We build Agents. You build the Future.
● B2B offer — Quality & continuous improvementQuality: every finding recorded, every action tracked
A non-conformity is worth what is done with it: a clear finding, a cause looked for, a corrective action and a check that it worked. Your agent records the finding from the reporting channel, attaches the documents, tracks the open actions and alerts you to deadlines coming up. Data hosted in France; the AI model is called at its publisher, except with the Sovereign model option. With the 100% Sovereign package, your findings and what follows from them stay with you. The quality manager decides on the actions and on closing them.
Updated on
Deployed in a few weeks
Quality tracking · hosted in France
Fictional simulation — not a client result
Record the non-conformity reported this morning on line 3.
Record created: the finding taken from the report, product and batch concerned, attachments linked.
The process concerned is identified from your quality framework.
Two similar non-conformities have been recorded on the same process: they are flagged for comparison.
🔗 Sourced · the report and the quality framework
The process concerned is identified from your quality framework.
Two similar non-conformities have been recorded on the same process: they are flagged for comparison.
🔗 Sourced · the report and the quality framework
Where do the open corrective actions stand?
State of the actions: completed, in progress, and those whose deadline is approaching, with their owner.
Each refers back to the original record and the documents filed.
Declaring an action effective and closing a record rest on your judgement.
✎ Support · state presented, closure decided by you
Each refers back to the original record and the documents filed.
Declaring an action effective and closing a record rest on your judgement.
✎ Support · state presented, closure decided by you
Data hosted in France · sovereign model as an option
Quality records hosted in France
Hosted in FranceAI model called at its publisher, except with the Sovereign model option
Quality findings protected
GDPR & AI Act: governed deploymentTraceability & human oversight
TurnkeyDesigned, installed and operated for you
The quality manager 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 quality agent records your non-conformities from your reporting channels, identifies the process concerned from your framework, attaches the documents and tracks the corrective actions with their deadlines. Similar findings are flagged for comparison. Data hosted in France; the AI model is called at its publisher, except with the Sovereign model option. With the 100% Sovereign package, your findings stay with you, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
100%
hosted in France in the target architecture
0
transfer outside the EU with the Sovereign model option (target architecture)
3
quality uses ready to deploy
0
decision with a legal or significant effect taken by an agent
These figures describe our offer, not results measured at a client. How large the gain is on your number of processes and findings tracked 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 quality agent (findings, processes, corrective actions), installed and operated for you.
Agility
Setup + controlled subscription
2,775 € excl. VAT setup
then 277 € 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
432 € 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
4,625 € excl. VAT setup
then 381 € 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, the whole cycle of a non-conformity, with a link to a neighbouring agent. 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 277 € excl. VAT / month
Recording the findings
Creates the record from the report and attaches the documents filed.
Linking to the process
Identifies the process concerned from your quality framework.
Tracking corrective actions
Shows the state of the open actions, their owner and their deadlines.
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
Explain alerts, scenarios and operational trade-offs
What the agent must be connected to
These 2 connections are required for the agent to work. They concern your information system and are scoped during the audit.
Connect ERP/WMS/MES/IoT without creating double entry
Honour constraints, statuses, approvals and error recovery
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.
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.
Quality tracking · hosted in France
Fictional simulation — not a client result
Three flags this morning across your 214 open non-conformities.
· 61 were closed on an action marked "done", with no check that it worked. Nine came back.
· 23 non-conformities declared separately share one cause. They are being handled 23 times.
· One workshop declares three times fewer than the others since it started being tracked on its NC count. morning-watch_3-flags.pdf61 unverified closures · 9 recurrences
⛓ Source · 214 open NCs, closure history, recurrence dates
· 61 were closed on an action marked "done", with no check that it worked. Nine came back.
· 23 non-conformities declared separately share one cause. They are being handled 23 times.
· One workshop declares three times fewer than the others since it started being tracked on its NC count. morning-watch_3-flags.pdf61 unverified closures · 9 recurrences
⛓ Source · 214 open NCs, closure history, recurrence dates
Closed but back again: what does that mean?
