● B2B offer — International support
International support: several languages, one source of truth
Serving several markets means saying the same thing everywhere, in each customer's language, on the channel they use. Your agent draws on a single source — your reference content — and answers in each language chosen, respecting the local particulars you have documented: lead times, warranties, information duties. Hosted in France: your customer exchanges stay with you. Your local teams keep control of what commits them.
Hosted in France
International exchanges protected
GDPR & AI Act: governed deployment
Human oversight
Updated on
Deployed in a few weeks
International support · hosted in France
A German customer is asking what warranty period applies where they are.
Answer written in German from your reference content, with the local particular documented for that market.
The source cited is the same as for the French version: one truth, translated.
The mandatory information notices for that market are carried over.
🔗 Sourced · reference content and local particulars
The source cited is the same as for the French version: one truth, translated.
The mandatory information notices for that market are carried over.
🔗 Sourced · reference content and local particulars
And if a market has no documented particular?
The answer is limited to the shared reference content, and the question is passed to the local team.
Anything touching a country's legal obligation commits the company in that market: it is documented before it is said.
✎ Support · passed to the local team
Anything touching a country's legal obligation commits the company in that market: it is documented before it is said.
✎ Support · passed to the local team
Local inference · no data outside the EU
Support 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 customer service department decidesThe agent prepares, never rules
✦ In brief
A Blue Lemon Agent international support agent answers in several languages and across several channels from a single source, applying the local particulars you have documented — lead times, warranties, mandatory notices. Anything not documented is passed to the local team. It runs on local inference or is hosted in France: your exchanges 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
support 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 languages, channels and markets is confirmed by a pilot.
The context
What does an AI agent bring to your international support?
An identical answer from one market to the next, adapted to the language and to local law: that is the mark of an organisation with its support under control.
! The issue
International support has to hold together consistent information and genuine local particulars. Maintaining one version per country multiplies the gaps; drawing on a single source, extended by documented particulars, avoids them. That is the architecture the agent applies, language by language and channel by channel.
✓ Our answer
Your customers get an immediate answer in their language, in line with what applies where they are. The local teams receive whatever is not documented and keep control of what commits the company in their country. Local inference or an isolated resource hosted in France: your customer exchanges around the world are entrusted to no third party, and transfers outside the European Union are ruled out by design.
The decisive point
Your exchanges with customers around the world: sovereignty & compliance
International support brings together personal data and the legal obligations of several countries. Here is how the architecture of our agents answers that.
Local inference
The agent can run on a machine belonging to your organisation: no customer exchange and no support content leaves the network.
Hosting in France
Otherwise, a dedicated and isolated resource hosted in France, under French law — your support content and your international channels: processing and access within the European Union targeted by the architecture.
Reduced extraterritorial exposure
For your exchanges with customers around the world, 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 markets.
One source, documented particulars
Every answer cites the reference content and, where relevant, the applicable local particular; encryption, role-based access and logging by market.
AI Act: governed deployment
The agent is strictly in support; no country's legal obligation is asserted without prior documentation; 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 international disputes and exchanges concerning a product recall, 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 companyAldevine — manufacturer of domestic air treatment appliances
- Sector
- Manufacture and sale of air treatment appliances — purifiers, dehumidifiers
- Headcount
- 340 staff, including an international support team of 16 across 3 sites
- Markets served
- 9 European markets, 9 languages — 210,000 appliances sold a year, an installed base of 840,000
- Volume
- 7,400 enquiries a month — 4,100 emails, 2,100 chat conversations, 1,200 calls; chat is open in 6 markets, phone in 4
- Tools in place
- Product knowledge base, ticketing tool, website chat, telephony, CRM and a shared workspace of 9 country folders — the agent plugs into them, nothing is replaced
- Who decides
- Each market manager validates the local specifics of their country; the customer service director sets the channel scope
- Room for improvement
- The 9 country versions of the knowledge base have diverged and keeping them aligned takes 60 % of a support writer's time; 3,500 enquiries a month wait for a translation or a local team; first-reply time ranges from 40 minutes in France to 3 days in Sweden and Poland
Aldevine opened its markets one at a time, and each country ended up holding its own version of the knowledge base. The agent runs on local inference on a machine at head office and plugs into the product base, the ticketing tool, the chat, the telephony and the CRM: it works from a single source, applies the specifics each market manager has validated, and works up whatever is not yet validated. The exchanges below cover one quarter, from comparing the 9 bases to preparing a tenth market.
