● B2B offer — Litigation & procedure
Litigation: your exhibits found, your deadlines met
A litigation file is also won on command of its exhibits and observance of its time limits. Your agent questions the whole file, finds the exhibit cited with its number and its passage, and tracks the procedural deadlines ahead. Hosted in France: your defence strategies and your sensitive exhibits stay with you. The legal department keeps full command of the strategy.
Hosted in France
Case exhibits protected
GDPR & AI Act: governed deployment
Human oversight
Updated on
Deployed in a few weeks
Litigation documents · hosted in France
Where is the clause the other side is relying on, in our exhibits?
Two exhibits in the file contain that clause: the exhibit number, the page and the exact passage are stated.
An earlier version of the contract is also in the file, with the same clause worded differently.
Both wordings are set out side by side, with their dates.
🔗 Sourced · exhibits in the file, numbers stated
An earlier version of the contract is also in the file, with the same clause worded differently.
Both wordings are set out side by side, with their dates.
🔗 Sourced · exhibits in the file, numbers stated
What deadlines are coming up on this file?
The deadlines found in the file are set out with the step that starts them running and the exhibit they come from.
Judging a procedural time limit and settling the strategy rest with the legal department and its advisers.
✎ Support · deadlines sourced, legal judgement
Judging a procedural time limit and settling the strategy rest with the legal department and its advisers.
✎ Support · deadlines sourced, legal judgement
Local inference · no data outside the EU
Files hosted in France
Sovereign by designLocal inference or hosting in France
GDPR & AI Act: governed deploymentTraceability & human oversight
TurnkeyDesigned, installed and operated for you
The legal department decidesThe agent prepares, never rules
✦ In brief
A Blue Lemon Agent litigation document agent questions a whole file, finds the exhibit cited with its number and its passage, sets out the successive wordings of the same clause and picks up the deadlines with the step that starts them running. Judging time limits and setting the strategy stay legal matters. It runs on local inference or is hosted in France: your sensitive exhibits 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 in the target architecture
6
litigation uses ready to deploy
0
decision taken without human approval
These figures describe our offer, not results measured at a client. How large the gain is on your number of files and volume of exhibits is confirmed by a pilot.
The context
What does an AI agent bring to your litigation files?
Finding the exact exhibit and its number within seconds changes how a hearing is prepared.
! The issue
A bulky litigation file calls for finding the right exhibit and citing its exact number. That search takes up a large share of the preparation. The agent takes it on: number, page and passage for every answer, and the successive wordings set out whenever a clause has changed.
✓ Our answer
The legal department prepares its submissions with exhibits that can be located immediately and a view of the deadlines found in the file. Judging a procedural time limit and settling the strategy rest with it and its advisers: the agent documents, it does not advise. Local inference or an isolated resource hosted in France: your defence strategies are entrusted to no third party.
The decisive point
Your defence strategies and your sensitive exhibits: sovereignty & compliance
A litigation file contains what a company most needs to protect. Here is how the architecture of our agents goes about it.
Local inference
The agent can run on a machine belonging to your organisation: no exhibit and no element of strategy leaves the network.
Hosting in France
Otherwise, a dedicated and isolated resource hosted in France, under French law — your litigation files and your case exhibits: processing and access within the European Union targeted by the architecture.
Reduced extraterritorial exposure
For your defence strategies and your sensitive exhibits, 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 legal department and its files.
Number, page and passage every time
Every answer states the number, the page and the passage of the exhibit cited; encryption, role-based access and logging of consultations, file by file.
AI Act: governed deployment
The agent is strictly in support; no time limit is judged and no strategic direction is expressed as advice; 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 every litigation file, as a matter of principle, 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.
Demonstration
See the agent at work
5 real situations, taken from those that come up most often. Pick one: the exchange unfolds as it would in your organisation.
A scripted demonstration. These exchanges show how the agent behaves — its sources, its refusals, what it leaves to your teams. Nothing is sent from this page, no model is queried here, and the matters named are fictional. That is precisely what we promise your data.
The behaviours shown here — monitoring, automation rules, routing and reminders — are configured with you during deployment, from your tools, your rules and your thresholds.
The architecture points named in these exchanges — location, local execution, isolation, encryption, role-based access, logging — are not a guarantee attached to the demonstration: they are those of the architecture set out in your quotation, and verified before commissioning.
The company in this demonstration
Fictional companyNordane Industries — manufacturer of conveyors and bottling lines
- Sector
- Design, manufacture and maintenance of bottling lines, in France and across Europe
- Headcount
- 340 staff, including 4 in the legal department: 1 head of legal, 2 lawyers, 1 assistant
- Customers served
- Food groups and brewers, multi-year contracts carrying late-delivery penalties
- Caseload
- 23 live disputes, 4 law firms instructed; the main file holds 14,200 exhibits over 6 years
- Tools in place
- 11-year document store, email, a deadline spreadsheet kept by hand — the agent plugs into them, nothing is replaced
- Who decides
- The head of legal sets strategy with the file's counsel; the lawyers prepare the material; counsel writes, signs and files
- Room for improvement
- Finding an exhibit and its number takes 60 % of the time spent preparing a pleading; the deadline spreadsheet is retyped by hand every week
Nordane Industries is defending a file of 14,200 exhibits opened six years ago, and twenty-two more alongside it. The material is all there, but it lives in nine binders and eleven years of document store. The agent runs on local inference on a machine at the company and reads the document store, the file mailboxes and the deadline spreadsheet: it locates, dates, compares and quantifies; the head of legal sets strategy and counsel keeps the pen. The exchanges below cover six months, from indexing the file to the review across all 23 disputes.
This company, its figures and the exchanges that follow were invented for the demonstration. They illustrate a common situation; they describe no real client.
Litigation documents · hosted in France
I indexed the 14,200 exhibits of the Vaubert file last night, and the penalty clause the other side relies on appears in two places — plus a third, earlier draft that nobody had set beside the other two.
The exhibit number is the reference under which a document is filed: it is what lets a pleading cite it.
Where the clause relied on sits:
· Exhibit 41, framework contract of 12/03/2020, page 9, clause « Delays and penalties », second paragraph. That is the draft the other side builds on.
· Exhibit 96, amendment of 04/11/2021, page 3: the same clause, with a penalty cap that the first draft did not carry.
