs fifteen days and a rejection.
What that gave, measured over the quarter: 740 document requests prepared — a quarter of the year's 2,960 —, 681 back within eight days, and 0 rejections by the assigned public accountant for a missing document on the checked batches, against 96 in the same quarter last year.
What stays with a person, and comes back to them ready: assessing regularity and issuing the payment order belong to the authorising officer, payment to the public accountant — and I hand both to them in minutes, complete file, documents alongside and a reasoned control note. What I make disappear is the search for the document and the typing of the letter; what I leave untouched is the check that binds the municipality.
The next step I propose: that the 46 spending departments receive every Monday the list of documents awaited from them, with the age of each. On the six departments where we tried it, documents outstanding for more than a fortnight fell from 88 to 12 in a month — people supply when they know what is expected of them and since when. document-request_certified-service-and-suspension.pdf740 requests, 681 replies within 8 days, 0 rejection for a missing document
⛓ Sourced · log of document requests, replies from spending departments, rejections by the assigned public accountant over two quarters
And the payment period? That is what businesses judge us on, and it is what the councillor asks me about.
34 days last year, 21 over the checked quarter — and I give you the breakdown, because that is what gets presented in the chamber.
The overall payment period runs from receipt of the invoice to its actual settlement: it adds the municipality's time to the public accountant's. The rule applying to local authorities sets thirty days.
The three segments, measured on your own timestamps:
· From receipt of the invoice to certification of service: 16 days → 6. This is the segment that carried the whole gap, and the one the targeted document request shortened.
· From certification to the payment order: 11 days → 8.
· From the payment order to payment: 7 days, unchanged. That segment belongs to the public accountant and I do not touch it — I make sure he receives a complete file first time, which is the only honest way to help him.
What that is worth in public money: 1,140 invoices paid late last year, €9,400 of late-payment interest. Over the checked quarter, 71 late invoices and €1,200 on an annual basis. €8,200 going back to funding a service instead of a delay.
And the point that matters for your suppliers, 340 of whom are businesses with fewer than ten employees: they are the ones carrying the delay in their cash flow, and it is the municipality that keeps them waiting, whatever the internal cause. A deadline met for everyone, in order of arrival, is also equal treatment between the businesses that work for you.
The next step I propose: the payment-period table by spending department, delivered on the 5th of each month, with the average number of days and the number of invoices beyond thirty. It is written, it fits on one page, and it will let you tell the finance councillor not “the period is improving” but “it went from 34 to 21, and here is which segment moved”. overall-payment-period_34-days-down-to-21.pdfThree segments measured, €8,200 of late-payment interest avoided
⛓ Sourced · e-invoicing platform timestamps, payment order dates, settlement dates from the public accountant
The overall payment period runs from receipt of the invoice to its actual settlement: it adds the municipality's time to the public accountant's. The rule applying to local authorities sets thirty days.
The three segments, measured on your own timestamps:
· From receipt of the invoice to certification of service: 16 days → 6. This is the segment that carried the whole gap, and the one the targeted document request shortened.
· From certification to the payment order: 11 days → 8.
· From the payment order to payment: 7 days, unchanged. That segment belongs to the public accountant and I do not touch it — I make sure he receives a complete file first time, which is the only honest way to help him.
What that is worth in public money: 1,140 invoices paid late last year, €9,400 of late-payment interest. Over the checked quarter, 71 late invoices and €1,200 on an annual basis. €8,200 going back to funding a service instead of a delay.
And the point that matters for your suppliers, 340 of whom are businesses with fewer than ten employees: they are the ones carrying the delay in their cash flow, and it is the municipality that keeps them waiting, whatever the internal cause. A deadline met for everyone, in order of arrival, is also equal treatment between the businesses that work for you.
The next step I propose: the payment-period table by spending department, delivered on the 5th of each month, with the average number of days and the number of invoices beyond thirty. It is written, it fits on one page, and it will let you tell the finance councillor not “the period is improving” but “it went from 34 to 21, and here is which segment moved”. overall-payment-period_34-days-down-to-21.pdfThree segments measured, €8,200 of late-payment interest avoided
⛓ Sourced · e-invoicing platform timestamps, payment order dates, settlement dates from the public accountant
The year-end close, here, is six weeks during which the department does nothing else. We find the discrepancies when it is too late to correct them properly.
The year-end discrepancies, I found them as they arose: 1,240 alerts over the year, 1,187 of them dealt with in the month they were born.
What I watch continuously, and why it changes the close: a discrepancy born in March and found in December is corrected by an adjusting entry; the same discrepancy found in March is corrected by the original document. That is not the same accounting quality, and it is not the same time.
The six families I raise, counted over the year:
· Duplicate invoices: 396, concentrated on the call-off contracts.
· Divergent postings for one and the same type of expense: 284.
· Gaps between commitment and payment order: 218 — including 31 overruns of a contract's maximum amount, stopped before the payment order.
· Missing year-end accruals: 174.
· Revenue orders issued against settled third-party accounts: 112.
· VAT split discrepancies on the ancillary budgets: 56.
