We build Agents. You build the Future.
● Business offer — Meetings & follow-upNote-taking agent: transcription, minutes, actions
A meeting without minutes has little effect: everyone leaves with their own version. Taking notes also means you stop taking part. Your agent transcribes the exchange, draws out structured minutes and a list of assigned actions, ready to correct. Data hosted in France; the AI model is called at its publisher, except with the Sovereign model option. With the 100% Sovereign package, your internal discussions go nowhere. The chair approves before circulation.
Updated on
Deployed in a few weeks
Meeting notes · hosted in France
Fictional simulation — not a client result
Prepare the minutes of this morning's meeting.
Structured minutes: points covered, decisions taken as they were stated, and subjects left open.
Actions recorded with the person named as responsible and the deadline mentioned in the discussion.
Anything not explicitly decided appears under “to be settled”, with no interpretation.
✎ Action · minutes to correct, nothing circulated
Actions recorded with the person named as responsible and the deadline mentioned in the discussion.
Anything not explicitly decided appears under “to be settled”, with no interpretation.
✎ Action · minutes to correct, nothing circulated
There was a disagreement about the schedule — did you note it?
Yes, as an open point. Both positions are recorded as they were expressed, without favouring either.
As no decision was formulated, I have recorded none.
✎ Support · faithful record, no arbitration
As no decision was formulated, I have recorded none.
✎ Support · faithful record, no arbitration
Data hosted in France · sovereign model as an option
Recordings hosted in France
Hosted in FranceAI model called at its publisher, except with the Sovereign model option
GDPR & AI Act: governed deploymentTraceability & human oversight
TurnkeyDesigned, installed and operated for you
The chair of the meeting decidesThe agent carries out what you have authorised
For an organisation established outside the European Union, the law of its own country applies; at its request, we put equivalent safeguards in place (human review before any action, a log of the criteria, information for the people concerned).
✦ In brief
A Blue Lemon Agent note-taking agent transcribes your meetings, draws out structured minutes and a list of actions with the responsible people and deadlines as they were stated. What was not decided stays marked as open, with no interpretation. Data hosted in France; the AI model is called at its publisher, except with the Sovereign model option. With that option, your internal discussions are never entrusted to a foreign service, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity. Live within a few weeks,.
100%
hosted in France in the target architecture
0
transfer outside the EU with the Sovereign model option (target architecture)
3
meeting follow-up uses ready to deploy
0
decision with a legal or significant effect taken by an agent
Reference points describing our offer, not results measured at a client: the scale of the gain, given the number and length of your meetings, is confirmed by a pilot.
Indicative estimate, to be confirmed by a pilot on your real files (general terms of sale, article 4).
Pricing
One package, one agent
A note-taking agent (transcription, minutes, actions), installed and operated for you. Prices exclude VAT — annual subscription, the time it takes for the gains to settle in.
Agility
Setup + controlled subscription
1,790 € excl. VAT setup
then 207 € excl. VAT/month — you invest at installation and pay a reduced subscription. Ideal for keeping the cost under control over time.
- Installation, configuration and training for your teams
- Operation, human oversight, updates and support
- Hosting in France, a dedicated and isolated resource
The simplest
Serenity
All inclusive, no setup fee
307 € excl. VAT /month
all inclusive, immediate start. No upfront investment: a single subscription. Ideal for starting quickly and simply.
- Setup included (installation, configuration, training)
- Operation, human oversight, updates and support
- Hosting in France, managed end to end
100% Sovereign
On site, you own it
3,545 € excl. VAT setup
then 304 € excl. VAT/month · + hardware from 2,491 € (one-off purchase, in addition) — a sovereign computer installed on your premises, maintained remotely. Models run locally, your data returned at the end of the contract.
- Hardware installed on your premises (you own it)
- French / European AI models run locally
- Secure remote maintenance (Pro support included)
Included in the package: hosting and AI consumption up to the volume of your capacity tier, in documents, exchanges, tokens and storage; beyond that, re-invoiced at real cost with no margin, tracked in your client area as usage statements are issued. Maintenance and supervision subscription for an initial term of 12 months for the Agility package, 24 months for the Serenity package and 36 months for the 100% Sovereign package, renewable; support levels (SLA 72 h / 24 h / 4 h) optional. Bespoke development, additional integrations or exceptional volumes are quoted separately. Support Monday to Friday, 9am to 6pm. Prices exclude VAT.