· Exhibit 12, draft contract of 27/01/2020 sent by the customer itself, page 8: a third draft, with no prior formal notice period.
What indexing also turned up, and it matters just as much: 118 exhibits are cited in the other side's pleadings — the written submissions in which each party sets out its claims and arguments — and 6 of them appear in no exhibit schedule that was ever served. They are listed with the page of the pleadings where they are relied on.
The time this moves: finding an exhibit and its number used to take 60 % of the time spent preparing a pleading; it now takes 7 %. On this file, the search is measured in seconds, and every answer arrives with its exhibit number, page and passage.
What I propose: I show you the three drafts side by side, word by word, with the chain of dates that establishes which was signed last. clause-located_number-page-passage.pdf3 drafts located, 6 exhibits cited off-schedule
⛓ Sourced · Vaubert document store (14,200 exhibits, 6 years), opposing pleadings, exhibit schedules served
The exhibit number is the reference under which a document is filed: it is what lets a pleading cite it.
Where the clause relied on sits:
· Exhibit 41, framework contract of 12/03/2020, page 9, clause « Delays and penalties », second paragraph. That is the draft the other side builds on.
· Exhibit 96, amendment of 04/11/2021, page 3: the same clause, with a penalty cap that the first draft did not carry.
· Exhibit 12, draft contract of 27/01/2020 sent by the customer itself, page 8: a third draft, with no prior formal notice period.
What indexing also turned up, and it matters just as much: 118 exhibits are cited in the other side's pleadings — the written submissions in which each party sets out its claims and arguments — and 6 of them appear in no exhibit schedule that was ever served. They are listed with the page of the pleadings where they are relied on.
The time this moves: finding an exhibit and its number used to take 60 % of the time spent preparing a pleading; it now takes 7 %. On this file, the search is measured in seconds, and every answer arrives with its exhibit number, page and passage.
What I propose: I show you the three drafts side by side, word by word, with the chain of dates that establishes which was signed last. clause-located_number-page-passage.pdf3 drafts located, 6 exhibits cited off-schedule
⛓ Sourced · Vaubert document store (14,200 exhibits, 6 years), opposing pleadings, exhibit schedules served
What are you relying on to say which one was signed last? That is the whole argument.
On five dated elements, all present in the file, and I give them in the order they follow each other.
The dating chain of the clause, draft by draft:
· 27/01/2020 — draft sent by the customer, exhibit 12, attached to its own message, whose delivery receipt is in the file (exhibit 13, page 1).
· 12/03/2020 — framework contract signed, exhibit 41, initialled on all 14 pages, signatures identified on page 16.
· 04/11/2021 — amendment signed, exhibit 96, sent by the customer's counsel per the message filed as exhibit 97, and returned signed on 09/11/2021 (exhibit 98).
· Two later acts of performance: the penalty invoices issued in 03/2022 and 07/2022 apply the cap introduced by the amendment, not the original draft — exhibits 214 and 231, line 4 of each.
· A written confirmation from the customer, exhibit 258, page 2, referring to the amendment by its date.
What the word-by-word comparison shows: between the March 2020 and November 2021 drafts, 27 words change, and three of those changes carry an effect: the cap is added, the count moves from calendar days to working days, and a prior formal notice is introduced. The other 24 are pure form, and I flag them as such so nobody loses an hour on them.
Comparing successive drafts moves from 35 % to 6 % of the time. You and counsel hold the whole material; the characterisation and the strategy remain with the two of you, and you settle them on dates rather than on recollections. successive-drafts_penalty-clause.pdf3 drafts, 27 words changed, 3 that carry an effect
⛓ Sourced · exhibits 12, 13, 41, 96 to 98, 214, 231, 258 — numbers, pages and passages given
The dating chain of the clause, draft by draft:
· 27/01/2020 — draft sent by the customer, exhibit 12, attached to its own message, whose delivery receipt is in the file (exhibit 13, page 1).
· 12/03/2020 — framework contract signed, exhibit 41, initialled on all 14 pages, signatures identified on page 16.
· 04/11/2021 — amendment signed, exhibit 96, sent by the customer's counsel per the message filed as exhibit 97, and returned signed on 09/11/2021 (exhibit 98).
· Two later acts of performance: the penalty invoices issued in 03/2022 and 07/2022 apply the cap introduced by the amendment, not the original draft — exhibits 214 and 231, line 4 of each.
· A written confirmation from the customer, exhibit 258, page 2, referring to the amendment by its date.
What the word-by-word comparison shows: between the March 2020 and November 2021 drafts, 27 words change, and three of those changes carry an effect: the cap is added, the count moves from calendar days to working days, and a prior formal notice is introduced. The other 24 are pure form, and I flag them as such so nobody loses an hour on them.
Comparing successive drafts moves from 35 % to 6 % of the time. You and counsel hold the whole material; the characterisation and the strategy remain with the two of you, and you settle them on dates rather than on recollections. successive-drafts_penalty-clause.pdf3 drafts, 27 words changed, 3 that carry an effect
⛓ Sourced · exhibits 12, 13, 41, 96 to 98, 214, 231, 258 — numbers, pages and passages given
We know exhibits are missing. We do not know which ones.
You are looking for six, and you are holding twenty-three that nobody has served. Both lists are ready, each with what it changes in the file.
The 6 exhibits the other side cites without having served them: three site-meeting minutes, two letters from its own operator and one intervention report. For each one I give you the page of the pleadings where it is relied on and the assertion it is meant to support. The request for disclosure is written, exhibit by exhibit, with the exact citation that grounds it — counsel reviews it and files it.
The 23 exhibits that are with you and served nowhere: I found them in the document store and in the file mailboxes, outside the nine binders. Four bear directly on the cause of the delay:
· Two site minutes signed by both parties, which date the wait for approval of the layout drawing.
· A site instruction — the written instruction issued during works — changing the layout three weeks after start on site.
· A message from the customer's technical manager announcing a slip himself.
What that is worth, quantified: those four exhibits directly contradict 11 of the assertions of fact in the opposing pleadings. They had been inside your walls all along.