This company, its figures and the exchanges that follow were invented for the demonstration. They illustrate a common situation; they describe no real client.
International support · hosted in France
I compared your nine country bases article by article overnight — 1,240 pages — and found 148 substantive gaps between versions of the same content. Twelve of them tell the customer different things depending on the language they read.
A substantive gap is not a difference of style: it is a piece of information whose meaning changes from one version to the next — a figure, a period, a condition, a procedure.
The clearest of the twelve: for the same purifier, your German version states 24 months of commercial warranty, your Dutch version states 12, and your Swedish version 36. Your contractual undertaking is 24 months everywhere, and 36 months in Sweden where your distributor extended it. The Dutch version is two rewrites behind — it dates from 2023.
What the divergence was costing, measured on your own tickets: keeping the languages consistent took 60 % of a support writer's time; it now takes 7 %. The rest went into working out which version was authoritative.
What I propose, and it is the heart of the arrangement: 312 reference articles, maintained once, of which each language is a view — not a copy. 2,808 pages kept current by 312 edits.
What is left for you to do: your nine market managers read through the list of 148 gaps, which I have already sorted and worked up. Let me show you the sort. single-source_312-articles-for-9-languages.pdf2,808 pages kept current by 312 edits
⛓ Sourced · 9 country bases (1,240 pages), distribution contracts, rewrite history
A substantive gap is not a difference of style: it is a piece of information whose meaning changes from one version to the next — a figure, a period, a condition, a procedure.
The clearest of the twelve: for the same purifier, your German version states 24 months of commercial warranty, your Dutch version states 12, and your Swedish version 36. Your contractual undertaking is 24 months everywhere, and 36 months in Sweden where your distributor extended it. The Dutch version is two rewrites behind — it dates from 2023.
What the divergence was costing, measured on your own tickets: keeping the languages consistent took 60 % of a support writer's time; it now takes 7 %. The rest went into working out which version was authoritative.
What I propose, and it is the heart of the arrangement: 312 reference articles, maintained once, of which each language is a view — not a copy. 2,808 pages kept current by 312 edits.
What is left for you to do: your nine market managers read through the list of 148 gaps, which I have already sorted and worked up. Let me show you the sort. single-source_312-articles-for-9-languages.pdf2,808 pages kept current by 312 edits
⛓ Sourced · 9 country bases (1,240 pages), distribution contracts, rewrite history
How do you know which of the nine versions is right?
I do not settle it alone, and that is precisely what makes the sort usable: every gap arrives with its origin, its date, its author and the document that ought to be authoritative. Here are the three families.
· 96 stale-translation gaps — the reference article changed, the country version stayed put. I have aligned them on the source, each reversible in one click, and I show the before and after side by side. None of them touches an undertaking: they are turns of phrase, screenshots and model numbers.
· 40 gaps that are in fact genuine local specifics — a return procedure particular to one market, a different service contact. They are not errors: I have taken them out of the common base and placed them in the register of specifics, for the relevant market manager to validate.
· 12 contradictions that commit the company — warranty, return period, transport costs. I have not aligned those: I have put them as questions, with the distribution contract clause or the product sheet that ought to settle them. Each question fits in three lines and is answered in two minutes.
What that produced: all 12 were settled in six days by the nine market managers, including the three warranty versions. The Dutch version moved to 24 months on the 9th, and the 1,400 Dutch customers who had received the old answer were contacted one by one — with a message telling them the correct period, not a shorter one.