That the action was carried out, and it did not work — with nobody finding out.
What I record: of 214 NCs, 61 were closed on the date the corrective action was carried out. Not on the date somebody checked it produced the intended effect. Those are two different things, and the average gap is zero days — which means the check never happened.
What that produces: 9 of those 61 gave rise to a new NC within six months, opened under another number. In the tables, that is two non-conformities handled. On the shop floor, it is the same problem twice.
What I do: I separate three states — action defined, action carried out, effectiveness verified. An NC is only closed at the third. I close nothing myself: I flag NCs stuck at the second state for over 30 days.
What it changed: the number of open NCs rose by 61 in the first month. No new one had appeared — they were already there, counted as closed. 61-closed_9-came-back.pdfAction carried out and effectiveness verified are two different dates
⛓ Source · 61 closures with no verification, 9 recurrences in six months
What I record: of 214 NCs, 61 were closed on the date the corrective action was carried out. Not on the date somebody checked it produced the intended effect. Those are two different things, and the average gap is zero days — which means the check never happened.
What that produces: 9 of those 61 gave rise to a new NC within six months, opened under another number. In the tables, that is two non-conformities handled. On the shop floor, it is the same problem twice.
What I do: I separate three states — action defined, action carried out, effectiveness verified. An NC is only closed at the third. I close nothing myself: I flag NCs stuck at the second state for over 30 days.
What it changed: the number of open NCs rose by 61 in the first month. No new one had appeared — they were already there, counted as closed. 61-closed_9-came-back.pdfAction carried out and effectiveness verified are two different dates
⛓ Source · 61 closures with no verification, 9 recurrences in six months
What exactly is the agent for?
It keeps the tracking and flags what has stopped moving. It grades no non-conformity and closes none.
What it tracks: the real state of each NC (defined, carried out, verified), the delays, the recurrences, causes shared between several NCs, and audit deadlines.
Routing follows who can decide: an NC with no defined action goes to the owner of the process concerned; an action carried out but unverified to whoever must verify it; a cause shared by several NCs to the quality manager, because that is a handling decision, not a correction; an NC late against an audit deadline to management, with the days remaining.
With a monthly summary: NCs by state, overdue delays, recurrences, shared causes, and NCs nobody has picked back up.
What this tracking has already given back: 61 non-conformities counted as closed and still open, 9 back under another number, and 23 findings that share one single cause and are handled 23 times over. From tomorrow: 22 treatments avoided on that cause alone, a closure reached only once effectiveness is verified, and no more audit deadline discovered the week before — management gets the days remaining, not the overrun. The access is yours: findings, actions and deadlines opened by role, logged, withdrawn with a word; your non-conformities and what follows them stay with you. No NC is counted per person or per team, and that is not a reservation, it is the engine: the workshop declaring three times fewer than the others since it has been tracked on its NC count is the proof — what gets counted per team ends up not being declared, and an undeclared NC is a defect that leaves for the customer. Severity and cause stay with the people who know the process, and I hand them back in minutes: the finding, the resembling NCs already brought together, the action history and its verification date — the quality manager arbitrates instead of reconstructing. The next step is ready: the 23 findings sharing a cause are grouped into a single file and the 61 closures re-sorted by real state; tell me whether we reopen them in one go or process by process.
✎ Framework · no severity graded · no cause written · no count per team at this stage
What it tracks: the real state of each NC (defined, carried out, verified), the delays, the recurrences, causes shared between several NCs, and audit deadlines.
Routing follows who can decide: an NC with no defined action goes to the owner of the process concerned; an action carried out but unverified to whoever must verify it; a cause shared by several NCs to the quality manager, because that is a handling decision, not a correction; an NC late against an audit deadline to management, with the days remaining.
With a monthly summary: NCs by state, overdue delays, recurrences, shared causes, and NCs nobody has picked back up.