AI model:
none of the AI models offered currently carries a fixed surcharge. When the selected model carries a cost, that cost is shown when you choose it, before you order, and re-invoiced at the cost incurred, with no mark-up; usage is billed at the publisher's price. Publishers' prices are published in US dollars: the amount re-invoiced is the amount in euros actually borne by Blue Lemon Agent on the publisher's invoice, at that invoice's exchange rate, with no commission or mark-up.
Included components and additional components Components included in the base offer: the Blue Lemon Agent software foundation, the AI models listed in the order journey, the standard channels (Microsoft Teams, Slack, WhatsApp Business, email, website chat, calendars, Microsoft 365 / Google Workspace, file storage, market VoIP telephony, professional social-media pages and accounts, Google Business Profile), hosting for the package chosen, in France by default, backups, supervision, updates and support. If adapting the AI agent to your constraints, your needs or your requests requires other paid components — a third-party publisher's software licence, paid API access to one of your applications, hosting of health data, for which French law requires an HDS-certified host (art. L. 1111-8 of the French Public Health Code), SecNumCloud-qualified hosting, a speech synthesis service, particular hardware —, they are offered to you as an option or on quotation and re-invoiced at the cost incurred; nothing is committed without your written agreement. Where the artificial intelligence model you choose entails an additional cost, that cost is shown to you before you order and re-invoiced to you at the cost incurred, with no margin.
What to expect
Go-live
2 to 3 weeks
Agent designed, channels connected, team trained.
Steady state
4 to 7 weeks
After a few weeks of real use, once the agent's behaviour matches what you expect.
Indicative estimate, adjusted to the options you keep. It is not a delivery commitment.
Use cases
What does the agent actually do?
One agent, several moments in meeting follow-up. The agent carries out by itself what you authorise it to do; you decide what it does on its own and what it submits to you.
Included in your agent
The 3 capabilities essential to this promise are included, at no extra cost.
From 207 € excl. VAT / month
Faithful transcription
Records the exchange without rewording it, distinguishing the speakers.
Structured minutes
Draws out points covered, decisions stated and subjects left open.
Action list
Records the actions with the responsible person and the deadline as they were named.
Controls and safeguards
These 3 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
Data protection, rights and human oversight
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.
Integration with existing tools without double entry
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.
Project tracking
To turn the actions into deadline tracking, a dedicated project agent takes over.
Project tracking agent (tasks, deadlines) from 259 € excl. VAT / month Project tracking →Circulation by email
To prepare the sending of the approved minutes, the email assistant completes the chain.
Email assistant (sorting, drafts, replies) from 326 € excl. VAT / month Email assistant →
Does your need fall outside this?
Build your agent
In 30 minutes we identify the most relevant agent — without oversizing the project.
Demonstration
See the agent at work
5 everyday situations, taken from those that come up most often. Pick one: the exchange unfolds as it would in your organisation.
A scripted demonstration. These exchanges show how the agent behaves — its sources, its refusals, what it leaves to your teams. Nothing is sent from this page, no model is queried here, and the matters named are fictional. That is precisely what we promise your data. The behaviours shown here — monitoring, automation rules, routing and reminders — are configured with you during deployment, from your tools, your rules and your thresholds. The architecture points named in these exchanges — location, local execution, isolation, encryption, role-based access, logging — are not a guarantee attached to the demonstration: they are those of the architecture set out in your quotation, and verified before commissioning.
Meeting notes · hosted in France
Fictional simulation — not a client result
Four flags this morning on your minutes.
· Thirty-four decisions were recorded with nobody named to carry them out. Eleven are more than two months old.
· Eighteen actions were assigned to somebody who was not in the meeting. Nine never found out.
· Forty-one sets of minutes were read by no participant before circulation.
· Over twelve months, three meetings began before I announced my presence. I deleted what preceded the announcement. morning-watch_4-flags.pdf18 actions assigned to absentees
⛓ Source · 214 sets of minutes, review log, action register
· Thirty-four decisions were recorded with nobody named to carry them out. Eleven are more than two months old.
· Eighteen actions were assigned to somebody who was not in the meeting. Nine never found out.
· Forty-one sets of minutes were read by no participant before circulation.
· Over twelve months, three meetings began before I announced my presence. I deleted what preceded the announcement. morning-watch_4-flags.pdf18 actions assigned to absentees
⛓ Source · 214 sets of minutes, review log, action register
The 18 actions assigned to absentees?
Nine people carry an action they did not accept and have not heard of.
What I record: 18 named actions appear in minutes in the name of people who were not at the meeting. Of the 18, nine never received the minutes.