What I propose now: that the sort runs continuously on new edits. A substantive gap can no longer live for nine months: it is flagged the day it appears, to the manager of the market concerned. 148-gaps-sorted_and-how-they-get-settled.pdf96 aligned, 40 became specifics, 12 settled in 6 days
⛓ Sourced · 9 country bases, distribution contracts, product sheets, rewrite log
· 96 stale-translation gaps — the reference article changed, the country version stayed put. I have aligned them on the source, each reversible in one click, and I show the before and after side by side. None of them touches an undertaking: they are turns of phrase, screenshots and model numbers.
· 40 gaps that are in fact genuine local specifics — a return procedure particular to one market, a different service contact. They are not errors: I have taken them out of the common base and placed them in the register of specifics, for the relevant market manager to validate.
· 12 contradictions that commit the company — warranty, return period, transport costs. I have not aligned those: I have put them as questions, with the distribution contract clause or the product sheet that ought to settle them. Each question fits in three lines and is answered in two minutes.
What that produced: all 12 were settled in six days by the nine market managers, including the three warranty versions. The Dutch version moved to 24 months on the 9th, and the 1,400 Dutch customers who had received the old answer were contacted one by one — with a message telling them the correct period, not a shorter one.
What I propose now: that the sort runs continuously on new edits. A substantive gap can no longer live for nine months: it is flagged the day it appears, to the manager of the market concerned. 148-gaps-sorted_and-how-they-get-settled.pdf96 aligned, 40 became specifics, 12 settled in 6 days
⛓ Sourced · 9 country bases, distribution contracts, product sheets, rewrite log
And when a reference article changes, what happens across the nine languages?
All nine versions are current within 40 minutes, local specifics reapplied, and every market manager receives what changed on their own market. You used to take 11 days, and two languages were missed on average.
What happens exactly, in order:
· The reference article is rewritten once, in French, by your support writer.
· The nine views are regenerated, and the market's specifics are put back in their place — the Swedish warranty stays at 36 months, the Polish return procedure stays the one in the register.
· Every market manager receives a before-and-after of what changes on their market, and nothing else. They have 48 hours to object, and they can freeze publication on their market in a word.
· The month's 41 updates went out that way. No language missed, and the log shows it market by market.
What that changes for your customers, and it can be measured: the manual for the quiet range changed on the 12th following a product update. The support answer was correct in all nine languages that same evening. The previous rewrite, in 2025, left the Portuguese version behind for seven weeks, and 340 Portuguese customers received an out-of-date maintenance instruction.
What I suggest next: extend the same mechanism to your manuals and packaging. Those are the two places where a country version ages without anyone noticing — and your document translation agent handles it, from the same single source. propagation_11-days-become-40-minutes.pdf41 updates, 9 languages, 0 missed
⛓ Sourced · publication log, register of specifics, history of the 2025 rewrite
What happens exactly, in order:
· The reference article is rewritten once, in French, by your support writer.
· The nine views are regenerated, and the market's specifics are put back in their place — the Swedish warranty stays at 36 months, the Polish return procedure stays the one in the register.
· Every market manager receives a before-and-after of what changes on their market, and nothing else. They have 48 hours to object, and they can freeze publication on their market in a word.
· The month's 41 updates went out that way. No language missed, and the log shows it market by market.
What that changes for your customers, and it can be measured: the manual for the quiet range changed on the 12th following a product update. The support answer was correct in all nine languages that same evening. The previous rewrite, in 2025, left the Portuguese version behind for seven weeks, and 340 Portuguese customers received an out-of-date maintenance instruction.
What I suggest next: extend the same mechanism to your manuals and packaging. Those are the two places where a country version ages without anyone noticing — and your document translation agent handles it, from the same single source. propagation_11-days-become-40-minutes.pdf41 updates, 9 languages, 0 missed
⛓ Sourced · publication log, register of specifics, history of the 2025 rewrite
Country specifics are what worries us. We do not want to say the wrong thing about a warranty or a right of return.