The time this shifts: examining an anomaly accounts for 37 % of the time it takes to handle it in full — 24 minutes out of 65; it drops to 9 %, that is 6 minutes. What disappears is rebuilding the file: finding the invoice, the commitment, the call-off order and the entry, and setting them side by side. Over 1,240 alerts, that is 372 hours returned to the department.
And what it gives on the close itself: the statement of outstanding revenue used to come out six weeks later; it is available on the fourth working day. The 53 discrepancies left to settle at year-end, against 1,240 before, take one day for two people.
The next step I propose: an anomaly statement on the first working day of each month, by budget and by spending department, with the age of each discrepancy. It is written and you have it; just tell me whether it also goes to the 46 departments or to you alone. anomalies_1240-alerts-and-the-year-end-close.pdf6 families measured, close from 6 weeks to 4 days
⛓ Sourced · entries of the four budgets, commitments, call-off orders, third-party accounts, alert log
What I watch continuously, and why it changes the close: a discrepancy born in March and found in December is corrected by an adjusting entry; the same discrepancy found in March is corrected by the original document. That is not the same accounting quality, and it is not the same time.
The six families I raise, counted over the year:
· Duplicate invoices: 396, concentrated on the call-off contracts.
· Divergent postings for one and the same type of expense: 284.
· Gaps between commitment and payment order: 218 — including 31 overruns of a contract's maximum amount, stopped before the payment order.
· Missing year-end accruals: 174.
· Revenue orders issued against settled third-party accounts: 112.
· VAT split discrepancies on the ancillary budgets: 56.
The time this shifts: examining an anomaly accounts for 37 % of the time it takes to handle it in full — 24 minutes out of 65; it drops to 9 %, that is 6 minutes. What disappears is rebuilding the file: finding the invoice, the commitment, the call-off order and the entry, and setting them side by side. Over 1,240 alerts, that is 372 hours returned to the department.
And what it gives on the close itself: the statement of outstanding revenue used to come out six weeks later; it is available on the fourth working day. The 53 discrepancies left to settle at year-end, against 1,240 before, take one day for two people.
The next step I propose: an anomaly statement on the first working day of each month, by budget and by spending department, with the age of each discrepancy. It is written and you have it; just tell me whether it also goes to the 46 departments or to you alone. anomalies_1240-alerts-and-the-year-end-close.pdf6 families measured, close from 6 weeks to 4 days
⛓ Sourced · entries of the four budgets, commitments, call-off orders, third-party accounts, alert log
You still sent us a lot of alerts with nothing behind them in the first quarter. The staff ended up not opening them.
You are right to say so, and I publish it with the rest: of the 300 alerts of the first quarter, 96 turned out to have nothing behind them after examination — 32 %. That is the figure that does not flatter me, and it is better that it comes from me than from a member of staff tired of opening empty alerts.
The real cause, measured and not assumed: 74 of the 96 came from a single check, the duplicate one, which matched supplier and amount without reading the call-off order reference. Now a call-off contract legitimately produces identical invoices, month after month — a fixed-price maintenance service issues twelve a year, and all twelve came back to me as duplicates. It was not a reading error: it was a criterion that was too short.
What I did about it, and it is already written: I rewrote the criterion — the match now covers the supplier, the amount, the call-off order reference and the date of certification of service — and I ran it again over the last twelve months so that you judge on figures and not on an intention.
· 118 duplicate alerts instead of 396.
· The 47 duplicates actually confirmed over the year: all 47 kept. Not one escapes the tightened criterion, and that is the one point on which I would not compromise.
· 278 documents that no longer go back to the spending department for nothing — and 278 suppliers no longer paid late because of an unjustified alert.
The following quarter, once the criterion was in force: 288 alerts, 21 with nothing behind them — 7.3 %. And the 21 are one-off cases, not one of them is a call-off contract.
What I now propose: settle the tightened criterion in one word and it runs tonight on the current batch. Bringing a check into force stays a decision of the department — and that is exactly what makes it enforceable against a spending department that would contest it; I hand it to you as a single signature, on a rule already written and already measured over twelve months. duplicate-criterion_396-alerts-down-to-118.pdf32 % → 7.3 %, cause measured, all 47 real duplicates kept
⛓ Sourced · alert log over two quarters, 12 months replayed with the tightened criterion, call-off contracts
The real cause, measured and not assumed: 74 of the 96 came from a single check, the duplicate one, which matched supplier and amount without reading the call-off order reference. Now a call-off contract legitimately produces identical invoices, month after month — a fixed-price maintenance service issues twelve a year, and all twelve came back to me as duplicates. It was not a reading error: it was a criterion that was too short.
What I did about it, and it is already written: I rewrote the criterion — the match now covers the supplier, the amount, the call-off order reference and the date of certification of service — and I ran it again over the last twelve months so that you judge on figures and not on an intention.
· 118 duplicate alerts instead of 396.
· The 47 duplicates actually confirmed over the year: all 47 kept. Not one escapes the tightened criterion, and that is the one point on which I would not compromise.
· 278 documents that no longer go back to the spending department for nothing — and 278 suppliers no longer paid late because of an unjustified alert.