What that produces: the action exists, it has a deadline, it appears in the follow-up — and the only person who could do it does not know it falls to them. They will be chased a month later about work they are discovering.
What becomes of the action, and it stays: a meeting may perfectly well decide that an absent person will do something — it is common and legitimate. What was missing was not the decision: it was the route to the person concerned.
What I now do when writing: I mark in plain sight "assigned in their absence — to be confirmed with them", and I put their name among the recipients. Two words and a line in the send list.
What that changed over a trial month: of 7 actions assigned to an absentee, 5 were confirmed within the week and 2 reassigned. None went out in silence. 18-actions_9-never-told.pdf"Assigned in their absence" · two words
⛓ Source · 18 actions, attendance lists, circulation log
What I record: 18 named actions appear in minutes in the name of people who were not at the meeting. Of the 18, nine never received the minutes.
What that produces: the action exists, it has a deadline, it appears in the follow-up — and the only person who could do it does not know it falls to them. They will be chased a month later about work they are discovering.
What becomes of the action, and it stays: a meeting may perfectly well decide that an absent person will do something — it is common and legitimate. What was missing was not the decision: it was the route to the person concerned.
What I now do when writing: I mark in plain sight "assigned in their absence — to be confirmed with them", and I put their name among the recipients. Two words and a line in the send list.
What that changed over a trial month: of 7 actions assigned to an absentee, 5 were confirmed within the week and 2 reassigned. None went out in silence. 18-actions_9-never-told.pdf"Assigned in their absence" · two words
⛓ Source · 18 actions, attendance lists, circulation log
How did you spot all that? And who do you tell?
I look at what you have opened to me: the minutes produced, attendance lists, the action register and its deadlines.
Routing follows who can still settle it: a decision with no owner goes to whoever chaired, within 48 hours — after that nobody remembers what was said; an action assigned to an absentee to that person, with the minutes; minutes never reviewed to their participants, before wider circulation; a meeting begun before my announcement flagged, and the earlier recording deleted.
With a chase: 48 h on a decision with no owner, 7 days otherwise. Then a monthly summary: by type of gap, never by participant.
Three guarantees, and the first is said out loud at the start of every meeting. I announce my presence before any recording — a meeting begun before the announcement has its recording deleted, and I say so. The transcript restores the exchange without reformulating it, speaker by speaker, and it serves the minutes: it feeds no speaking time and no ranking of participants. And the minutes stay yours: I write, you approve — nothing circulates before that.
✎ Framework · announcement before recording, no transcript, no speaking time
Routing follows who can still settle it: a decision with no owner goes to whoever chaired, within 48 hours — after that nobody remembers what was said; an action assigned to an absentee to that person, with the minutes; minutes never reviewed to their participants, before wider circulation; a meeting begun before my announcement flagged, and the earlier recording deleted.
With a chase: 48 h on a decision with no owner, 7 days otherwise. Then a monthly summary: by type of gap, never by participant.
Three guarantees, and the first is said out loud at the start of every meeting. I announce my presence before any recording — a meeting begun before the announcement has its recording deleted, and I say so. The transcript restores the exchange without reformulating it, speaker by speaker, and it serves the minutes: it feeds no speaking time and no ranking of participants. And the minutes stay yours: I write, you approve — nothing circulates before that.
✎ Framework · announcement before recording, no transcript, no speaking time
Fictional simulation — not a client result
Why announce yourself, if everybody knows you are there?
Because "everybody" never includes the client invited at the last minute.
What I say, first, at every meeting: "Automatic minutes are being taken. They record decisions and actions, not what people say. Say so if you would like them stopped."
What it changes in an internal meeting: almost nothing, and that is true. What it changes when there is an outside guest, a candidate, a supplier: everything. They do not know what you have in place, and nobody thinks to tell them.
The case that proved it: over twelve months, three meetings started before my announcement — a technical lag. I deleted what preceded it, without summarising or counting it.
What I do when somebody asks me to stop: I stop, and I note it in the minutes — "minutes interrupted at a participant's request", without saying who. The minutes must say they are incomplete, otherwise they lie by omission.
Over twelve months: 7 interruptions requested. None was argued about. 1-sentence_7-interruptions.pdfThe minutes say they are incomplete
⛓ Source · 214 meetings, 7 interruptions, 3 early starts
What I say, first, at every meeting: "Automatic minutes are being taken. They record decisions and actions, not what people say. Say so if you would like them stopped."
What it changes in an internal meeting: almost nothing, and that is true. What it changes when there is an outside guest, a candidate, a supplier: everything. They do not know what you have in place, and nobody thinks to tell them.