That is why they live in a register and not in someone's head. 94 documented specifics across your nine markets, each with its source, its date and the market manager who validated it — and every answer to a customer cites them.
A local specific here is a rule particular to one market that your company has written and validated: it is added to the common reference article rather than replacing it.
What the register holds today:
· 23 warranty and extension periods, including the three extensions negotiated by your distributors.
· 17 withdrawal and return periods, with each market's exact procedure.
· 21 pieces of information to be included in the answer depending on the market.
· 14 appliance take-back and recycling obligations, with the collection point to state.
· 19 service and consumer mediation contacts, specific to each country.
What each line carries, and this is what makes it hold up: the wording in French, the validated translation, the source — contract clause, official market page, internal note —, the validation date and the market manager's name. An answer to a customer cites both: the reference article and the applicable specific.
What that shifts: checking local specifics took 35 % of your support teams' time; it now takes 6 %. And the remaining 6 % is reading over a new specific, which is exactly the work your market managers should be doing.
What I suggest: that every specific carries a review date, chosen by whoever validated it. Three of the 94 are more than 18 months old; I have raised them with the precise point to re-check, and two have already been confirmed. register-of-specifics_94-dated-entries.pdf5 families, source and validator for each
⛓ Sourced · distribution contracts, internal notes from the 9 markets, register of specifics
A local specific here is a rule particular to one market that your company has written and validated: it is added to the common reference article rather than replacing it.
What the register holds today:
· 23 warranty and extension periods, including the three extensions negotiated by your distributors.
· 17 withdrawal and return periods, with each market's exact procedure.
· 21 pieces of information to be included in the answer depending on the market.
· 14 appliance take-back and recycling obligations, with the collection point to state.
· 19 service and consumer mediation contacts, specific to each country.
What each line carries, and this is what makes it hold up: the wording in French, the validated translation, the source — contract clause, official market page, internal note —, the validation date and the market manager's name. An answer to a customer cites both: the reference article and the applicable specific.
What that shifts: checking local specifics took 35 % of your support teams' time; it now takes 6 %. And the remaining 6 % is reading over a new specific, which is exactly the work your market managers should be doing.
What I suggest: that every specific carries a review date, chosen by whoever validated it. Three of the 94 are more than 18 months old; I have raised them with the precise point to re-check, and two have already been confirmed. register-of-specifics_94-dated-entries.pdf5 families, source and validator for each
⛓ Sourced · distribution contracts, internal notes from the 9 markets, register of specifics
And when a question lands on something that is not in the register?
The customer gets an answer within the minute on whatever is common and sourced, plus a dated commitment; and the market manager receives the question already worked up. They answer in 40 minutes instead of three days, and their answer enters the register.
What the customer receives straight away: the reference article that applies everywhere, what is settled without argument, and the exact sentence: “the point specific to your country will be confirmed by our local team before noon tomorrow”. A date that is kept beats an approximate answer, and your customers notice: 4.6 out of 5 on those exchanges.
What the market manager receives, and this is what makes the 40 minutes: the question restated in one sentence · the exact passage of the reference article concerned · what the other eight markets have documented on the same point · the source to check, which I name for them · and the answer already drafted in their language, which they only have to confirm or correct.
What that produced over the quarter: 3,500 enquiries a month were waiting for a translation or a local team; 610 remain. Of those 610, 214 produced a new specific: the register went from 94 to 308 entries, and the residual volume fell to 180 a month.
The reading, and it holds for what comes next: customers in one market ask the same questions as those in the other eight. Of the 214 new specifics, 168 were already documented elsewhere — only the local confirmation was missing, and that takes two minutes when the question is worked up.
What I suggest: a monthly thirty-minute review per market, on new specifics and those reaching their review date. It is the only standing meeting this arrangement asks of your market managers. undocumented-question_the-40-minute-protocol.pdf3,500 → 610 → 180 a month, register 94 → 308
⛓ Sourced · queue of pending enquiries, register of specifics, response times across the 9 markets
What the customer receives straight away: the reference article that applies everywhere, what is settled without argument, and the exact sentence: “the point specific to your country will be confirmed by our local team before noon tomorrow”. A date that is kept beats an approximate answer, and your customers notice: 4.6 out of 5 on those exchanges.