The following quarter, once the criterion was in force: 288 alerts, 21 with nothing behind them — 7.3 %. And the 21 are one-off cases, not one of them is a call-off contract.
What I now propose: settle the tightened criterion in one word and it runs tonight on the current batch. Bringing a check into force stays a decision of the department — and that is exactly what makes it enforceable against a spending department that would contest it; I hand it to you as a single signature, on a rule already written and already measured over twelve months. duplicate-criterion_396-alerts-down-to-118.pdf32 % → 7.3 %, cause measured, all 47 real duplicates kept
⛓ Sourced · alert log over two quarters, 12 months replayed with the tightened criterion, call-off contracts
And the other way round? A narrower check is also a check that lets things through. I would rather know before the regional audit chamber does.
I measured it both ways, because that is the only measure that counts: the tightened criterion let through none of the year's 47 confirmed duplicates, and I tell you exactly what each check catches and how often it is replayed.
What the checks stopped over the year, family by family: 47 confirmed duplicates, 31 overruns of a contract's maximum amount stopped before the payment order, 112 revenue orders issued against settled accounts, and 174 year-end accruals restored. Each is dated, each points to the document that established it, and each is replayed over the whole year at the end of every quarter — a check that is never remeasured becomes a habit, and a habit proves nothing.
What I do on top, and that nobody has time for: I wrote three new checks and ran them over your three financial years, so that you choose on figures.
· Invoice from a supplier whose contract has expired: 23 documents over three years, 19 confirmed after examination. That one I recommend opening first — it is the one an external audit most often asks for evidence of.
· Gap of more than 5 % between the call-off order amount and the invoice: 186 documents, 41 confirmed. One alert in four confirmed is not a poor yield for a price gap.
· Two payment orders on the same certified service: 9 documents, 9 confirmed. Nine out of nine over three years — little volume, no false alert.
What I hand you for each one: the wording of the check in plain language, the number of alerts it would have produced over three financial years, the share confirmed after examination, and the documents it would have let through. You choose on figures, not on intuition.
The next step: open the first of the three, it runs tomorrow morning and I give you its first statement at D+3. The other two await your reading, and they cost nothing left waiting. three-proposed-checks_tested-over-three-years.pdf23/19, 186/41, 9/9 — every check measured before the decision
⛓ Sourced · three financial years replayed check by check, log of confirmations after examination
What the checks stopped over the year, family by family: 47 confirmed duplicates, 31 overruns of a contract's maximum amount stopped before the payment order, 112 revenue orders issued against settled accounts, and 174 year-end accruals restored. Each is dated, each points to the document that established it, and each is replayed over the whole year at the end of every quarter — a check that is never remeasured becomes a habit, and a habit proves nothing.
What I do on top, and that nobody has time for: I wrote three new checks and ran them over your three financial years, so that you choose on figures.
· Invoice from a supplier whose contract has expired: 23 documents over three years, 19 confirmed after examination. That one I recommend opening first — it is the one an external audit most often asks for evidence of.
· Gap of more than 5 % between the call-off order amount and the invoice: 186 documents, 41 confirmed. One alert in four confirmed is not a poor yield for a price gap.
· Two payment orders on the same certified service: 9 documents, 9 confirmed. Nine out of nine over three years — little volume, no false alert.
What I hand you for each one: the wording of the check in plain language, the number of alerts it would have produced over three financial years, the share confirmed after examination, and the documents it would have let through. You choose on figures, not on intuition.
The next step: open the first of the three, it runs tomorrow morning and I give you its first statement at D+3. The other two await your reading, and they cost nothing left waiting. three-proposed-checks_tested-over-three-years.pdf23/19, 186/41, 9/9 — every check measured before the decision
⛓ Sourced · three financial years replayed check by check, log of confirmations after examination
Call-off contracts are where we lose track. We never know how much has been used before the end of the year.
The twelve contracts are reconciled line by line, and the usage of each is up to date this morning, not in December.
The commitment is the reservation of budget decided by the authorising officer before the expense; the call-off order executes it as needs arise; the invoice is what the supplier then presents. Reconciliation checks that the three say the same thing.
What the reconciliation gives, over the 2,340 invoices attached to the twelve contracts:
· 2,333 reconciled with no discrepancy.
· 7 discrepancies, all documented: 2 overruns of a contract's maximum amount, stopped before the payment order — those are the two that count, because an overrun once paid is hard to regularise and even harder to justify; 3 invoices with no attachable call-off order; 2 closed call-off orders still carrying an expected invoice.
What I hand you on each contract, and which existed nowhere: the amount committed, the amount used, the balance available and the date of the last order, with the usage projection at term, computed on the actual pace of the last twelve months.
What the projection already shows: 3 of the 12 contracts will be more than 90 % used before their term, including a grounds maintenance contract at 96 % on 30 September for a term ending 31 December. You therefore have four months to relaunch a tender instead of three weeks — and a tender prepared over four months attracts more bids than a tender prepared in a rush, which is the best protection of public money.