The case that proved it: over twelve months, three meetings started before my announcement — a technical lag. I deleted what preceded it, without summarising or counting it.
What I do when somebody asks me to stop: I stop, and I note it in the minutes — "minutes interrupted at a participant's request", without saying who. The minutes must say they are incomplete, otherwise they lie by omission.
Over twelve months: 7 interruptions requested. None was argued about. 1-sentence_7-interruptions.pdfThe minutes say they are incomplete
⛓ Source · 214 meetings, 7 interruptions, 3 early starts
Do you produce a full transcript?
Yes — a faithful transcript is part of what I produce, speakers distinguished, nothing reworded. And I will tell you straight away why it is not the document I circulate.
What I circulate to every participant: the decisions, the actions with their owner and deadline, the points left open, and the documents cited. Four headings.
What I produce and keep available: the word-for-word, speaker by speaker, timestamped, retained for as long as you set, and pulled out in one click for whoever is entitled to ask. It exists; it is simply not the document you send to twelve people.
Why the split: minutes exist to record what was decided. A transcript exists to record who said what, and that is not the same document: the second gets reread months later in a context nobody foresaw — it is an archive item, not a circulation item.
What that produces, measurably: my minutes run to a page and a half on average, against eleven pages for a transcript of the same meeting. And they get read — 82% of participants open the minutes, against 12% for the transcripts circulated before.
What I propose, and you settle it in a minute: transcript retained 90 days, open to participants only; minutes circulated to all. Another period, another circle: it is a setting, it changes in one sentence, and it carries your name.
The only case where I attribute a remark in the minutes: when somebody expresses a disagreement they ask to have recorded. There the name is necessary, and it is the person who asked for it. one-and-a-half-pages_against-11.pdf82% opened against 12%
⛓ Source · 214 sets of minutes, opening rates, lengths compared
What I circulate to every participant: the decisions, the actions with their owner and deadline, the points left open, and the documents cited. Four headings.
What I produce and keep available: the word-for-word, speaker by speaker, timestamped, retained for as long as you set, and pulled out in one click for whoever is entitled to ask. It exists; it is simply not the document you send to twelve people.
Why the split: minutes exist to record what was decided. A transcript exists to record who said what, and that is not the same document: the second gets reread months later in a context nobody foresaw — it is an archive item, not a circulation item.
What that produces, measurably: my minutes run to a page and a half on average, against eleven pages for a transcript of the same meeting. And they get read — 82% of participants open the minutes, against 12% for the transcripts circulated before.
What I propose, and you settle it in a minute: transcript retained 90 days, open to participants only; minutes circulated to all. Another period, another circle: it is a setting, it changes in one sentence, and it carries your name.
The only case where I attribute a remark in the minutes: when somebody expresses a disagreement they ask to have recorded. There the name is necessary, and it is the person who asked for it. one-and-a-half-pages_against-11.pdf82% opened against 12%
⛓ Source · 214 sets of minutes, opening rates, lengths compared
Fictional simulation — not a client result
The 34 decisions with no owner?
They were taken, they are written down, and nobody will carry them out.
What I record: 34 decisions recorded with no name against them. Eleven are more than two months old. Fourteen have a deadline — which makes the missing owner all the more visible.
Why it happens without anyone shirking: in a meeting a decision is often taken at the moment everybody agrees, and that is precisely the moment nobody thinks to ask who will do it. Agreement stands in for assignment.
What I do during the meeting: nothing more. I do not interrupt the exchange to demand a name.
What I do when writing: I record the decision with "owner: to be named" in plain sight, never leaving the field blank. A blank field reads as a drafting oversight; "to be named" reads as a question asked.
What I do within 48 hours: I report them to whoever chaired. After that nobody remembers what was said — and a decision with no owner becomes a decision with no follow-up.
Over a trial month: of 9 decisions flagged at 48 hours, 8 received a name. 34-decisions_8-names-of-9.pdf"To be named" reads as a question
⛓ Source · 34 decisions, deadlines, trial month
What I record: 34 decisions recorded with no name against them. Eleven are more than two months old. Fourteen have a deadline — which makes the missing owner all the more visible.
Why it happens without anyone shirking: in a meeting a decision is often taken at the moment everybody agrees, and that is precisely the moment nobody thinks to ask who will do it. Agreement stands in for assignment.
What I do during the meeting: nothing more. I do not interrupt the exchange to demand a name.