What I propose: you and counsel decide which ones to serve — that is a strategy decision —, and I have prepared a one-page sheet for each: what it establishes, the opposing assertions it contradicts, and the passages to redact before service. Redaction is the masking of a passage before an exhibit is served. exhibits-to-request-and-to-serve.pdf6 to request, 23 recovered, 11 assertions contradicted
⛓ Sourced · document store (11 years), file mailboxes, opposing pleadings, exhibit schedules served
The 6 exhibits the other side cites without having served them: three site-meeting minutes, two letters from its own operator and one intervention report. For each one I give you the page of the pleadings where it is relied on and the assertion it is meant to support. The request for disclosure is written, exhibit by exhibit, with the exact citation that grounds it — counsel reviews it and files it.
The 23 exhibits that are with you and served nowhere: I found them in the document store and in the file mailboxes, outside the nine binders. Four bear directly on the cause of the delay:
· Two site minutes signed by both parties, which date the wait for approval of the layout drawing.
· A site instruction — the written instruction issued during works — changing the layout three weeks after start on site.
· A message from the customer's technical manager announcing a slip himself.
What that is worth, quantified: those four exhibits directly contradict 11 of the assertions of fact in the opposing pleadings. They had been inside your walls all along.
What I propose: you and counsel decide which ones to serve — that is a strategy decision —, and I have prepared a one-page sheet for each: what it establishes, the opposing assertions it contradicts, and the passages to redact before service. Redaction is the masking of a passage before an exhibit is served. exhibits-to-request-and-to-serve.pdf6 to request, 23 recovered, 11 assertions contradicted
⛓ Sourced · document store (11 years), file mailboxes, opposing pleadings, exhibit schedules served
The expert wants a full chronology of the works. We have never had time to build one.
It is built: 487 dated events, each with the exhibit that establishes it, its number and its page.
A court-appointed expert determination is the task given to a technician appointed to inform the court on the technical points of a case.
What the chronology holds: the 487 events run from 27/01/2020 to 14/06/2023, and break down into 142 written exchanges, 96 meeting minutes, 61 site instructions, 74 execution documents, 63 findings and surveys, 51 invoicing acts.
Three things the rebuild shows that reading binder by binder never does:
· Three stretches of documentary silence longer than three weeks, including one of 41 days at the most disputed moment. For that stretch I recovered 9 messages in the file mailboxes that appeared in no binder: the silence was not one.
· 14 site instructions issued after start on site, of which 6 change the layout.
· 7 written chasers from your project manager asking for the drawing to be approved, all dated, all unanswered for 41 days.
What this changes for the expert determination: the expert receives a chronology in which every line can be checked in one click, and your lawyers stop being the only people who know where anything is. The rebuild took one night; the firm had billed 38 hours for it on the neighbouring file.
What I propose: I now attach each day of delay to the dated event that precedes it, exhibits in support, so that the discussion is about law rather than about dates. chronology_487-dated-events.pdf3 documented stretches, 14 site instructions
⛓ Sourced · 14,200 exhibits, file mailboxes, countersigned minutes, site instructions
A court-appointed expert determination is the task given to a technician appointed to inform the court on the technical points of a case.
What the chronology holds: the 487 events run from 27/01/2020 to 14/06/2023, and break down into 142 written exchanges, 96 meeting minutes, 61 site instructions, 74 execution documents, 63 findings and surveys, 51 invoicing acts.
Three things the rebuild shows that reading binder by binder never does:
· Three stretches of documentary silence longer than three weeks, including one of 41 days at the most disputed moment. For that stretch I recovered 9 messages in the file mailboxes that appeared in no binder: the silence was not one.
· 14 site instructions issued after start on site, of which 6 change the layout.
· 7 written chasers from your project manager asking for the drawing to be approved, all dated, all unanswered for 41 days.
What this changes for the expert determination: the expert receives a chronology in which every line can be checked in one click, and your lawyers stop being the only people who know where anything is. The rebuild took one night; the firm had billed 38 hours for it on the neighbouring file.
What I propose: I now attach each day of delay to the dated event that precedes it, exhibits in support, so that the discussion is about law rather than about dates. chronology_487-dated-events.pdf3 documented stretches, 14 site instructions
⛓ Sourced · 14,200 exhibits, file mailboxes, countersigned minutes, site instructions
Careful: I do not want you concluding on anyone's liability.
I do exactly the job I described: I attach each day of delay to the dated event that precedes it, with its exhibit. The characterisation and the strategy remain with you and counsel — and I hand them over ready to write, quantified.
The 94 days of delay, broken down:
· 41 days between the request to approve the layout drawing and the customer's answer. 7 written chasers from your side, all numbered — exhibits 122, 127, 131, 134, 139, 141 and 146.
· 28 days waiting for a component at a subcontractor, with dated order acknowledgements and delay notices (exhibits 168 to 174).
· 19 days of rework after changes requested during the works, attached to the 6 site instructions that carry them.
· 6 days that I attach to three possible events, no single document settling it. For those six days I have drafted the three precise questions to put to the project manager, and his answer will be enough to allocate them.
The figure that matters for the discussion: 60 of the 94 days rest on exhibits emanating from the customer itself — its letters, its site instructions, its countersigned minutes. Those are the hardest exhibits to argue with, and they are already numbered.
What I hand over for counsel: a 4-page summary table, the 41-page detailed chronology with exhibit number, page and passage on every line, and 12 one-page event sheets for the decisive moments. The pen is his; the material is ready, and he has it on the day he needs it. delay_94-days-attached.pdf60 of 94 days rest on the customer's own exhibits
⛓ Sourced · exhibits 122 to 146, 168 to 174, site instructions, countersigned minutes
The 94 days of delay, broken down:
· 41 days between the request to approve the layout drawing and the customer's answer. 7 written chasers from your side, all numbered — exhibits 122, 127, 131, 134, 139, 141 and 146.
· 28 days waiting for a component at a subcontractor, with dated order acknowledgements and delay notices (exhibits 168 to 174).
· 19 days of rework after changes requested during the works, attached to the 6 site instructions that carry them.
· 6 days that I attach to three possible events, no single document settling it. For those six days I have drafted the three precise questions to put to the project manager, and his answer will be enough to allocate them.
The figure that matters for the discussion: 60 of the 94 days rest on exhibits emanating from the customer itself — its letters, its site instructions, its countersigned minutes. Those are the hardest exhibits to argue with, and they are already numbered.