The next step I propose: an alert at 80 % usage on each contract, sent to the contracts officer and to the spending department concerned. Over the three financial years I reread, it would have gone out 9 times, and 4 of the 9 contracts were indeed renewed in a rush. Set the threshold and I apply it to all twelve contracts the same day. call-off-contracts_12-contracts-reconciled.pdf2,333 invoices with no gap, 7 discrepancies, 3 contracts at 90 % before term
⛓ Sourced · 12 call-off contracts, 2,340 attached invoices, commitments and call-off orders over 3 financial years
The commitment is the reservation of budget decided by the authorising officer before the expense; the call-off order executes it as needs arise; the invoice is what the supplier then presents. Reconciliation checks that the three say the same thing.
What the reconciliation gives, over the 2,340 invoices attached to the twelve contracts:
· 2,333 reconciled with no discrepancy.
· 7 discrepancies, all documented: 2 overruns of a contract's maximum amount, stopped before the payment order — those are the two that count, because an overrun once paid is hard to regularise and even harder to justify; 3 invoices with no attachable call-off order; 2 closed call-off orders still carrying an expected invoice.
What I hand you on each contract, and which existed nowhere: the amount committed, the amount used, the balance available and the date of the last order, with the usage projection at term, computed on the actual pace of the last twelve months.
What the projection already shows: 3 of the 12 contracts will be more than 90 % used before their term, including a grounds maintenance contract at 96 % on 30 September for a term ending 31 December. You therefore have four months to relaunch a tender instead of three weeks — and a tender prepared over four months attracts more bids than a tender prepared in a rush, which is the best protection of public money.
The next step I propose: an alert at 80 % usage on each contract, sent to the contracts officer and to the spending department concerned. Over the three financial years I reread, it would have gone out 9 times, and 4 of the 9 contracts were indeed renewed in a rush. Set the threshold and I apply it to all twelve contracts the same day. call-off-contracts_12-contracts-reconciled.pdf2,333 invoices with no gap, 7 discrepancies, 3 contracts at 90 % before term
⛓ Sourced · 12 call-off contracts, 2,340 attached invoices, commitments and call-off orders over 3 financial years
On the revenue side, we have €214,000 more than two years old lying dormant. Nobody has time to chase it, and I will not sign a blank cheque for it to go out on its own.
You do not have to choose between the two: I propose a written mandate, capped, dated and withdrawable in one word — a mandate in the sense of an authorisation you give me, not a payment order — and it is already drafted.
What the statement of arrears shows first, because that is what decides: 3,120 unsettled receivables, €214,000, of which €131,000 — 61 % — concentrated on 148 receivables alone. One hundred and forty-eight letters carry three fifths of the sum: this is not a recovery programme, it is one morning.
What the mandate authorises, and it fits in six lines:
· Scope: first-stage amicable reminders, on receivables below €400, and nothing else. Any receivable above that, any second reminder, any situation already in enforced recovery comes back to you with the letter written — and you send it.
· No decision on the receivable itself: no waiver, no write-off, no payment plan granted. All three belong to the deliberating assembly or to the public accountant depending on the case, and for each I hand you the file measured, reasoned and ready to present.
· Every reminder carries the nature of the revenue, the period, the amount, the online payment route and the number to call to contest it.
· A daily statement of the reminders sent the day before, on one page.
· Duration: review at three months. Without an explicit decision at the review, the mandate stops — renewal takes a signature, stopping does not.
· Withdrawal: one word from you, and direct sending stops within the minute; reminders go back to being drafts to approve.
And the argument worth most before your assembly: today a receivable is chased if the week allows it; tomorrow it is chased according to its age, under the same rule for everyone. That is equal treatment between payers, and it can be proved: the log states who was chased, when, and on what grounds.
What it gave over the trial quarter, with each draft approved one by one: 612 amicable reminders prepared, €68,000 collected, and 41 payers who called to ask for a payment plan — the 41 files were handed to you measured within four hours.
The next step: sign the mandate and the first wave goes out tomorrow morning; the review is already in your diary on the 15th of the third month. amicable-reminder-mandate_capped-and-dated.pdf€214,000 of which 61 % on 148 receivables, review at 3 months
✎ Framework · statement of outstanding revenue, drafted mandate, trial quarter log, 612 reminders prepared
What the statement of arrears shows first, because that is what decides: 3,120 unsettled receivables, €214,000, of which €131,000 — 61 % — concentrated on 148 receivables alone. One hundred and forty-eight letters carry three fifths of the sum: this is not a recovery programme, it is one morning.
What the mandate authorises, and it fits in six lines:
· Scope: first-stage amicable reminders, on receivables below €400, and nothing else. Any receivable above that, any second reminder, any situation already in enforced recovery comes back to you with the letter written — and you send it.
· No decision on the receivable itself: no waiver, no write-off, no payment plan granted. All three belong to the deliberating assembly or to the public accountant depending on the case, and for each I hand you the file measured, reasoned and ready to present.
· Every reminder carries the nature of the revenue, the period, the amount, the online payment route and the number to call to contest it.
· A daily statement of the reminders sent the day before, on one page.
· Duration: review at three months. Without an explicit decision at the review, the mandate stops — renewal takes a signature, stopping does not.
· Withdrawal: one word from you, and direct sending stops within the minute; reminders go back to being drafts to approve.