What this tracking has already given back: 61 non-conformities counted as closed and still open, 9 back under another number, and 23 findings that share one single cause and are handled 23 times over. From tomorrow: 22 treatments avoided on that cause alone, a closure reached only once effectiveness is verified, and no more audit deadline discovered the week before — management gets the days remaining, not the overrun. The access is yours: findings, actions and deadlines opened by role, logged, withdrawn with a word; your non-conformities and what follows them stay with you. No NC is counted per person or per team, and that is not a reservation, it is the engine: the workshop declaring three times fewer than the others since it has been tracked on its NC count is the proof — what gets counted per team ends up not being declared, and an undeclared NC is a defect that leaves for the customer. Severity and cause stay with the people who know the process, and I hand them back in minutes: the finding, the resembling NCs already brought together, the action history and its verification date — the quality manager arbitrates instead of reconstructing. The next step is ready: the 23 findings sharing a cause are grouped into a single file and the 61 closures re-sorted by real state; tell me whether we reopen them in one go or process by process.
✎ Framework · no severity graded · no cause written · no count per team at this stage
Fictional simulation — not a client result
How can 23 NCs share a cause without anybody seeing it?
Because they were declared on 23 dates, by 4 workshops, under 23 different wordings.
What I record: 23 NCs concern distinct symptoms — a surface defect, a dimensional deviation, two scrapped parts, a customer return — and all occur after a material batch change, within a window of under 48 hours.
What I do: I bring those 23 NCs together and pass the grouping to the quality manager, with the dates, batches and workshops.
What I put on the table, with figures, rather than an asserted cause: the hypothesis the data supports best — all 23 NCs occur within 48 h of a batch change, across 4 workshops and under 23 different wordings, which no card-by-card reading brings out — and the two checks that would settle it: the settings restored after maintenance over the period, and the batches delivered with no NC at all. The cause analysis that enters the audit file is signed by the team, the only ones who know the process and the material, and that is precisely what makes it defensible in session. I hand them all of the work except the conclusion.
What it gave: the analysis run by the team found two causes, not one — the batch for 16 NCs, a setting not restored after maintenance for 7. My grouping was useful and partly wrong, which is why it had to stay a flag.
What it avoids: 23 separate corrective actions, of which 21 would have treated a symptom. 23-ncs_2-causes-retained.pdfA coincidence of dates is not a cause
⛓ Source · 23 NCs grouped, 2 causes retained by the team, 21 actions avoided
What I record: 23 NCs concern distinct symptoms — a surface defect, a dimensional deviation, two scrapped parts, a customer return — and all occur after a material batch change, within a window of under 48 hours.
What I do: I bring those 23 NCs together and pass the grouping to the quality manager, with the dates, batches and workshops.
What I put on the table, with figures, rather than an asserted cause: the hypothesis the data supports best — all 23 NCs occur within 48 h of a batch change, across 4 workshops and under 23 different wordings, which no card-by-card reading brings out — and the two checks that would settle it: the settings restored after maintenance over the period, and the batches delivered with no NC at all. The cause analysis that enters the audit file is signed by the team, the only ones who know the process and the material, and that is precisely what makes it defensible in session. I hand them all of the work except the conclusion.
What it gave: the analysis run by the team found two causes, not one — the batch for 16 NCs, a setting not restored after maintenance for 7. My grouping was useful and partly wrong, which is why it had to stay a flag.
What it avoids: 23 separate corrective actions, of which 21 would have treated a symptom. 23-ncs_2-causes-retained.pdfA coincidence of dates is not a cause
⛓ Source · 23 NCs grouped, 2 causes retained by the team, 21 actions avoided
Could you not propose the analysis anyway?
I can supply everything it rests on. I cannot conclude it in your place.
What I supply: the grouped NCs, what they have in common and what separates them, the history of batches and interventions over the period, earlier NCs at the same place, and actions already tried on neighbouring cases with their outcome.
Why I stop there: a cause is established by going and looking, by stripping down, by rerunning the test. What I handle are records — that is, what was written down, not what happened.
The case that serves as my rule: on a run of scrapped parts, everything pointed to a worn tool — its age, the gradual drift, the last change date. The real cause was a sensor refitted wrongly after cleaning, invisible in every record I can read. An automatic analysis would have produced a coherent, sourced and wrong file.