What the market manager receives, and this is what makes the 40 minutes: the question restated in one sentence · the exact passage of the reference article concerned · what the other eight markets have documented on the same point · the source to check, which I name for them · and the answer already drafted in their language, which they only have to confirm or correct.
What that produced over the quarter: 3,500 enquiries a month were waiting for a translation or a local team; 610 remain. Of those 610, 214 produced a new specific: the register went from 94 to 308 entries, and the residual volume fell to 180 a month.
The reading, and it holds for what comes next: customers in one market ask the same questions as those in the other eight. Of the 214 new specifics, 168 were already documented elsewhere — only the local confirmation was missing, and that takes two minutes when the question is worked up.
What I suggest: a monthly thirty-minute review per market, on new specifics and those reaching their review date. It is the only standing meeting this arrangement asks of your market managers. undocumented-question_the-40-minute-protocol.pdf3,500 → 610 → 180 a month, register 94 → 308
⛓ Sourced · queue of pending enquiries, register of specifics, response times across the 9 markets
It is still our responsibility. What if a rule changes in a country and we do not see it?
I watch the sources your market managers designated themselves, and I warn them the day those sources move, with the text before and after side by side. Four alerts this quarter, and all four were handled within the week.
What I watch, market by market: the official pages the manager entered in the register · the clauses of your distribution contracts and their amendments · your internal service notes · and the product sheets a specific depends on.
What I hand you on an alert: the source, the date of the change, the exact passage modified, the list of register entries that rest on it, and the number of answers sent last month on that point. Reading it belongs to your market manager; I hand them the file done.
The four this quarter: two distribution amendments — including the Swedish warranty extension, raised from 36 to 48 months — and two changes to official pages on return procedures. On the Swedish 48 months, the specific was current in the register 3 hours after the amendment, and Swedish answers were stating it that same evening.
And what protects the arrangement in the meantime: any answer resting on a specific under review says so to the customer and offers a call back from the local team. That happened 11 times this quarter; all 11 customers had their confirmation the same day.
What I suggest: that each market manager designate two further sources they already check by hand. They gain the watching, and the register gains a review date that keeps itself. source-watching_4-alerts-this-quarter.pdfBefore and after side by side, dependent specifics listed
✎ Framework · sources designated by the 9 market managers, distribution amendments, alert log
What I watch, market by market: the official pages the manager entered in the register · the clauses of your distribution contracts and their amendments · your internal service notes · and the product sheets a specific depends on.
What I hand you on an alert: the source, the date of the change, the exact passage modified, the list of register entries that rest on it, and the number of answers sent last month on that point. Reading it belongs to your market manager; I hand them the file done.
The four this quarter: two distribution amendments — including the Swedish warranty extension, raised from 36 to 48 months — and two changes to official pages on return procedures. On the Swedish 48 months, the specific was current in the register 3 hours after the amendment, and Swedish answers were stating it that same evening.
And what protects the arrangement in the meantime: any answer resting on a specific under review says so to the customer and offers a call back from the local team. That happened 11 times this quarter; all 11 customers had their confirmation the same day.
What I suggest: that each market manager designate two further sources they already check by hand. They gain the watching, and the register gains a review date that keeps itself. source-watching_4-alerts-this-quarter.pdfBefore and after side by side, dependent specifics listed
✎ Framework · sources designated by the 9 market managers, distribution amendments, alert log
In concrete terms, which languages do you answer in, and on what?
Nine languages, on the channels each market has chosen. 6,790 of your 7,400 monthly enquiries are handled end to end, or 92 %.
What the month gave, channel by channel:
· Email — 4,100 enquiries, all nine markets. First reply in 4 minutes.