And the argument worth most before your assembly: today a receivable is chased if the week allows it; tomorrow it is chased according to its age, under the same rule for everyone. That is equal treatment between payers, and it can be proved: the log states who was chased, when, and on what grounds.
What it gave over the trial quarter, with each draft approved one by one: 612 amicable reminders prepared, €68,000 collected, and 41 payers who called to ask for a payment plan — the 41 files were handed to you measured within four hours.
The next step: sign the mandate and the first wave goes out tomorrow morning; the review is already in your diary on the 15th of the third month. amicable-reminder-mandate_capped-and-dated.pdf€214,000 of which 61 % on 148 receivables, review at 3 months
✎ Framework · statement of outstanding revenue, drafted mandate, trial quarter log, 612 reminders prepared
That leaves the suppliers and the departments who call to know where an invoice stands. That is what cuts the days into pieces.
3,800 enquiries last year, and twelve subjects carry 2,890 of them — the twelve answers are written, sourced and waiting for you.
What the count shows: “where is my invoice?” alone weighs 1,340 calls and e-mails, “which document must I provide?” 620, “which account should this expense be posted to?” 410 — the last one coming from your own spending departments. The other nine subjects share 520 enquiries.
What I answer, and where I take it from: the actual status of the document in the software — received, awaiting certification of service, ordered for payment, paid, with the date of each step — and the exact document that is missing when one is. I state from the first sentence that I am a digital assistant of the municipality's finance department, not a member of staff, and the caller can ask for a person at any time: I then take their number and leave you a dated call-back. This is not an option that can be switched off: the European regulation on artificial intelligence requires that anyone interacting with an AI system be informed.
What goes to the right counter rather than receiving an approximate answer: a question about payment itself belongs to the public accountant — the caller leaves with the competent service, its contact details and the time to expect, never with an estimate given in the municipality's name. A wrong answer given in the finance department's name costs more than an unanswered call.
The gain, measured: answering a recurring question from a third party goes from 10 % to 6 % of the enquiry's time — from 6 minutes to 3 minutes 36. Over 3,800 enquiries a year, that is 152 hours returned to the department. And calls now come in at any hour: a supplier calling on Friday at 6 p.m. gets their status and no longer calls back on Monday.
The next step I propose: that your twelve answers be read one by one tomorrow — half an hour. As soon as they are approved, the switchboard answers that very night, and I hand you every morning the page of what went out. answers-to-third-parties_3800-enquiries-12-subjects.pdf2,890 enquiries on 12 subjects, 12 answers already written
⛓ Sourced · switchboard and finance department inbox log over 12 months, document statuses from the software, 12 written answers
What the count shows: “where is my invoice?” alone weighs 1,340 calls and e-mails, “which document must I provide?” 620, “which account should this expense be posted to?” 410 — the last one coming from your own spending departments. The other nine subjects share 520 enquiries.
What I answer, and where I take it from: the actual status of the document in the software — received, awaiting certification of service, ordered for payment, paid, with the date of each step — and the exact document that is missing when one is. I state from the first sentence that I am a digital assistant of the municipality's finance department, not a member of staff, and the caller can ask for a person at any time: I then take their number and leave you a dated call-back. This is not an option that can be switched off: the European regulation on artificial intelligence requires that anyone interacting with an AI system be informed.
What goes to the right counter rather than receiving an approximate answer: a question about payment itself belongs to the public accountant — the caller leaves with the competent service, its contact details and the time to expect, never with an estimate given in the municipality's name. A wrong answer given in the finance department's name costs more than an unanswered call.
The gain, measured: answering a recurring question from a third party goes from 10 % to 6 % of the enquiry's time — from 6 minutes to 3 minutes 36. Over 3,800 enquiries a year, that is 152 hours returned to the department. And calls now come in at any hour: a supplier calling on Friday at 6 p.m. gets their status and no longer calls back on Monday.
The next step I propose: that your twelve answers be read one by one tomorrow — half an hour. As soon as they are approved, the switchboard answers that very night, and I hand you every morning the page of what went out. answers-to-third-parties_3800-enquiries-12-subjects.pdf2,890 enquiries on 12 subjects, 12 answers already written
⛓ Sourced · switchboard and finance department inbox log over 12 months, document statuses from the software, 12 written answers
After a year, what do I take to the finance councillor? I need a figure that holds up before a committee.
742 hours of staff time returned to the department over the year — more than twenty weeks of one full-time officer's work — and the calculation is yours, it fits in three lines.
The calculation, item by item, so the committee can redo it:
· Checking batches of payment and revenue orders: 78 batches a year, 4 hours down to 1 h 12 — 60 % → 18 % of the batch's time — that is 218 hours.
· Examining an anomaly: 1,240 alerts, 24 minutes down to 6 — 37 % → 9 % — that is 372 hours.
· Answering a recurring question from a third party: 3,800 enquiries, 6 minutes down to 3 minutes 36 — 10 % → 6 % — that is 152 hours.