What I do instead: I flag when a cause analysis has been followed by no recurrence for six months, and when it has been followed by a recurrence despite an action judged effective. The second is the only one that proves the analysis was wrong. what-i-supply_what-i-do-not-conclude.pdfA coherent, sourced and wrong file
⛓ Source · elements supplied, conclusion left to the team, 1 counter-example
What I supply: the grouped NCs, what they have in common and what separates them, the history of batches and interventions over the period, earlier NCs at the same place, and actions already tried on neighbouring cases with their outcome.
Why I stop there: a cause is established by going and looking, by stripping down, by rerunning the test. What I handle are records — that is, what was written down, not what happened.
The case that serves as my rule: on a run of scrapped parts, everything pointed to a worn tool — its age, the gradual drift, the last change date. The real cause was a sensor refitted wrongly after cleaning, invisible in every record I can read. An automatic analysis would have produced a coherent, sourced and wrong file.
What I do instead: I flag when a cause analysis has been followed by no recurrence for six months, and when it has been followed by a recurrence despite an action judged effective. The second is the only one that proves the analysis was wrong. what-i-supply_what-i-do-not-conclude.pdfA coherent, sourced and wrong file
⛓ Source · elements supplied, conclusion left to the team, 1 counter-example
Fictional simulation — not a client result
A workshop declares fewer: does that mean fewer errors?
Its scrap has not fallen. Only its declarations have.
What I record: since that workshop appeared in a tracking table by number of NCs declared, its declarations have fallen by two thirds. Over the same period, its scrap rate and customer returns are stable.
What that means: non-conformities are not fewer, they are less written down. A workshop compared on its NC count has a simple way to improve: declare fewer. And a quality system that no longer sees deviations shows exactly the same figures as one that no longer has any.
What I produce on no condition: an NC count per person, a ranking, a comparative trend, or an inference about someone. Per workshop, per team or per line, the aggregate exists only if you configure it, for authorised recipients and above a headcount threshold that prevents tracing back to a person. Declaration is what keeps the system alive: measuring it kills it.
What I count instead: NCs by cause, by process, by recurrence, and by handling time. A process has no bonus, no annual review and no reason to conceal.
What I do flag: a fall in declarations with no matching fall in output indicators — as here. That is a signal about the system, not about the workshop. declarations-down_outputs-stable.pdfA system that no longer sees deviations shows the figures of one that has none
⛓ Source · declarations down two thirds, scrap and returns stable
What I record: since that workshop appeared in a tracking table by number of NCs declared, its declarations have fallen by two thirds. Over the same period, its scrap rate and customer returns are stable.
What that means: non-conformities are not fewer, they are less written down. A workshop compared on its NC count has a simple way to improve: declare fewer. And a quality system that no longer sees deviations shows exactly the same figures as one that no longer has any.
What I produce on no condition: an NC count per person, a ranking, a comparative trend, or an inference about someone. Per workshop, per team or per line, the aggregate exists only if you configure it, for authorised recipients and above a headcount threshold that prevents tracing back to a person. Declaration is what keeps the system alive: measuring it kills it.
What I count instead: NCs by cause, by process, by recurrence, and by handling time. A process has no bonus, no annual review and no reason to conceal.
What I do flag: a fall in declarations with no matching fall in output indicators — as here. That is a signal about the system, not about the workshop. declarations-down_outputs-stable.pdfA system that no longer sees deviations shows the figures of one that has none
⛓ Source · declarations down two thirds, scrap and returns stable
And for the certification audit? We are asked for figures.
You will have them, and they will hold — because they are about handling, not about counts.
What I prepare: the real state of each NC with its three dates, handling times, recurrences, the actions whose effectiveness was verified and by whom, and the cause analyses with their supporting items.
What an auditor looks at: not how many NCs you have, but whether the ones you have are handled, verified and free of recurrence. A falling NC count with no documented recurrence is an unfavourable signal, not a good result.
What I flag before the deadline: NCs whose action is done but unverified, those carrying a cause analysis followed by a recurrence, and those open longer than your own target time.