· Chat — 2,100 conversations, six markets. Immediate reply, with the customer's appliance already loaded if they registered their warranty.
· Phone — 1,200 calls, four markets. Identified by number, answered in the caller's language, and handed to the local team with the file on screen before they even pick up.
The figure worth the most commercially, and it belongs to your small markets: first-reply time ranged from 40 minutes in France to 3 days in Sweden and Poland. It is 4 minutes everywhere. It is not your big markets that gain most, it is the ones where you had a single person.
What that shifts: routine questions took 50 % of your teams' time; they now take 8 %.
And consistency can be checked, which is the point of the page: the same question asked on the same day in all nine languages gets the same substantive answer, with each market's specific added. I had your market managers test it on 40 questions: 40 times nine concordant answers.
What I suggest: open chat in the three markets that do not have it. Across the six where it is open, it carries 28 % of the volume and has the best first-contact resolution — 91 %. 9-languages-3-channels_what-the-month-gave.pdf3 days → 4 minutes in the small markets
⛓ Sourced · ticketing tool, chat log, telephony, first-reply times by market
What the month gave, channel by channel:
· Email — 4,100 enquiries, all nine markets. First reply in 4 minutes.
· Chat — 2,100 conversations, six markets. Immediate reply, with the customer's appliance already loaded if they registered their warranty.
· Phone — 1,200 calls, four markets. Identified by number, answered in the caller's language, and handed to the local team with the file on screen before they even pick up.
The figure worth the most commercially, and it belongs to your small markets: first-reply time ranged from 40 minutes in France to 3 days in Sweden and Poland. It is 4 minutes everywhere. It is not your big markets that gain most, it is the ones where you had a single person.
What that shifts: routine questions took 50 % of your teams' time; they now take 8 %.
And consistency can be checked, which is the point of the page: the same question asked on the same day in all nine languages gets the same substantive answer, with each market's specific added. I had your market managers test it on 40 questions: 40 times nine concordant answers.
What I suggest: open chat in the three markets that do not have it. Across the six where it is open, it carries 28 % of the volume and has the best first-contact resolution — 91 %. 9-languages-3-channels_what-the-month-gave.pdf3 days → 4 minutes in the small markets
⛓ Sourced · ticketing tool, chat log, telephony, first-reply times by market
Our customers must know who they are talking to, in their own language.
They know from the first word, in all nine languages, and the text of each disclosure was written by the manager of the market concerned — not translated by me.
Why the disclosure is there: the European regulation on artificial intelligence has required, since 2 August 2026, that anyone interacting with an AI system be informed of it, unless it is obvious. Here it is done on all three channels, in the caller's language, and it is logged market by market.
And moving to a person takes one word, in every language: your nine market managers wrote the recognised wordings themselves. The Swedish manager chose a word that is not the literal translation of “adviser” — it is the one his customers actually use, and he was right: that word accounts for 61 % of Swedish handover requests.
What that gives: 288 handover requests over the month, 288 granted, with the full file in front of the local team and without the customer repeating anything.
What the disclosure produced, since that was your concern: satisfaction on exchanges announced as automated is 4.4 out of 5, against 4.5 for exchanges held by a person. In the markets where the first reply used to take three days, it went from 3.2 to 4.4. What counts for a customer is being understood in their own language, and quickly.
What I suggest: that the handover word stays on screen permanently on chat, in the six markets where it is open. A customer who knows they can step out of an automated exchange stays in it more willingly, and your own figures bear that out. ai-disclosure_9-languages-and-the-handover-word.pdf288 handovers asked for, 288 granted with the file
✎ Framework · welcome messages for the 9 markets, handover request log, satisfaction surveys
Why the disclosure is there: the European regulation on artificial intelligence has required, since 2 August 2026, that anyone interacting with an AI system be informed of it, unless it is obvious. Here it is done on all three channels, in the caller's language, and it is logged market by market.