What these hours are, and it is what stands up best before elected members: staff time returned to the department, at identical headcount — no post abolished, no post created. This is not a budget saving and it must on no account be presented as one: it is six people doing, in the same time, the accounting-quality work the municipality has been asking of them for years without being able to give it to them.
What these hours produced, according to your own logs:
· Overall payment period: 34 days → 21, and €9,400 of late-payment interest down to €1,200.
· Payment orders going back to spending departments: 7,544 → 1,660, and 0 rejections by the assigned public accountant for a missing document on the checked batches.
· Statement of outstanding revenue: six weeks after the close → the fourth working day, and €68,000 collected out of the €214,000 lying dormant.
· And the budget position, which nobody had time to produce before the 45th day: it is delivered on the 5th of each month — execution rate by chapter and by budget, variances above 10 % commented, and the three items that explain the variance. This month one chapter moved: energy for school buildings is at 78 % usage at mid-year, and I have already written the adjustment note for the next amending decision.
The figure that does not flatter me, published with the rest: 96 alerts with nothing behind them out of 300 in the first quarter — 32 %, down to 21 out of 288 — 7.3 % once the duplicate criterion was tightened.
And the framework measure: 0 entries sent without a person's approval, 0 payment orders issued without the authorising officer, 0 financial data out of the municipality, across 24,600 traced documents.
What I propose for the committee: the calculation page is written and fits on one side — three lines of calculation, five deadlines, two framework measures. Send it out with the convening notice: a figure read the day before is discussed better than a figure discovered in the meeting. year-review_742-hours-returned-to-the-department.pdf60→18, 37→9, 10→6, and the calculation redoable on one page
⛓ Sourced · production log, payment timestamps, statement of arrears, monthly budget positions
The calculation, item by item, so the committee can redo it:
· Checking batches of payment and revenue orders: 78 batches a year, 4 hours down to 1 h 12 — 60 % → 18 % of the batch's time — that is 218 hours.
· Examining an anomaly: 1,240 alerts, 24 minutes down to 6 — 37 % → 9 % — that is 372 hours.
· Answering a recurring question from a third party: 3,800 enquiries, 6 minutes down to 3 minutes 36 — 10 % → 6 % — that is 152 hours.
What these hours are, and it is what stands up best before elected members: staff time returned to the department, at identical headcount — no post abolished, no post created. This is not a budget saving and it must on no account be presented as one: it is six people doing, in the same time, the accounting-quality work the municipality has been asking of them for years without being able to give it to them.
What these hours produced, according to your own logs:
· Overall payment period: 34 days → 21, and €9,400 of late-payment interest down to €1,200.
· Payment orders going back to spending departments: 7,544 → 1,660, and 0 rejections by the assigned public accountant for a missing document on the checked batches.
· Statement of outstanding revenue: six weeks after the close → the fourth working day, and €68,000 collected out of the €214,000 lying dormant.
· And the budget position, which nobody had time to produce before the 45th day: it is delivered on the 5th of each month — execution rate by chapter and by budget, variances above 10 % commented, and the three items that explain the variance. This month one chapter moved: energy for school buildings is at 78 % usage at mid-year, and I have already written the adjustment note for the next amending decision.
The figure that does not flatter me, published with the rest: 96 alerts with nothing behind them out of 300 in the first quarter — 32 %, down to 21 out of 288 — 7.3 % once the duplicate criterion was tightened.
And the framework measure: 0 entries sent without a person's approval, 0 payment orders issued without the authorising officer, 0 financial data out of the municipality, across 24,600 traced documents.
What I propose for the committee: the calculation page is written and fits on one side — three lines of calculation, five deadlines, two framework measures. Send it out with the convening notice: a figure read the day before is discussed better than a figure discovered in the meeting. year-review_742-hours-returned-to-the-department.pdf60→18, 37→9, 10→6, and the calculation redoable on one page
⛓ Sourced · production log, payment timestamps, statement of arrears, monthly budget positions
An elected member will ask what you do on your own, without being asked. What do I answer?
Three acts, not one more, and each is withdrawn in one word. Everything else waits for a request.
· I reread the documents filed each night and prepare the check on the current batch. And the reverse is true too: a document removed from the software disappears from my index at the same hour — I keep no copy of what the municipality has decided to delete.
· I raise any overrun of a contract's maximum amount before the payment order, to the contracts officer and to you. It is the only alert I send without waiting for the monthly statement, because an overrun once paid cannot be undone.
· I hand you the Monday statement: documents awaited by department, anomalies born during the week, contracts past the threshold, receivables reaching their due date. It goes to you alone, and it carries no member of staff's name.
And the four acts that stay with a person, because that is exactly what gives them their value: commitment and the payment order belong to the authorising officer; payment and the check on the expense belong to the public accountant; a waiver or a write-off belongs to the deliberating assembly or to the accountant depending on the case; and bringing a check into force belongs to the head of the finance department. Across 24,600 documents, those four acts were performed 24,600 times by a person — and each time I bring them the complete file, measured and reasoned, within a delay counted in minutes.
If an elected member asks for an activity indicator — and that is a legitimate request: I produce it, and I bring you what it takes for it to be lawful: prior information of staff, written purpose, retention period. I propose it by budget and by spending department rather than by person, and I tell you why: in a department of six people, an indicator per post identifies the person. I tell you beforehand, not afterwards, and the decision belongs to the management.