What I hand you is the real state, and that is what holds in session: nothing is closed ahead of the audit, nothing is consolidated, no NC is removed from a count — a file more favourable than reality is precisely the one an auditor takes apart. And three weeks is the notice you need to put things right: for every NC done but unverified, the effectiveness check is already written — the measurement to repeat, the evidence to produce, the reference date, its three dates on file. All that is missing is the visit to the floor and the owner's signature to close it properly before the deadline. what-is-prepared-before-an-audit.pdfA falling count with no documented recurrence is an unfavourable signal
✎ Framework · no closure before an audit · no removal from a count
What I prepare: the real state of each NC with its three dates, handling times, recurrences, the actions whose effectiveness was verified and by whom, and the cause analyses with their supporting items.
What an auditor looks at: not how many NCs you have, but whether the ones you have are handled, verified and free of recurrence. A falling NC count with no documented recurrence is an unfavourable signal, not a good result.
What I flag before the deadline: NCs whose action is done but unverified, those carrying a cause analysis followed by a recurrence, and those open longer than your own target time.
What I hand you is the real state, and that is what holds in session: nothing is closed ahead of the audit, nothing is consolidated, no NC is removed from a count — a file more favourable than reality is precisely the one an auditor takes apart. And three weeks is the notice you need to put things right: for every NC done but unverified, the effectiveness check is already written — the measurement to repeat, the evidence to produce, the reference date, its three dates on file. All that is missing is the visit to the floor and the owner's signature to close it properly before the deadline. what-is-prepared-before-an-audit.pdfA falling count with no documented recurrence is an unfavourable signal
✎ Framework · no closure before an audit · no removal from a count
Fictional simulation — not a client result
An effectiveness check this morning falls outside your procedure, and the NC has not been closed for it.
What happened: NC 2026-0187 — a dimensional deviation on line 3 — carries a corrective action verified effective on 10/08, after 45 days of observation. While checking it, I found the same deviation, on the same station, on lines 1 and 5, which have declared nothing about it.
The rule to apply here has not yet been defined: your procedure says an NC closes at verified effectiveness, and it says nothing about the scope to keep when the check reveals the same deviation elsewhere. So 2026-0187 stays open until you settle it — and lines 1 and 5 are already documented.
Three ways forward. Close 2026-0187 and open two new NCs: each line carries its own file, action and check, at the cost of two full investigations and a root-cause analysis done three times. Widen 2026-0187 to all three lines and close it only at the third check: one root-cause analysis, one audit file — and your average handling time grows by as much. Close 2026-0187 and open a single preventive action on lines 1 and 5: that is the shortest path, but in your procedure a preventive action carries no effectiveness check, and the deviation's return would go unnoticed.
I recommend the second. It is the only one that rests the closure on effectiveness verified everywhere the deviation exists — and your system already knows how to hold one finding across several workshops, which is what brought the 23 NCs of a single batch change together.
You can also give me another rule of your own — for instance: the NC widens when the deviation is found on an identical station, and two separate NCs open when the stations differ.
Shall I remember it for the next cases? Of your 214 open NCs, 17 carry a deviation found on one line and present on at least one other. If you approve the rule, I will apply it to those 17 instead of sending them up to you one by one, and I will give you the count at each review.
⛓ Source · NC 2026-0187, check of 10/08, records from lines 1, 3 and 5 · 17 comparable NCs out of 214 open
What happened: NC 2026-0187 — a dimensional deviation on line 3 — carries a corrective action verified effective on 10/08, after 45 days of observation. While checking it, I found the same deviation, on the same station, on lines 1 and 5, which have declared nothing about it.
The rule to apply here has not yet been defined: your procedure says an NC closes at verified effectiveness, and it says nothing about the scope to keep when the check reveals the same deviation elsewhere. So 2026-0187 stays open until you settle it — and lines 1 and 5 are already documented.
Three ways forward. Close 2026-0187 and open two new NCs: each line carries its own file, action and check, at the cost of two full investigations and a root-cause analysis done three times. Widen 2026-0187 to all three lines and close it only at the third check: one root-cause analysis, one audit file — and your average handling time grows by as much. Close 2026-0187 and open a single preventive action on lines 1 and 5: that is the shortest path, but in your procedure a preventive action carries no effectiveness check, and the deviation's return would go unnoticed.