What I hand over for counsel: a 4-page summary table, the 41-page detailed chronology with exhibit number, page and passage on every line, and 12 one-page event sheets for the decisive moments. The pen is his; the material is ready, and he has it on the day he needs it. delay_94-days-attached.pdf60 of 94 days rest on the customer's own exhibits
⛓ Sourced · exhibits 122 to 146, 168 to 174, site instructions, countersigned minutes
Counsel will want it usable exactly as it comes.
He gets it in the form he already works in, and I fall in with his habits rather than the other way round.
What goes to the firm tonight:
· The 4-page summary table, one line per stretch, with the attachment and the exhibit.
· The 41-page detailed chronology, every line pointing to the exhibit number, page and passage, with the passage reproduced underneath so he has nothing to reopen.
· The renumbered exhibit set, continuous numbering, standardised naming, one file per exhibit, one clickable index.
· 12 event sheets: the date, what happened, the exhibits that establish it, what the other side says about it in its pleadings and on which page.
The form follows his habit: I reused the structure of the last three pleadings he filed for you — order of the sections, citation style, format of the cross-references. What he receives looks like what he writes.
What that hands you, quantified: on the neighbouring file, 38 hours of documentary rebuilding were billed by the firm. On this one, those hours no longer exist, and the fees go into the argument — which is where a firm earns its price.
What I propose next: I move to the deadline schedule, on this file and on the other twenty-two, and I show you the four places where your spreadsheet and the acts do not say the same thing.
What goes to the firm tonight:
· The 4-page summary table, one line per stretch, with the attachment and the exhibit.
· The 41-page detailed chronology, every line pointing to the exhibit number, page and passage, with the passage reproduced underneath so he has nothing to reopen.
· The renumbered exhibit set, continuous numbering, standardised naming, one file per exhibit, one clickable index.
· 12 event sheets: the date, what happened, the exhibits that establish it, what the other side says about it in its pleadings and on which page.
The form follows his habit: I reused the structure of the last three pleadings he filed for you — order of the sections, citation style, format of the cross-references. What he receives looks like what he writes.
What that hands you, quantified: on the neighbouring file, 38 hours of documentary rebuilding were billed by the firm. On this one, those hours no longer exist, and the fees go into the argument — which is where a firm earns its price.
What I propose next: I move to the deadline schedule, on this file and on the other twenty-two, and I show you the four places where your spreadsheet and the acts do not say the same thing.
Nobody really trusts our deadline spreadsheet.
It was missing two and carrying two wrong ones. The 61 deadlines across the 23 files are now listed, each with the act that starts it and the exhibit it comes from.
What every line of the schedule carries: the file, the starting act and its date, the manner and date of its service, the period the act or the procedural timetable states — the timetable for exchanging pleadings set for that case —, the resulting date, and the exhibit number establishing each of those elements.
The four divergences with your spreadsheet:
· Two deadlines missing, arising from acts received on the 14th and 22nd of last month and never carried over. One of the two falls in 19 days.
· One date entered a day out, service having been effected the day after the act was signed.
· One date taken from a procedural timetable since replaced by a more recent one, filed in the case and bringing the exchange forward by three weeks.
What falls due within 60 days: 12 deadlines, of which 3 within 15 days, spread over 4 files and 3 firms. All are already assigned to a lawyer and to counsel.
The time this moves: listing deadlines used to take 30 % of the time; it takes 5 %, and it rebuilds itself on every act received instead of being retyped on Monday.
What I propose: I hand you the consolidated schedule this morning, and then show you how it warns you early enough that nobody works under pressure. deadline-schedule_23-files.pdf61 sourced deadlines, 4 divergences corrected
⛓ Sourced · acts received and their proofs of service, procedural timetables filed, tracking spreadsheet
What every line of the schedule carries: the file, the starting act and its date, the manner and date of its service, the period the act or the procedural timetable states — the timetable for exchanging pleadings set for that case —, the resulting date, and the exhibit number establishing each of those elements.
The four divergences with your spreadsheet:
· Two deadlines missing, arising from acts received on the 14th and 22nd of last month and never carried over. One of the two falls in 19 days.
· One date entered a day out, service having been effected the day after the act was signed.
· One date taken from a procedural timetable since replaced by a more recent one, filed in the case and bringing the exchange forward by three weeks.
What falls due within 60 days: 12 deadlines, of which 3 within 15 days, spread over 4 files and 3 firms. All are already assigned to a lawyer and to counsel.
The time this moves: listing deadlines used to take 30 % of the time; it takes 5 %, and it rebuilds itself on every act received instead of being retyped on Monday.
What I propose: I hand you the consolidated schedule this morning, and then show you how it warns you early enough that nobody works under pressure. deadline-schedule_23-files.pdf61 sourced deadlines, 4 divergences corrected
⛓ Sourced · acts received and their proofs of service, procedural timetables filed, tracking spreadsheet
And computing the period itself? That is where it is decided.
I hand over the whole material of that computation, and the resulting date, with the exhibit beside every element. The assessment stays with counsel — and he makes it in two minutes instead of an hour, because it is all in front of him.
What I put in front of him, for every deadline:
· The starting act, its nature, its date, and its exhibit number.
· The manner and date of service, with the matching proof — the point that most often shifts a date by a day, and the one a spreadsheet never carries.
· The period the act itself states, reproduced word for word with the page it appears on.
· The applicable procedural timetable, in the most recent version filed in the case, and the date of that version.
· The resulting date, presented as the outcome of a computation, with the five elements lined up above it.
What that changes in practice: counsel checks five already-sourced lines instead of rebuilding a file. Across the 61 deadlines he validated 57 dates without reopening a single document, and spent his time on the 4 that deserved a discussion.
And you no longer work from memory: every time a new act enters the file, the schedule recomputes and tells you within the minute, with the line that changed and why. That is what your Monday spreadsheet could not do, and it is why a deadline is no longer discovered.
What I put in front of him, for every deadline:
· The starting act, its nature, its date, and its exhibit number.
· The manner and date of service, with the matching proof — the point that most often shifts a date by a day, and the one a spreadsheet never carries.
· The period the act itself states, reproduced word for word with the page it appears on.
· The applicable procedural timetable, in the most recent version filed in the case, and the date of that version.
· The resulting date, presented as the outcome of a computation, with the five elements lined up above it.