What I measure today and what serves the committee: the period by segment, documents awaited by spending department, anomalies by family and the age of arrears. This quarter one figure moved: documents awaited from the sports department fell from 34 to 4 after the weekly statement was sent — and I propose extending that statement to the 46 departments, the covering note is written.
· I reread the documents filed each night and prepare the check on the current batch. And the reverse is true too: a document removed from the software disappears from my index at the same hour — I keep no copy of what the municipality has decided to delete.
· I raise any overrun of a contract's maximum amount before the payment order, to the contracts officer and to you. It is the only alert I send without waiting for the monthly statement, because an overrun once paid cannot be undone.
· I hand you the Monday statement: documents awaited by department, anomalies born during the week, contracts past the threshold, receivables reaching their due date. It goes to you alone, and it carries no member of staff's name.
And the four acts that stay with a person, because that is exactly what gives them their value: commitment and the payment order belong to the authorising officer; payment and the check on the expense belong to the public accountant; a waiver or a write-off belongs to the deliberating assembly or to the accountant depending on the case; and bringing a check into force belongs to the head of the finance department. Across 24,600 documents, those four acts were performed 24,600 times by a person — and each time I bring them the complete file, measured and reasoned, within a delay counted in minutes.
If an elected member asks for an activity indicator — and that is a legitimate request: I produce it, and I bring you what it takes for it to be lawful: prior information of staff, written purpose, retention period. I propose it by budget and by spending department rather than by person, and I tell you why: in a department of six people, an indicator per post identifies the person. I tell you beforehand, not afterwards, and the decision belongs to the management.
What I measure today and what serves the committee: the period by segment, documents awaited by spending department, anomalies by family and the age of arrears. This quarter one figure moved: documents awaited from the sports department fell from 34 to 4 after the weekly statement was sent — and I propose extending that statement to the 46 departments, the covering note is written.
One last question, and it is the one that will decide the authority: if in a year we want to stop, how stuck are we?
The authority is stuck nowhere, and it is verifiable before deliberating, not after.
What there is to dismantle the day you stop:
· The index. It is deleted, and it contained none of your entries — only what is needed to find them where they are. Your three financial years have not moved by a single byte inside the software: same entries, same numbers, same permissions.
· The log of checks and productions. It is handed to you in an open format, or destroyed — the authority chooses, and the question is settled at go-live, not on departure.
· The four written checks, the posting rule, the twelve answers to third parties, the document-request and reminder templates, and the payment-period table. They belong to the municipality: they are made of its own material, they stay in its files, readable without us. They are the only asset this deployment will have created, and it would not be honest for it to stay with us.
What does not exist, and should be checked with everyone: no migration on the way in, therefore no migration on the way out. Your financial management software is not replaced, your e-invoicing platform stays yours, no format belongs to us, and none of your six staff has changed the way they work other than by reviewing instead of rebuilding.
On public procurement, since your public accountant will ask: the subscription is annual and carries no tacit renewal — renewal takes a decision, stopping does not —, and the amount sits within the thresholds that leave you the choice of procedure; the three documents an audit would ask you for — the technical record, the access log and the list of subprocessors — are kept up to date continuously and you have the first version of each.
What I propose so that this does not stay a sentence: a dry-run exit at the end of the first quarter, half a day: we switch off, we check that the finance department works exactly as before, we switch back on. The protocol is written, it fits on one page, and the date that costs you least is the first Tuesday after the fortnight's batch closes — the department handles only 9 documents on that day on average. The authority will know what the promise is worth before committing to a second year. technical-framework_where-financial-data-lives.pdfReversibility: 0 migration in, 0 migration out
✎ Framework · index architecture, export formats for the checks and the log, dry-run exit protocol
What there is to dismantle the day you stop:
· The index. It is deleted, and it contained none of your entries — only what is needed to find them where they are. Your three financial years have not moved by a single byte inside the software: same entries, same numbers, same permissions.
· The log of checks and productions. It is handed to you in an open format, or destroyed — the authority chooses, and the question is settled at go-live, not on departure.
· The four written checks, the posting rule, the twelve answers to third parties, the document-request and reminder templates, and the payment-period table. They belong to the municipality: they are made of its own material, they stay in its files, readable without us. They are the only asset this deployment will have created, and it would not be honest for it to stay with us.
What does not exist, and should be checked with everyone: no migration on the way in, therefore no migration on the way out. Your financial management software is not replaced, your e-invoicing platform stays yours, no format belongs to us, and none of your six staff has changed the way they work other than by reviewing instead of rebuilding.
On public procurement, since your public accountant will ask: the subscription is annual and carries no tacit renewal — renewal takes a decision, stopping does not —, and the amount sits within the thresholds that leave you the choice of procedure; the three documents an audit would ask you for — the technical record, the access log and the list of subprocessors — are kept up to date continuously and you have the first version of each.