I recommend the second. It is the only one that rests the closure on effectiveness verified everywhere the deviation exists — and your system already knows how to hold one finding across several workshops, which is what brought the 23 NCs of a single batch change together.
You can also give me another rule of your own — for instance: the NC widens when the deviation is found on an identical station, and two separate NCs open when the stations differ.
Shall I remember it for the next cases? Of your 214 open NCs, 17 carry a deviation found on one line and present on at least one other. If you approve the rule, I will apply it to those 17 instead of sending them up to you one by one, and I will give you the count at each review.
⛓ Source · NC 2026-0187, check of 10/08, records from lines 1, 3 and 5 · 17 comparable NCs out of 214 open
The second. But the widening must be dated and carry who asked for it — otherwise, at audit, nobody will know why the NC stayed open five months. And yes, keep it.
Written exactly as you just said it, and it carries your correction.
The rule, in one sentence: an effectiveness check that reveals the same deviation on another station widens the original NC; the widening carries its date and the name of whoever asked for it; the NC closes only at the last effectiveness check of the widened scope.
Why your correction matters more than it looks: an auditor seeing an NC open for five months looks for a delay first. A dated, signed widening turns that delay into an owned scope, and that is what holds up in the room.
What it changes from this morning: 2026-0187 now carries lines 1, 3 and 5, the widening is dated 11/08 in your quality manager's name, and the two remaining checks are set at 45 days — the observation period used on line 3.
What lets you undo it: the rule is written into your procedure under its own name, with its date, the person who approved it and the NC that gave rise to it. You remove one line, and I go back to sending you these cases one by one.
✎ Rule approved · NC handling procedure, entry “scope widening”, approved on 11/08 by the quality manager
The rule, in one sentence: an effectiveness check that reveals the same deviation on another station widens the original NC; the widening carries its date and the name of whoever asked for it; the NC closes only at the last effectiveness check of the widened scope.
Why your correction matters more than it looks: an auditor seeing an NC open for five months looks for a delay first. A dated, signed widening turns that delay into an owned scope, and that is what holds up in the room.
What it changes from this morning: 2026-0187 now carries lines 1, 3 and 5, the widening is dated 11/08 in your quality manager's name, and the two remaining checks are set at 45 days — the observation period used on line 3.
What lets you undo it: the rule is written into your procedure under its own name, with its date, the person who approved it and the NC that gave rise to it. You remove one line, and I go back to sending you these cases one by one.
✎ Rule approved · NC handling procedure, entry “scope widening”, approved on 11/08 by the quality manager
The 45 days of observation — did you set those?
No — they were on line 3's own record, and that is exactly what made me look at the others.
What I looked at: the “observation period” field of your effectiveness check record, across the 214 open NCs, and the recurrences observed.
What comes out of it: the field is filled on 39 NCs and empty on 175. Across the 39, no recurrence in six months. Across the 175, 9 recurrences — the same nine that came back under another number.
What that means, and it is not a question of diligence: a check with no observation period happens on the day of the action, when nothing has had time to come back. The field exists, it is simply optional, and an optional field on a 22-line record gets skipped.
What I propose, and it is one line to change on your side: that the observation period become mandatory in order to close, and that it arrive pre-filled with the median observed on NCs from the same process — 42 days over your last twelve months. You keep it or you change it, but it no longer stays empty.
What it is worth: those 9 recurrences each reopened a full file — 11 hours of handling on average, that is 99 hours over twelve months, without counting what the deviation cost on the shop floor while it was counted as settled.
The decision is yours: it is your procedure, a mandatory field is decided with the people who fill it in, and I do not touch it before you have written it.
✎ Proposal · “observation period” field filled on 39 NCs out of 214, 9 recurrences on the side where it is empty — the procedure belongs to the quality manager
What I looked at: the “observation period” field of your effectiveness check record, across the 214 open NCs, and the recurrences observed.