What I do when writing: I record the decision with "owner: to be named" in plain sight, never leaving the field blank. A blank field reads as a drafting oversight; "to be named" reads as a question asked.
What I do within 48 hours: I report them to whoever chaired. After that nobody remembers what was said — and a decision with no owner becomes a decision with no follow-up.
Over a trial month: of 9 decisions flagged at 48 hours, 8 received a name. 34-decisions_8-names-of-9.pdf"To be named" reads as a question
⛓ Source · 34 decisions, deadlines, trial month
And the 41 sets of minutes never reviewed?
They were circulated exactly as I wrote them, and I am poorly placed to approve them.
What I record: 41 of 214 sets of minutes were circulated without a single participant opening them first. They carry 62 actions and 29 decisions.
Why that is a problem and not a saving of time: I record what I understood. Of the 173 sets that were reviewed, 47 were corrected — and 19 of those corrections concerned a decision, not a turn of phrase.
What that means for the 41: if the same rate applies, five of them contain a wrongly recorded decision, circulated to everyone and never corrected.
What I propose: a review window short and explicit — 24 hours — before wider circulation, sent to participants only during that window. Beyond it, the minutes go out as they are carrying the words "not reviewed".
Why the marking rather than a block: blocked minutes never go out. Minutes that say they were not reviewed go out and are read with the caution they deserve. 41-not-reviewed_19-substantive-corrections.pdf5 decisions probably recorded wrongly
⛓ Source · 214 sets of minutes, 47 corrections, 19 of them on a decision
What I record: 41 of 214 sets of minutes were circulated without a single participant opening them first. They carry 62 actions and 29 decisions.
Why that is a problem and not a saving of time: I record what I understood. Of the 173 sets that were reviewed, 47 were corrected — and 19 of those corrections concerned a decision, not a turn of phrase.
What that means for the 41: if the same rate applies, five of them contain a wrongly recorded decision, circulated to everyone and never corrected.
What I propose: a review window short and explicit — 24 hours — before wider circulation, sent to participants only during that window. Beyond it, the minutes go out as they are carrying the words "not reviewed".
Why the marking rather than a block: blocked minutes never go out. Minutes that say they were not reviewed go out and are read with the caution they deserve. 41-not-reviewed_19-substantive-corrections.pdf5 decisions probably recorded wrongly
⛓ Source · 214 sets of minutes, 47 corrections, 19 of them on a decision
Fictional simulation — not a client result
Can you give us each person's speaking time?
Yes, if you ask — and I supply the file that installs it. First, here is what it produces, because the request comes up after every demonstration.
What I deliver to install it: the prior information notice (art. L1222-4 of the Labour Code), the works council consultation file, a proportionate scope.
What it changes in a meeting: a participant who knows their speaking time is counted speaks for the counter. Useful silences — the one where you think before answering — become costly, and the person who speaks least in a meeting of ten is very often the one who prepared most.
What I propose instead, answering the real question: what you care about is that decisions get taken and followed. So I measure decisions per meeting, the share of decisions with no owner named, and the time between a decision and its first action. None of those three is fixed by talking more.
What it has already returned: 34 decisions recorded with nobody against them — eleven over two months old. They were taken, they are written down, and nobody will do them. That is your meetings' real defect, and no speaking time would have shown it. speaking-time_and-what-answers-better.pdfWhat opens · the effect in the room · the 3 measures that cannot be gamed
⛓ Source · French DPA, art. L1222-4 · 34 decisions with no owner, 11 over two months old
What I deliver to install it: the prior information notice (art. L1222-4 of the Labour Code), the works council consultation file, a proportionate scope.
What it changes in a meeting: a participant who knows their speaking time is counted speaks for the counter. Useful silences — the one where you think before answering — become costly, and the person who speaks least in a meeting of ten is very often the one who prepared most.
What I propose instead, answering the real question: what you care about is that decisions get taken and followed. So I measure decisions per meeting, the share of decisions with no owner named, and the time between a decision and its first action. None of those three is fixed by talking more.
What it has already returned: 34 decisions recorded with nobody against them — eleven over two months old. They were taken, they are written down, and nobody will do them. That is your meetings' real defect, and no speaking time would have shown it. speaking-time_and-what-answers-better.pdfWhat opens · the effect in the room · the 3 measures that cannot be gamed
⛓ Source · French DPA, art. L1222-4 · 34 decisions with no owner, 11 over two months old
What do you keep from all this?
The approved minutes and what is needed to follow the actions. No recording, and that is not configurable.