And moving to a person takes one word, in every language: your nine market managers wrote the recognised wordings themselves. The Swedish manager chose a word that is not the literal translation of “adviser” — it is the one his customers actually use, and he was right: that word accounts for 61 % of Swedish handover requests.
What that gives: 288 handover requests over the month, 288 granted, with the full file in front of the local team and without the customer repeating anything.
What the disclosure produced, since that was your concern: satisfaction on exchanges announced as automated is 4.4 out of 5, against 4.5 for exchanges held by a person. In the markets where the first reply used to take three days, it went from 3.2 to 4.4. What counts for a customer is being understood in their own language, and quickly.
What I suggest: that the handover word stays on screen permanently on chat, in the six markets where it is open. A customer who knows they can step out of an automated exchange stays in it more willingly, and your own figures bear that out. ai-disclosure_9-languages-and-the-handover-word.pdf288 handovers asked for, 288 granted with the file
✎ Framework · welcome messages for the 9 markets, handover request log, satisfaction surveys
And the quality of the language? We have been pulled up on a Polish text before.
It is scored every week by your own local teams, and I give you the first month's figure exactly as it stands: 214 answers judged clumsy out of 6,790, or 3.2 %.
The cause, and it is clear-cut: 171 of the 214 were in Polish and Swedish, and concerned forms of politeness rather than substance. Those two markets had only 40 and 60 validated texts in the style memory, against 900 in French. The style memory is the set of texts a market has already validated: it is by reading them that I recover its way of addressing a customer.
What I did with that, and the measured result: the two market managers each validated 120 texts, taken from their own past answers. The following month: 41 answers judged clumsy out of 6,790, or 0.6 %. The 214 became 41 in thirty days, and the fix cost two half-days of reading.
What I do with the remaining 41: each one is reviewed with the local team, and the wording chosen enters the style memory the same day. The second time round, the phrasing is right on its own.
What I suggest, and your team does the scoring, never me: a weekly review of 30 exchanges drawn at random plus the 10 longest, per market. Twenty minutes a week per manager — and in the two markets that have already done this work, the language score went from 3.8 to 4.7 out of 5 in six weeks. language-quality_214-then-41.pdf3.2 % → 0.6 % in thirty days, cause and fix
⛓ Sourced · weekly reviews by the 9 local teams, style memory by market
The cause, and it is clear-cut: 171 of the 214 were in Polish and Swedish, and concerned forms of politeness rather than substance. Those two markets had only 40 and 60 validated texts in the style memory, against 900 in French. The style memory is the set of texts a market has already validated: it is by reading them that I recover its way of addressing a customer.
What I did with that, and the measured result: the two market managers each validated 120 texts, taken from their own past answers. The following month: 41 answers judged clumsy out of 6,790, or 0.6 %. The 214 became 41 in thirty days, and the fix cost two half-days of reading.
What I do with the remaining 41: each one is reviewed with the local team, and the wording chosen enters the style memory the same day. The second time round, the phrasing is right on its own.
What I suggest, and your team does the scoring, never me: a weekly review of 30 exchanges drawn at random plus the 10 longest, per market. Twenty minutes a week per manager — and in the two markets that have already done this work, the language score went from 3.8 to 4.7 out of 5 in six weeks. language-quality_214-then-41.pdf3.2 % → 0.6 % in thirty days, cause and fix
⛓ Sourced · weekly reviews by the 9 local teams, style memory by market
We want to open Czechia next year. Poland took us seven months.
The file is ready, and the route is five weeks — four of them waiting on your answers, not mine.
What actually has to be done, and what it costs:
· Translate the 312 reference articles into Czech: 40 minutes. They are already generated and waiting to be read over.
· Document 11 local specifics. I have identified and worked them up: warranty and extension, return period and procedure, information to be included, appliance take-back, service and mediation contacts. For each one, the precise question and the source to check.
· Appoint a market manager. That is the one thing I cannot prepare, and it is the one that sets the calendar.
· Write the AI disclosure and the handover word, in Czech, by that manager.