What comes out of it: the field is filled on 39 NCs and empty on 175. Across the 39, no recurrence in six months. Across the 175, 9 recurrences — the same nine that came back under another number.
What that means, and it is not a question of diligence: a check with no observation period happens on the day of the action, when nothing has had time to come back. The field exists, it is simply optional, and an optional field on a 22-line record gets skipped.
What I propose, and it is one line to change on your side: that the observation period become mandatory in order to close, and that it arrive pre-filled with the median observed on NCs from the same process — 42 days over your last twelve months. You keep it or you change it, but it no longer stays empty.
What it is worth: those 9 recurrences each reopened a full file — 11 hours of handling on average, that is 99 hours over twelve months, without counting what the deviation cost on the shop floor while it was counted as settled.
The decision is yours: it is your procedure, a mandatory field is decided with the people who fill it in, and I do not touch it before you have written it.
✎ Proposal · “observation period” field filled on 39 NCs out of 214, 9 recurrences on the side where it is empty — the procedure belongs to the quality manager
Fictional simulation — not a client result
We would like to track non-conformities by team. Is that possible?
Yes, it is lawful, and I set up the conditions that make it unassailable — and the last one, the anti-reidentification rule, is still to be defined with you: until it is, the team-level aggregate is blocked and referred for human validation.
What the French DPA restates: an employer has the power to frame and monitor staff activity. It is the conditions that are regulated, not the principle.
What I prepare for you: the prior information notice to the people concerned (art. L1222-4 of the Labour Code), the works council consultation file with the purpose, the data used and the retention period, and a proportionate scope — strictly the indicator your objective requires.
What it saves you: that file takes half a day to assemble by hand. I produce it filled in, dated and ready to present. On the four team indicators you introduced last year, no audit observation at all.
What I add: a measured observation before you decide. In a workshop tracked on its number of declared NCs, declarations fell by two thirds while scrap and customer returns stayed flat. Tracking the HANDLING rather than the COUNT gives the same steering without that effect — and that is what I propose by default. The choice stays yours, and I equip it either way. team-tracking_conditions-of-use.pdf3 conditions met, 1 to be defined · works council file supplied
⛓ Source · French DPA, art. L1222-4 · 4 team indicators, 0 audit observation
What the French DPA restates: an employer has the power to frame and monitor staff activity. It is the conditions that are regulated, not the principle.
What I prepare for you: the prior information notice to the people concerned (art. L1222-4 of the Labour Code), the works council consultation file with the purpose, the data used and the retention period, and a proportionate scope — strictly the indicator your objective requires.
What it saves you: that file takes half a day to assemble by hand. I produce it filled in, dated and ready to present. On the four team indicators you introduced last year, no audit observation at all.
What I add: a measured observation before you decide. In a workshop tracked on its number of declared NCs, declarations fell by two thirds while scrap and customer returns stayed flat. Tracking the HANDLING rather than the COUNT gives the same steering without that effect — and that is what I propose by default. The choice stays yours, and I equip it either way. team-tracking_conditions-of-use.pdf3 conditions met, 1 to be defined · works council file supplied
⛓ Source · French DPA, art. L1222-4 · 4 team indicators, 0 audit observation
And on audit day, what does it change concretely?
You walk in with the complete file, and the auditor gets every answer live.
What you have to hand: the real state of each NC with its three dates, times against your own target, the actions whose effectiveness was verified and by whom, the cause analyses with their supporting items, and recurrences attached to the original NC.
What it represented: preparing the last audit took two hours instead of two days — Claire Fontanel, your quality manager, spent those two days on the substance rather than reconstructing dates.
What the auditor looked at: whether the NCs you have are handled, verified and free of recurrence. On that point you show a complete chain — and a complete chain is worth more than a falling counter nobody can explain.
What I flag three weeks ahead: actions carried out whose effectiveness is still to be verified, and NCs open beyond your target time. Three weeks is exactly what it takes to clear them before the auditor arrives. what-you-keep_audit-file.pdf7 items to hand · retention set by you
⛓ Source · preparation cut from 2 days to 2 hours, alert 3 weeks ahead
What you have to hand: the real state of each NC with its three dates, times against your own target, the actions whose effectiveness was verified and by whom, the cause analyses with their supporting items, and recurrences attached to the original NC.