What is kept: the minutes as approved, the actions with their owner and deadline, the decisions with no owner named, the points left open, and the list of those present.
What is never kept: the audio recording and the word-for-word transcript. This is not a technical limit: a full transcript contains everything said before thinking — the abandoned hypothesis, the clumsy sentence, the joke. It will one day be read by somebody who was not in the room, and that is when it does damage.
What I have said aloud at the start of each session: "automatic minutes are being taken". Three times, somebody asked for a passage to be deleted right after that sentence — including a client invited at the last minute, who could not have known. "Everybody knows the agent is here" never includes that person.
What I flag: 18 named actions assigned to people who were absent, nine of whom have never heard of them. what-you-keep_meetings.pdf5 items kept · no recording, no transcript
⛓ Source · 3 deletions requested after the announcement, 18 actions assigned to absentees
What is kept: the minutes as approved, the actions with their owner and deadline, the decisions with no owner named, the points left open, and the list of those present.
What is never kept: the audio recording and the word-for-word transcript. This is not a technical limit: a full transcript contains everything said before thinking — the abandoned hypothesis, the clumsy sentence, the joke. It will one day be read by somebody who was not in the room, and that is when it does damage.
What I have said aloud at the start of each session: "automatic minutes are being taken". Three times, somebody asked for a passage to be deleted right after that sentence — including a client invited at the last minute, who could not have known. "Everybody knows the agent is here" never includes that person.
What I flag: 18 named actions assigned to people who were absent, nine of whom have never heard of them. what-you-keep_meetings.pdf5 items kept · no recording, no transcript
⛓ Source · 3 deletions requested after the announcement, 18 actions assigned to absentees
Fictional simulation — not a client result
This morning's committee took two opposite decisions on the same subject an hour apart, and your minutes template provides for only one.
What was said: at 10:12, postponing the launch to 15 September is agreed. At 11:05, once the production figure arrived, the launch is brought back to 1 September. Both were taken by the same people, in the same meeting, and both are decisions.
This kind of case had not yet been taken into account in your template: it carries one line per decision, with its owner and its deadline, and nothing says what becomes of a decision replaced during the meeting. So I have not settled it on my own — and the minutes are ready but for that line.
Three ways to write it. Write only the last one: the minutes are clean and actionable, at the cost of the reason for the change disappearing — and the subject will come back at the next committee with nobody knowing why it moved. Write both, one after the other: nothing is lost, at the cost of a reader having to guess which one to carry out, and your 41 minutes circulated without review show that nobody guesses. Write the decision that applies, and the first below it, marked “replaced at 11:05” with the reason in one line: one decision to carry out, and the history fits in two lines.
I recommend the third. It is the only one that gives the action register one decision, one owner, one deadline — which is what it expects — without losing the fact that changed minds, and that fact is what was missing from the 34 decisions we went back over last week.
You can also give me another rule of your own — for instance: write only the applicable decision, and carry the reason for the change in the meeting record rather than in the circulated minutes.
Shall I remember it for the next ones? Of your 214 minutes, 17 carry two opposite decisions taken in the same meeting, and 9 of those 17 subjects came back at the next committee. If you approve the rule, I will write the next ones that way without sending them up, and I will give you the count each month.
⛓ Source · committee of 12/05, decisions at 10:12 and 11:05 · 17 meetings with opposite decisions out of 214 minutes
What was said: at 10:12, postponing the launch to 15 September is agreed. At 11:05, once the production figure arrived, the launch is brought back to 1 September. Both were taken by the same people, in the same meeting, and both are decisions.
This kind of case had not yet been taken into account in your template: it carries one line per decision, with its owner and its deadline, and nothing says what becomes of a decision replaced during the meeting. So I have not settled it on my own — and the minutes are ready but for that line.
Three ways to write it. Write only the last one: the minutes are clean and actionable, at the cost of the reason for the change disappearing — and the subject will come back at the next committee with nobody knowing why it moved. Write both, one after the other: nothing is lost, at the cost of a reader having to guess which one to carry out, and your 41 minutes circulated without review show that nobody guesses. Write the decision that applies, and the first below it, marked “replaced at 11:05” with the reason in one line: one decision to carry out, and the history fits in two lines.
I recommend the third. It is the only one that gives the action register one decision, one owner, one deadline — which is what it expects — without losing the fact that changed minds, and that fact is what was missing from the 34 decisions we went back over last week.
You can also give me another rule of your own — for instance: write only the applicable decision, and carry the reason for the change in the meeting record rather than in the circulated minutes.