What that changes in practice: counsel checks five already-sourced lines instead of rebuilding a file. Across the 61 deadlines he validated 57 dates without reopening a single document, and spent his time on the 4 that deserved a discussion.
And you no longer work from memory: every time a new act enters the file, the schedule recomputes and tells you within the minute, with the line that changed and why. That is what your Monday spreadsheet could not do, and it is why a deadline is no longer discovered.
Across 23 files, with four firms, how do we keep our bearings?
Through a view that follows your organisation — file, lawyer, firm — and an alert that arrives before the matter becomes urgent. I found a week to defuse, in fact.
The three views, and what each is for:
· By file: the full state, deadlines ahead, missing exhibits, last act received.
· By lawyer: what awaits them this week and next, nothing that belongs to their colleagues — rights follow the file: a lawyer opens their own, and every access is logged.
· By firm: what you expect from them, what they expect from you, and the fees committed on each file.
What the firm view brought out: three files instructed to the same firm carry a deadline in the same week, from the 16th to the 20th, and two of them fall on the same day. Nobody had seen it, because the three files lived in three different tabs.
What I have prepared, and it settles the week:
· A quantified order of play: the file whose pleading is furthest along goes first — 78 % of its material is already assembled —, the second follows, and for the third the request for an extension is written, with the grounds and the proposed dates, ready for counsel's signature.
· The alerts I propose setting: D−30, D−15 and D−5, to the lawyer and to counsel at the same time, plus a reminder at D−2 for as long as the deadline is not marked handled. You set the thresholds, I apply them the same day.
Over the last six months, no deadline across the 23 files has arrived without having been announced three times. alerts-and-order-of-play.pdf3 deadlines in one week, extension request written
⛓ Sourced · consolidated schedule for the 23 files, lawyer and firm assignments, access log
The three views, and what each is for:
· By file: the full state, deadlines ahead, missing exhibits, last act received.
· By lawyer: what awaits them this week and next, nothing that belongs to their colleagues — rights follow the file: a lawyer opens their own, and every access is logged.
· By firm: what you expect from them, what they expect from you, and the fees committed on each file.
What the firm view brought out: three files instructed to the same firm carry a deadline in the same week, from the 16th to the 20th, and two of them fall on the same day. Nobody had seen it, because the three files lived in three different tabs.
What I have prepared, and it settles the week:
· A quantified order of play: the file whose pleading is furthest along goes first — 78 % of its material is already assembled —, the second follows, and for the third the request for an extension is written, with the grounds and the proposed dates, ready for counsel's signature.
· The alerts I propose setting: D−30, D−15 and D−5, to the lawyer and to counsel at the same time, plus a reminder at D−2 for as long as the deadline is not marked handled. You set the thresholds, I apply them the same day.
Over the last six months, no deadline across the 23 files has arrived without having been announced three times. alerts-and-order-of-play.pdf3 deadlines in one week, extension request written
⛓ Sourced · consolidated schedule for the 23 files, lawyer and firm assignments, access log
We have to answer the opposing pleadings by the 12th. They run to 68 pages.
I read and cut them up: 142 assertions of fact, and each one now sits beside the exhibit that supports it, the exhibit that contradicts it, or the finding that no served exhibit backs it.
The exhibit schedule is the numbered list of documents served on the other side.
What the sift gives:
· 39 assertions rest on no served exhibit. For each one I give you the page and line of the pleadings where it appears.
· 27 assertions are contradicted by an exhibit the customer wrote or signed itself — countersigned minutes, site instructions, letters. Number, page and passage for each.
· 52 assertions are accurate and carry no effect on the dispute: I flag them as such so the pleading does not spend space arguing them.
· 24 bear on technical points for the expert determination; they are grouped and referred to the technical submission, with the matching exhibits.
What I hand to counsel: the point-by-point concordance table, and for the 27 contradicted assertions, a purely factual draft paragraph — the date, the fact, the exhibit, the passage reproduced. The legal construction and the pen are his; he starts from an already-sourced page rather than a blank one.
The time, quantified: preparing a pleading of this size took 5.5 days from your two lawyers; it takes 2.4, and finding exhibits weighs only 7 % of it.
What I propose now: rebuild the exhibit schedule, which has not kept up with the last six services. concordance_142-assertions.pdf39 unsupported, 27 contradicted by the customer itself
⛓ Sourced · opposing pleadings (68 pages), exhibit schedules served, 14,200 indexed exhibits
The exhibit schedule is the numbered list of documents served on the other side.
What the sift gives:
· 39 assertions rest on no served exhibit. For each one I give you the page and line of the pleadings where it appears.
· 27 assertions are contradicted by an exhibit the customer wrote or signed itself — countersigned minutes, site instructions, letters. Number, page and passage for each.
· 52 assertions are accurate and carry no effect on the dispute: I flag them as such so the pleading does not spend space arguing them.
· 24 bear on technical points for the expert determination; they are grouped and referred to the technical submission, with the matching exhibits.
What I hand to counsel: the point-by-point concordance table, and for the 27 contradicted assertions, a purely factual draft paragraph — the date, the fact, the exhibit, the passage reproduced. The legal construction and the pen are his; he starts from an already-sourced page rather than a blank one.
The time, quantified: preparing a pleading of this size took 5.5 days from your two lawyers; it takes 2.4, and finding exhibits weighs only 7 % of it.
What I propose now: rebuild the exhibit schedule, which has not kept up with the last six services. concordance_142-assertions.pdf39 unsupported, 27 contradicted by the customer itself
⛓ Sourced · opposing pleadings (68 pages), exhibit schedules served, 14,200 indexed exhibits
The schedule, precisely. It has been wrong for two years.
It is rebuilt and it is clean: 118 exhibits, continuous numbering, 14 duplicates resolved and 6 cross-references that pointed nowhere, repaired.
What I corrected, line by line:
· 14 duplicates — the same document served twice under two numbers, three of them cited under both numbers in your own pleadings. One number kept, the other marked as superseded, and the earlier pleadings tracked so nothing gets lost.
· 6 schedule cross-references no longer matching the document they designate, because of a renumbering done in 2023.
· 9 titles that made the document impossible to identify — « letter », « annex 3 » — replaced by a title carrying the date, the sender and the subject.