What it represented: preparing the last audit took two hours instead of two days — Claire Fontanel, your quality manager, spent those two days on the substance rather than reconstructing dates.
What the auditor looked at: whether the NCs you have are handled, verified and free of recurrence. On that point you show a complete chain — and a complete chain is worth more than a falling counter nobody can explain.
What I flag three weeks ahead: actions carried out whose effectiveness is still to be verified, and NCs open beyond your target time. Three weeks is exactly what it takes to clear them before the auditor arrives. what-you-keep_audit-file.pdf7 items to hand · retention set by you
⛓ Source · preparation cut from 2 days to 2 hours, alert 3 weeks ahead
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 quality system?
A quality system lives on steadiness: findings recorded as they come and actions tracked through to their verification.
! The issue
Recording a finding and tracking the actions are two routine acts, and it is that steadiness that makes the system valuable. The agent keeps it up: it creates the record from the report, attaches the documents, identifies the process concerned and shows at any moment the state of the open actions with their deadlines.
✓ Our answer
The quality manager has a system that is up to date and an automatic flagging of similar findings — valuable material for root-cause analysis. Deciding on a corrective action, declaring it effective and closing a record rest on their professional judgement. Data hosted in France on an isolated resource; the AI model is called at its publisher, except with the Sovereign model option. With the 100% Sovereign package, your findings and what follows from them do not leave the company.
The decisive point
The findings and what follows from them: sovereignty & compliance
Your quality findings and what follows from them are sensitive information, internally as well as towards your clients and your auditors. Here is how they are protected.
Local inference
With the 100% Sovereign package, the agent runs on a machine belonging to your organisation: no finding and no quality document leaves the network.
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 findings and what follows from them, 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: an environment strictly dedicated to your company and its quality framework.
Every record traced from end to end
The finding, the documents, the actions and the deadlines are kept with their history; encryption, role-based access and logging of every change.
AI Act: governed deployment
The agent carries out what you have authorised; no corrective action is decided and no record is closed 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.
- 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 non-conformities tied to a product recall, a major client complaint or an audit in progress, SecNumCloud and reinforced hosting are options 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 quality department give back to analysis?
By taking on the recording and the tracking, the effort shifts towards looking for causes and improving. How large the gain is depends on your volume and remains to be confirmed by a pilot.
Recording the findings
Tracking the open actions
Comparing similar findings
Indicative figures, not contractual, to be confirmed by a pilot on your number of processes and findings tracked. Deciding on a corrective action, declaring it effective and closing a record rest on the quality manager's judgement.
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 quality system
Your findings stay internalData 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 finding and no document entrusted to a third party, no data used to train a model.
Data in France, under French lawThe findings and what follows from them: minimisation and location in France, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The quality manager keeps the decisionThe agent produces non-conformity records that are tracked, which can be checked and altered; the decisions remain yours.
Human oversight & traceabilityOn your number of processes and findings tracked: systematic logging and tracking, in line with the AI Act.
Going further
Related resources
Frequently asked questions
Your questions, our answers
Does the agent decide on the corrective actions?
No. It records, links and tracks. Deciding on an action, declaring it effective and closing a record rest on the quality manager's judgement.
How does it identify the process concerned?
From your quality framework, as you have structured it. The link proposed stays changeable by the department.
What does it do with similar findings?
It flags them for comparison, without merging them. Grouping several findings under a single cause is an analysis that falls to you.
Is it compatible with an ISO certification?
The agent keeps the full history of every record — finding, documents, actions, deadlines — which makes preparing an audit easier. Your framework's requirements are taken in at the design stage.
Are our findings 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 quality data is 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 the number of processes covered and your reporting channels, after a meeting with our sales team then a phase of design, integration and testing.
Related use cases
Other agents for production
Let's talk
Let's size up the potential in your quality system
30 minutes to frame your processes and your reports — hosted in France, supervised, with no commitment.