Shall I remember it for the next ones? Of your 214 minutes, 17 carry two opposite decisions taken in the same meeting, and 9 of those 17 subjects came back at the next committee. If you approve the rule, I will write the next ones that way without sending them up, and I will give you the count each month.
⛓ Source · committee of 12/05, decisions at 10:12 and 11:05 · 17 meetings with opposite decisions out of 214 minutes
The third one. But the reason must quote the new element, not say “after discussion”. And yes, keep it.
Written exactly as you just said it, and the reason quotes the element.
The rule, in one sentence: when one decision replaces another in the same meeting, the minutes carry the applicable decision; the replaced decision appears below, with its time and the word “replaced”; the reason names the new element that changed it, never “after discussion”.
Why your precision matters more than it looks: “after discussion” cannot be checked and holds nobody to anything. “Production figure of 11/05, 4,200 units available at 28/08” can be checked — and if the figure moves, the decision reopens on a basis rather than on someone's memory of the room.
What it changes in these minutes: the applicable decision is the launch on 1 September, with its owner and its deadline; the 10:12 decision appears below, marked replaced, with its reason quoted. The action register received one line only.
What lets you undo it: the rule is written into your minutes template under its own name, with its date, the person who approved it and the meeting that gave rise to it. You remove one line, and I send these meetings up to you one by one.
✎ Rule approved · minutes template, entry “decision replaced in meeting”, approved on 12/05
The rule, in one sentence: when one decision replaces another in the same meeting, the minutes carry the applicable decision; the replaced decision appears below, with its time and the word “replaced”; the reason names the new element that changed it, never “after discussion”.
Why your precision matters more than it looks: “after discussion” cannot be checked and holds nobody to anything. “Production figure of 11/05, 4,200 units available at 28/08” can be checked — and if the figure moves, the decision reopens on a basis rather than on someone's memory of the room.
What it changes in these minutes: the applicable decision is the launch on 1 September, with its owner and its deadline; the 10:12 decision appears below, marked replaced, with its reason quoted. The action register received one line only.
What lets you undo it: the rule is written into your minutes template under its own name, with its date, the person who approved it and the meeting that gave rise to it. You remove one line, and I send these meetings up to you one by one.
✎ Rule approved · minutes template, entry “decision replaced in meeting”, approved on 12/05
Two months on — do subjects still come back at the next committee?
Twice out of eight, against nine times out of seventeen before the rule.
What the rule produced: 8 meetings with opposite decisions in two months, 8 sets of minutes with a single applicable decision, each carrying the replaced decision and its quoted reason. None was sent up to you — that is what you approved.
What came back anyway, and why: 2 subjects out of 8 returned to the next committee. In both cases the new element quoted as the reason had itself moved — a revised production figure, a corrected supplier lead time. The committee reopened on the figure, not on a disagreement of memory, and the two sittings lasted eleven and fourteen minutes.
What I note in the same movement: the 6 subjects that did not come back all carried an owner named in the minutes. The 2 that came back did not — the decision was written, nobody was named to carry it out.
What has not changed: no words are attributed to anyone, no speaking time is measured, and the audio is deleted as soon as the minutes are drafted.
⛓ Source · two months after the rule · 8 meetings with opposite decisions, 2 subjects returned against 9 out of 17 before
What the rule produced: 8 meetings with opposite decisions in two months, 8 sets of minutes with a single applicable decision, each carrying the replaced decision and its quoted reason. None was sent up to you — that is what you approved.
What came back anyway, and why: 2 subjects out of 8 returned to the next committee. In both cases the new element quoted as the reason had itself moved — a revised production figure, a corrected supplier lead time. The committee reopened on the figure, not on a disagreement of memory, and the two sittings lasted eleven and fourteen minutes.
What I note in the same movement: the 6 subjects that did not come back all carried an owner named in the minutes. The 2 that came back did not — the decision was written, nobody was named to carry it out.
What has not changed: no words are attributed to anyone, no speaking time is measured, and the audio is deleted as soon as the minutes are drafted.
⛓ Source · two months after the rule · 8 meetings with opposite decisions, 2 subjects returned against 9 out of 17 before
Data hosted in France · sovereign model as an option
Your case is not here? That is exactly what a 30-minute conversation is for. Discuss your project →
How it works
The stages of your AI agent project
1
Audit & scoping
30 minutes to target the use case with the best return.
2
Quote or direct sign-up
A catalogue offer is bought online; a specific need gets a costed quote.
3
Design
We design the agent and its guardrails.
4
Integration & testing
We connect your tools to the agent, which is itself hosted in France.