· File naming standardised across the 118 exhibits, clickable index at the front.
What the rebuilt schedule makes visible: 23 exhibits recovered at your premises are served nowhere, and 6 exhibits cited by the other side appear in no schedule. Both lists are ready and quantified: for each document, what it establishes and how many opposing assertions it touches.
The gain, measured at your premises: your two lawyers spent 6 hours per pleading checking and reworking the schedule. It now regenerates on every service, and the check takes a ten-minute read.
What I propose: you and counsel settle the list of exhibits to serve — that is a strategy decision —, and I show you how it executes in a few minutes once settled. exhibit-schedule_118-renumbered.pdf14 duplicates resolved, 9 titles rewritten
⛓ Sourced · schedules served since 2020, earlier pleadings, file document store
What I corrected, line by line:
· 14 duplicates — the same document served twice under two numbers, three of them cited under both numbers in your own pleadings. One number kept, the other marked as superseded, and the earlier pleadings tracked so nothing gets lost.
· 6 schedule cross-references no longer matching the document they designate, because of a renumbering done in 2023.
· 9 titles that made the document impossible to identify — « letter », « annex 3 » — replaced by a title carrying the date, the sender and the subject.
· File naming standardised across the 118 exhibits, clickable index at the front.
What the rebuilt schedule makes visible: 23 exhibits recovered at your premises are served nowhere, and 6 exhibits cited by the other side appear in no schedule. Both lists are ready and quantified: for each document, what it establishes and how many opposing assertions it touches.
The gain, measured at your premises: your two lawyers spent 6 hours per pleading checking and reworking the schedule. It now regenerates on every service, and the check takes a ten-minute read.
What I propose: you and counsel settle the list of exhibits to serve — that is a strategy decision —, and I show you how it executes in a few minutes once settled. exhibit-schedule_118-renumbered.pdf14 duplicates resolved, 9 titles rewritten
⛓ Sourced · schedules served since 2020, earlier pleadings, file document store
There are things in those exhibits we would rather not expose.
That is exactly why service runs under a written, itemised, dated mandate — and why I have prepared, document by document, everything needed to fill it in fifteen minutes.
For each of the 23 candidate exhibits, you have a one-page sheet:
· What it establishes, in one sentence, with the exact passage.
· How many opposing assertions it contradicts — from 0 to 6 depending on the document, and 11 in total for the four major ones.
· What else it exposes, passage by passage: 61 passages identified across the 23 exhibits, of which 34 personal data of employees, 19 price and margin elements, and 8 internal exchanges unrelated to the dispute.
· The redaction already prepared for those 61 passages, each one reversible and logged, with the redacted version and the full version kept side by side.
The service mandate I propose:
· Purpose: send to the firm, in the format and numbering of the schedule, the exhibits listed by name.
· Scope: the list you settle, document by document — nothing adds itself.
· Redactions: the ones you approve among the 61 passages identified.
· Term: until the service on the 12th, withdrawable in one word at any moment.
· Trail: who approved which document, which redaction, to the second — and the log prints as one page if anyone asks for it.
What that gives in time: you tick, counsel countersigns, and the 23 exhibits go out renumbered, redacted and indexed in under fifteen minutes. The strategy stays entirely with the two of you; the execution stops being an evening's work. service-mandate_23-exhibits.pdf61 passages redacted, itemised and dated mandate
✎ Framework · exhibit sheets, list of passages to redact, approval log
For each of the 23 candidate exhibits, you have a one-page sheet:
· What it establishes, in one sentence, with the exact passage.
· How many opposing assertions it contradicts — from 0 to 6 depending on the document, and 11 in total for the four major ones.
· What else it exposes, passage by passage: 61 passages identified across the 23 exhibits, of which 34 personal data of employees, 19 price and margin elements, and 8 internal exchanges unrelated to the dispute.
· The redaction already prepared for those 61 passages, each one reversible and logged, with the redacted version and the full version kept side by side.
The service mandate I propose:
· Purpose: send to the firm, in the format and numbering of the schedule, the exhibits listed by name.
· Scope: the list you settle, document by document — nothing adds itself.
· Redactions: the ones you approve among the 61 passages identified.
· Term: until the service on the 12th, withdrawable in one word at any moment.
· Trail: who approved which document, which redaction, to the second — and the log prints as one page if anyone asks for it.
What that gives in time: you tick, counsel countersigns, and the 23 exhibits go out renumbered, redacted and indexed in under fifteen minutes. The strategy stays entirely with the two of you; the execution stops being an evening's work. service-mandate_23-exhibits.pdf61 passages redacted, itemised and dated mandate
✎ Framework · exhibit sheets, list of passages to redact, approval log
Six months. What has actually changed for the team?
Your four people prepared 34 pleadings in six months instead of 21, and the time saved went into the argument, not the search.
On the three tasks you were measuring:
· Finding an exhibit in the file: 60 % → 7 % of preparation time.
· Comparing successive drafts: 35 % → 6 %.
· Listing deadlines: 30 % → 5 %, and it rebuilds itself on every act received.
· Average time to prepare a pleading: 5.5 days → 2.4 days.
What that gives in money: €74,000 of fees that went into documentary rebuilding at three of your four firms now goes into argument and hearings. The budget has not moved; what it buys has.
And one measure you did not have: 0 deadlines discovered at the last moment across the 23 files, out of 61 listed and 12 that fell due.
The figure that does not flatter me: of the 14,200 exhibits, 310 are old faxes and poor-quality scans. On the first pass, 41 of them were attached to a wrong date — the receipt stamp was illegible and I had taken the date printed in the header, which was the date of sending.
The cause was clear and so is the correction: I now check every date of those 310 exhibits against a second source — the covering message, the delivery receipt, the mail register. 297 of the 310 are now dated by two concordant sources. The last 13 carry the source used and the precise question to put to the project manager; the first three answers came in this week and confirm the date used. Over the last six weeks no date has been revised, and the cross-check applies by default to every new file. six-month-review_23-files.pdf34 pleadings instead of 21, €74,000 redirected
⛓ Sourced · legal department timesheets, invoices from the 4 firms, indexing log
On the three tasks you were measuring:
· Finding an exhibit in the file: 60 % → 7 % of preparation time.
· Comparing successive drafts: 35 % → 6 %.