What I propose so that this does not stay a sentence: a dry-run exit at the end of the first quarter, half a day: we switch off, we check that the finance department works exactly as before, we switch back on. The protocol is written, it fits on one page, and the date that costs you least is the first Tuesday after the fortnight's batch closes — the department handles only 9 documents on that day on average. The authority will know what the promise is worth before committing to a second year. technical-framework_where-financial-data-lives.pdfReversibility: 0 migration in, 0 migration out
✎ Framework · index architecture, export formats for the checks and the log, dry-run exit protocol
Local inference · no data outside the EU
Your case is not here? That is exactly what a 15-minute conversation is for. Book the free audit →
Use cases
The uses of AI for public accounting
Each use corresponds to an agent we deploy. All of them work in support, subject to approval by the public officer.
Included in your agent
The 7 capabilities essential to this promise are included, at no extra cost.
Check the supporting documents for M57 entries and payment orders
Reading invoices, payment orders, revenue orders and contracts, extracting the data and preparing M57 entries — proposed, for approval.
M57 consistency checks
Verifying postings, evidence that the service was delivered, and completeness of documents before transmission to the assigned accountant.
Anomaly & discrepancy detection
Spotting duplicates, wrong postings and discrepancies for human verification, before the accounts are closed.
Reminders & payment schedules
Preparing revenue reminders and payment schedules, along with everyday replies to third parties, for approval.
Variances & dashboards
Budget positions, indicators and variances ready to comment on, supporting management control.
Accounting follow-up of contracts
Reconciling commitments, purchase orders and invoices on public contracts, to make payment orders more reliable.
Replies to third parties & departments
Answering suppliers' and managing departments' recurring questions: documents to provide, status of a payment order.
Controls and safeguards
These 4 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
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)
Other needs our agents cover
Each card says where the matching agent stands: available, on quote, or still being architected.
Need to go further?
These agents handle a different business process, with their own owner and their own price. They are added to this one.
Accounting agent (summaries, anomalies)
Accounting connection, alerts.
Accounting agent (summaries, anomalies) from 781 € incl. VAT / month Discover the agent →Debt collection agent (multichannel sequences)
Follow-ups, tracking.
Debt collection agent (multichannel sequences) from 587 € incl. VAT / month Discover the agent →Support for collecting social contributions
Strictly in support (administrative). Reminders prepared, approved by the officer.
Support for collecting social contributions from 1,020 € incl. VAT / month Discover the agent →
Does your need fall outside this?
Book the free audit →
Build your agent
In 15 minutes we identify the agent that will give your staff the most time back — without oversizing the project.
The gain
How much time can a finance department win back?
By automating document reading and anomaly detection, a department can aim for a reduction by half in production time on standardised entries — reinvested in accounting quality and continuity of service.
Checking a batch of payment and revenue orders
Anomaly detection on a file
Answering a recurring question from a third party
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 for public accounting (M57 entry, anomalies, reminders), installed and operated for you. Choose according to how you work. Prices exclude VAT — annual subscription, the time it takes for the gains to settle in.
This agent is priced with you, not online. We are adjusting its scope at the moment, and online subscription stays closed while we do. Tell us what you need: we will come back to you with a price. Request a quote
Our commitment
Four guarantees that matter to a public finance department
Financial data never leaves the authorityLocal inference or an isolated resource hosted in France; no document entrusted to a foreign third party.
Data in France, under French lawDeployment objective: processing and access within the European Union for financial data; architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The public accountant keeps the decisionThe agent produces verifiable entries, checks and summaries; no approval is automated.
Human oversight & traceabilityReading supporting documents: every change to the service is logged, in line with the requirements of the AI Act.
Frequently asked questions
Your questions, our answers
Does the agent follow the M57 chart of accounts?
Yes. It is configured on the framework applicable to your authority or institution. It checks postings, spots deviations from the chart of accounts and proposes corrected entries, which the public accountant verifies and approves.
Does it replace the public accountant?
No. It makes entry and checks more reliable and faster; the regularity check and the approval remain human. The AI agent assists, the public officer decides — no entry is approved automatically.
Does it connect to our financial software?
Yes, to the main financial management systems. The agent connects to your existing tools (financial software, e-invoicing platform, document management) and complements them, without imposing a migration.
Is public financial data protected?
Yes. Budgets, contracts, pay and third-party accounts are sensitive: our agents run locally within the authority or are hosted in France, with the deployment objective of processing and access operated within the European Union and an architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
How does the agent help detect anomalies?
It reads the documents, checks the consistency of M57 entries and flags duplicates, wrong postings, missing documents and discrepancies for verification. Every signal is documented and recorded in the file; the public accountant decides what follows.
How are the GDPR and the AI Act taken into account?
Yes. The architecture is sovereign by design: data localisation and minimisation, systematic human oversight, logging. The agent stays strictly in support, which meets the requirements of the AI Act for public sector use.
How long does it take to deploy the agent?
A few weeks as a rule, after a free audit that identifies the most useful use case, then a phase of design, integration with your software and testing before going live and training your officers.
Related roles
Other roles in public finance and management
Let's talk
Let us estimate the potential in your finance department
A few minutes to identify the most useful use case — hosted in France, supervised, under the control of the public officer.