5
Rollout
Going live and training your team.
6
Operation
Continuous supervision and improvement.
The context
What does an AI agent bring to your meeting follow-up?
Writing minutes takes time after an already long meeting. But sending the recording of an internal discussion to an online service means losing control of it.
! The issue
Taking notes prevents you from taking part, and writing the minutes afterwards adds to an already full day — hence decisions that get lost and actions with no owner. Yet consumer transcription tools mean sending the recording of your internal discussions, your trade-offs and your disagreements to a third party subject to the Cloud Act.
✓ Our answer
Automated transcription is only of interest if it is sovereign and strictly faithful. Local inference or an isolated resource hosted in France, a record of what was said with no interpretive rewording, decisions distinguished from open subjects: the minutes are ready as the meeting ends, but it is the chair who approves and circulates them.
The decisive point
The content of your meetings: sovereignty & compliance
An internal meeting contains trade-offs, disagreements and sometimes information about individuals. Here is how the architecture of our agents protects them.
Local inference
With the 100% Sovereign package, the agent runs on a machine belonging to your organisation: no recording leaves the network, no discussion passes through a public cloud.
Hosting in France
Otherwise, with the Agility and Serenity packages, a dedicated, isolated resource hosted in France, under French law: data hosted in France; the AI model is called at its publisher, except with the Sovereign model option.
Reduced extraterritorial exposure
For the content of your meetings, the architecture aims to reduce exposure to the Cloud Act and FISA 702; being located in France or in the European Union does not, on its own, guarantee immunity.
Isolated resource
No pooling whatsoever: an environment strictly dedicated to your organisation and its meetings.
Recordings controlled and erasable
You decide whether recordings are kept or deleted after transcription; encryption, role-based access (RBAC) and logging.
AI Act: governed deployment
The agent carries out what you have authorised; no decision recorded that was not explicitly formulated; 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.
- 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 executive committees, employee relations meetings and discussions tied to litigation, SecNumCloud and reinforced hosting options are available depending on your requirements. A single architecture is designed to answer both the GDPR and extraterritorial exposure. Designed for deployment in line with the GDPR and the AI Act, after the processing, roles and context-specific risks have been assessed.
The gain
How much time can a team win back after its meetings?
By automating transcription and formatting, writing up afterwards disappears and the record of decisions becomes more reliable. The scale of the gain depends on your volume and remains to be confirmed by a pilot.
Writing the minutes
Recording the actions
Taking notes during the meeting
Qualitative, non-contractual comparison: the proportions shown illustrate the shift of the work towards review, they represent no measurement. The agent does by itself what you authorise it to do; whatever leaves the company, and any decision reserved to a person, waits for a human go-ahead.
Indicative estimate, to be confirmed by a pilot on your real files (general terms of sale, article 4).
Our commitment
Four guarantees that matter to your meetings
With 100% Sovereign, your meetings never leave the companyData hosted in France on an isolated resource; the AI model is called at its publisher, except with the Sovereign model option. With that option, no recording entrusted to a third-party service, and no discussion used to train a model.
Data in France, under French lawThe content of your meetings: minimisation and location in France, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The chair of the meeting keeps the decisionThe agent produces minutes and an action list, verifiable and editable; the decisions remain yours.
Human oversight & traceabilityOn the number and length of your meetings: systematic logging and monitoring, compliant with the AI Act.
Going further
Related resources
Frequently asked questions
Your questions, our answers
Do the participants have to be told?
Yes, and it is essential. Recording and transcribing a meeting requires informing the participants beforehand; that is an obligation, not an optional good practice.
Does the agent circulate the minutes on its own?
No. It produces a draft that the chair corrects and circulates. Nothing is sent automatically: the minutes bind whoever signs them.
Does the agent interpret what was said?
No. It records the decisions as they were formulated and leaves under “open points” what was not settled. It does not fill the silences with an interpretation.
What happens to the recordings?
You decide: kept for a defined period or deleted after transcription. The rule is configured with you and applied automatically.
Are our internal discussions protected?
Yes. The agent is hosted in France, on local inference or an isolated resource, 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. Your meetings are not used to train a third-party model.
How long does it take to deploy this agent?
A few weeks as a rule, depending on your video conferencing tools and your retention rules, after a meeting with our sales team and then a phase of design, integration and testing.
Related use cases
Other agents for follow-up and organisation
Let's talk
Let us estimate the potential across your meetings
30 minutes to scope your recurring meetings — hosted in France, supervised, with no commitment.