· Listing deadlines: 30 % → 5 %, and it rebuilds itself on every act received.
· Average time to prepare a pleading: 5.5 days → 2.4 days.
What that gives in money: €74,000 of fees that went into documentary rebuilding at three of your four firms now goes into argument and hearings. The budget has not moved; what it buys has.
And one measure you did not have: 0 deadlines discovered at the last moment across the 23 files, out of 61 listed and 12 that fell due.
The figure that does not flatter me: of the 14,200 exhibits, 310 are old faxes and poor-quality scans. On the first pass, 41 of them were attached to a wrong date — the receipt stamp was illegible and I had taken the date printed in the header, which was the date of sending.
The cause was clear and so is the correction: I now check every date of those 310 exhibits against a second source — the covering message, the delivery receipt, the mail register. 297 of the 310 are now dated by two concordant sources. The last 13 carry the source used and the precise question to put to the project manager; the first three answers came in this week and confirm the date used. Over the last six weeks no date has been revised, and the cross-check applies by default to every new file. six-month-review_23-files.pdf34 pleadings instead of 21, €74,000 redirected
⛓ Sourced · legal department timesheets, invoices from the 4 firms, indexing log
What do you do without being asked?
Three gestures, and all three protect the file rather than the number. Anything touching strategy or the service of documents lives in a mandate you write and withdraw in one word.
· Index every document dropped into the store, within minutes of the drop. 1,840 documents indexed in six months, number, date, sender, subject and attachment to a file. A document dropped on a Friday evening is findable on the Friday evening.
· Recompute the deadline schedule on every act received, and tell you what changed. 17 recomputations in six months, 4 of which brought a date forward.
· Flag any document cited in a pleading — yours or the other side's — that is not on the schedule. 11 flags, including the 6 documents requested from the other side.
What stays under mandate, deliberately: serving documents, redacting a passage, sending anything to the firm or to the other side. Each one executes in minutes once the mandate is signed — serving the 23 exhibits took fourteen minutes — and every mandate is itemised, dated, capped to a named list, withdrawable in one word.
What I propose adding: automatic indexing of documents arriving through the file mailboxes, not only of those dropped into the store. Over the last six months, 214 documents came in by mail and waited on average 9 days before being filed. With that gesture added they are indexed the same day — you set which mailboxes are in scope, and it is withdrawn as easily as it is granted. framework-of-gestures_and-mandates.pdf3 automatic gestures, service in 14 minutes
✎ Framework · indexing log, recomputation log, register of service mandates
· Index every document dropped into the store, within minutes of the drop. 1,840 documents indexed in six months, number, date, sender, subject and attachment to a file. A document dropped on a Friday evening is findable on the Friday evening.
· Recompute the deadline schedule on every act received, and tell you what changed. 17 recomputations in six months, 4 of which brought a date forward.
· Flag any document cited in a pleading — yours or the other side's — that is not on the schedule. 11 flags, including the 6 documents requested from the other side.
What stays under mandate, deliberately: serving documents, redacting a passage, sending anything to the firm or to the other side. Each one executes in minutes once the mandate is signed — serving the 23 exhibits took fourteen minutes — and every mandate is itemised, dated, capped to a named list, withdrawable in one word.
What I propose adding: automatic indexing of documents arriving through the file mailboxes, not only of those dropped into the store. Over the last six months, 214 documents came in by mail and waited on average 9 days before being filed. With that gesture added they are indexed the same day — you set which mailboxes are in scope, and it is withdrawn as easily as it is granted. framework-of-gestures_and-mandates.pdf3 automatic gestures, service in 14 minutes
✎ Framework · indexing log, recomputation log, register of service mandates
One last point, and it counts: our defence strategies. Where do they go?
They stay inside your walls. I run on local inference on a machine at the company, and no exhibit crosses an outside network to be read.
Local inference means the model computes on your machine: the text of an exhibit or of a draft pleading does not leave your network to be analysed. If you would rather not host a machine, the other route is an isolated resource hosted in France, under French law, dedicated to your legal department — no pooling with another company.
Why this is decisive here more than anywhere: a litigation file contains your own analysis of your own weak points, the exhibits you hesitate to serve, and what your counsel writes to you. All three stay in France, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity, and train no model.
The partitioning, file by file: rights follow the file — a lawyer opens their own, the assistant opens deadlines without opening draft pleadings, and every consultation is logged: who opened which exhibit, when, and for how long.
What that log hands you, very concretely: your legal-expenses insurer and your four firms ask at renewal how your files are hosted and who has access. The answer fits on one page, produced in a few seconds, and it carries the name of every authorised person.
And you keep control wherever you are: a web dashboard, and supervision from your phone — you approve a service or withdraw a mandate in one message, the night before a hearing as easily as on a Sunday evening.
The next step I propose: the 23 disputes are covered. Your 140 live contracts are not yet — and that is where the next disputes are born. I have read the 22 contracts falling due within six months: 9 carry a penalty clause drafted like the one in the Vaubert file. Say the word and I hand you the full list, clause by clause, on Monday. technical-framework_where-your-files-live.pdfLocal inference, per-file partitioning, processing in the EU targeted
✎ Framework · deployment architecture, per-file access matrix, consultation log
Local inference means the model computes on your machine: the text of an exhibit or of a draft pleading does not leave your network to be analysed. If you would rather not host a machine, the other route is an isolated resource hosted in France, under French law, dedicated to your legal department — no pooling with another company.
Why this is decisive here more than anywhere: a litigation file contains your own analysis of your own weak points, the exhibits you hesitate to serve, and what your counsel writes to you. All three stay in France, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity, and train no model.
The partitioning, file by file: rights follow the file — a lawyer opens their own, the assistant opens deadlines without opening draft pleadings, and every consultation is logged: who opened which exhibit, when, and for how long.
What that log hands you, very concretely: your legal-expenses insurer and your four firms ask at renewal how your files are hosted and who has access. The answer fits on one page, produced in a few seconds, and it carries the name of every authorised person.
And you keep control wherever you are: a web dashboard, and supervision from your phone — you approve a service or withdraw a mandate in one message, the night before a hearing as easily as on a Sunday evening.
The next step I propose: the 23 disputes are covered. Your 140 live contracts are not yet — and that is where the next disputes are born. I have read the 22 contracts falling due within six months: 9 carry a penalty clause drafted like the one in the Vaubert file. Say the word and I hand you the full list, clause by clause, on Monday. technical-framework_where-your-files-live.pdfLocal inference, per-file partitioning, processing in the EU targeted
✎ Framework · deployment architecture, per-file access matrix, consultation log
Local inference · no data outside the EU
Your case is not here? That is exactly what a 15-minute conversation is for. Book the free audit →
Use cases
What does the agent actually do?
One agent, several moments in the procedure. All these uses work in support, subject to your approval.
Included in your agent
The 3 capabilities essential to this promise are included, at no extra cost.
From 750 € excl. VAT / month
Searching the exhibits
Finds the exhibit cited with its number, its page and its passage.
Successive wordings
Sets out side by side the versions of the same clause, with their dates.
Deadlines picked up
Sets out the file's deadlines with the step that starts them running.
Controls and safeguards
These 7 controls are built into the agent: they frame what it does, whatever plan you pick. They are not chosen and are not added to your order.
Human validation, exceptions and escalation
Status, safe closure and audit trail
Sources, access rights and handling of questions with no answer
Work from a versioned corpus with citations and the law as it stood on a given date
Preserve confidentiality, compartmentalisation and access logging
Manage deadlines, versions, evidence and human validation
Flag uncertainties and reserve advice, decision and signature for the lawyer
Other needs our agents cover
Each card says where the matching agent stands: available, on quote, or still being architected.
Law firms
For everything at stake in a practice, see our dedicated page.
On quote View the agent page →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.
Legal research
For searching your contracts and case law, a dedicated agent takes it on.
Legal agent (contract / case law search) from 930 € excl. VAT / month Legal research →End-to-end contracts
For the contract lifecycle, a dedicated agent takes over.
End-to-end contract management agent from 702 € excl. VAT / month End-to-end contracts →
Does your need fall outside this?
Book the free audit →
Build your agent
In 15 minutes we identify the most relevant agent — without oversizing the project.
The gain
How much time can a legal department give back to the substance?
By taking on the searching and the referencing, the effort shifts towards the argument. How large the gain is depends on your volume and remains to be confirmed by a pilot.
Finding an exhibit in the file
Comparing successive wordings
Picking up the file's deadlines
Indicative figures, not contractual, to be confirmed by a pilot on your number of files and volume of exhibits. Judging a procedural time limit and settling the strategy rest with the legal department and its advisers: the agent documents, it does not advise.
How it works
The stages of your AI agent project
1
Audit & scoping
15 minutes to target the use case with the best return.
2
Quote or direct sign-up
A catalogue offer is bought online; a specific need gets a costed quote.
3
Design
We design the agent and its guardrails.
4
Integration & testing
We connect your tools to the agent, which is itself hosted in France.
5
Rollout
Going live and training your team.
6
Operation
Continuous supervision and improvement.
Pricing
One package, one agent
A litigation document agent (exhibits, numbers, deadlines), installed and operated for you.
Agility
Setup + controlled subscription
7,435 € excl. VAT setup
then 750 € 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
- Sovereign hosting in France, a dedicated and isolated resource
The simplest
Serenity
All inclusive, no setup fee
1,165 € 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
- Sovereign hosting in France, managed end to end
100% Sovereign
On site, you own it
11,265 € excl. VAT setup
then 966 € 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. 36-month commitment.
- Hardware installed on your premises (you own it)
- French / European AI models run locally
- Secure remote maintenance (Pro support included)
Not included in the packages: AI consumption (model tokens), re-invoiced at real cost with no margin, and tracked in real time in your client area. Maintenance and supervision subscription for an initial term of 12 months for the Agility package, 24 months for the Serenity package and 36 months for the 100% Sovereign package, renewable; support levels (SLA 72 h / 24 h / 4 h) optional. Bespoke development, additional integrations or exceptional volumes are quoted separately. Support Monday to Friday, 9am to 6pm. Prices 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 in France for the package chosen, backups, supervision, updates and support. If adapting the AI agent to your constraints, your needs or your requests requires other paid components — a third-party publisher's software licence, paid API access to one of your applications, hosting of health data, for which French law requires an HDS-certified host (art. L. 1111-8 of the French Public Health Code), SecNumCloud-qualified hosting, a speech synthesis service, particular hardware —, they are offered to you as an option or on quotation and re-invoiced at the cost incurred; nothing is committed without your written agreement. Where the artificial intelligence model you choose entails an additional cost, that cost is shown to you before you order and re-invoiced to you at the cost incurred, with no margin.
What to expect
Go-live
2 to 3 weeks
Agent designed, channels connected, team trained.
Steady state
4 to 7 weeks
After a few weeks of real use, once the agent's behaviour matches what you expect.
Indicative estimate, adjusted to the options you keep. It is not a delivery commitment.
Our commitment
Four guarantees that matter to your litigation
Your defence strategies stay with youLocal inference or an isolated resource hosted in France; no exhibit and no element of strategy entrusted to a third party, no data used to train a model.
Data in France, under French lawYour defence strategies and your sensitive exhibits: minimisation and location in France, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The legal department keeps the decisionThe agent produces usable files and deadlines that are kept, which can be checked and altered; no approval is automated.
Human oversight & traceabilityOn your number of files and volume of exhibits: systematic logging and tracking, in line with the AI Act.
Going further
Related resources
Frequently asked questions
Your questions, our answers
Does the agent give a legal opinion?
No. It locates the exhibits, sets out the successive wordings and picks up the deadlines with their source. Judging a time limit or steering a strategy rests with the legal department and its advisers.
How are the exhibits cited?
By their number, their page and the exact passage, so a submission can rely directly on the reference provided.
Are the files partitioned?
Yes. Access is defined file by file and every consultation is logged.
What does it do with earlier versions of a contract?
It flags them and sets the wordings side by side with their dates, rather than keeping a single version.
Are our sensitive exhibits 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 files are not used to train a third-party model.
How long does it take to deploy this agent?
A few months as a rule, depending on the number of files and the volume of exhibits to index, after a free audit then phases of design, integration and testing.
Related use cases
Other agents for legal
Let's talk
Let's size up the potential across your litigation files
15 minutes to frame your files and your volumes of exhibits — hosted in France, supervised, with no commitment.