What Poland cost, according to your own records: seven months, €41,000 of translation and writing, and two temporary staff for four months to absorb the launch — €27,000. That is €68,000 and half a year.
What Czechia costs in this arrangement: translation is included, the writing is done, and support runs from day one. What is left is 11 validations and one appointment.
And one thing I will say before you ask: if a market opens in a single language and has no specifics of its own, a lighter multilingual support offer is enough — and I will say so every time. Czechia is not in that case: it has 11 specifics and you are opening three channels there. opening-a-10th-market_the-5-week-route.pdf312 articles, 11 specifics, 1 appointment
⛓ Sourced · register of specifics across the 9 markets, records of the Polish launch, reference base
What actually has to be done, and what it costs:
· Translate the 312 reference articles into Czech: 40 minutes. They are already generated and waiting to be read over.
· Document 11 local specifics. I have identified and worked them up: warranty and extension, return period and procedure, information to be included, appliance take-back, service and mediation contacts. For each one, the precise question and the source to check.
· Appoint a market manager. That is the one thing I cannot prepare, and it is the one that sets the calendar.
· Write the AI disclosure and the handover word, in Czech, by that manager.
What Poland cost, according to your own records: seven months, €41,000 of translation and writing, and two temporary staff for four months to absorb the launch — €27,000. That is €68,000 and half a year.
What Czechia costs in this arrangement: translation is included, the writing is done, and support runs from day one. What is left is 11 validations and one appointment.
And one thing I will say before you ask: if a market opens in a single language and has no specifics of its own, a lighter multilingual support offer is enough — and I will say so every time. Czechia is not in that case: it has 11 specifics and you are opening three channels there. opening-a-10th-market_the-5-week-route.pdf312 articles, 11 specifics, 1 appointment
⛓ Sourced · register of specifics across the 9 markets, records of the Polish launch, reference base
And support through the first six months? That is where we cracked in Poland.
It runs at full capacity on day one, across the three channels. What Poland took five months to reach is there from day 1.
The comparison, month by month, on your own Polish figures:
· First-reply time: Poland, 3 days in month 1, under 10 minutes by month 5. Czechia: 4 minutes on day 1.
· Share of enquiries handled end to end: Poland, 34 % in month 1, 71 % by month 6. Czechia: 88 % expected in month 1, because the 312 articles are already there and 94 % of a new market's questions already exist in the register of the other nine.
· Human reinforcement: Poland, two temporary staff for four months. Czechia: none.
What that frees up, and it is the real subject: your 16 people no longer absorb a launch. On the Polish launch they spent 640 hours handling first-level questions in approximate Polish. Those 640 hours go to supporting Czech resellers, which is the only work that sells appliances.
The figure I suggest tracking from month one: the number of reseller and installer enquiries handled the same day. Across your nine current markets, that figure went from 61 to 340 a quarter, and 96 of those enquiries produced an order, worth €214,000.
What I suggest: open Czechia at the same time as the monthly review of specifics, so the new market manager walks into a meeting that already exists. The only launch that goes off the rails is the one handled separately. launch_poland-compared-with-czechia.pdf88 % from month 1, 0 temporary staff, 640 hours freed
⛓ Sourced · records of the Polish launch, register of specifics, CRM (reseller enquiries)
The comparison, month by month, on your own Polish figures:
· First-reply time: Poland, 3 days in month 1, under 10 minutes by month 5. Czechia: 4 minutes on day 1.
· Share of enquiries handled end to end: Poland, 34 % in month 1, 71 % by month 6. Czechia: 88 % expected in month 1, because the 312 articles are already there and 94 % of a new market's questions already exist in the register of the other nine.
· Human reinforcement: Poland, two temporary staff for four months. Czechia: none.
What that frees up, and it is the real subject: your 16 people no longer absorb a launch. On the Polish launch they spent 640 hours handling first-level questions in approximate Polish. Those 640 hours go to supporting Czech resellers, which is the only work that sells appliances.
The figure I suggest tracking from month one: the number of reseller and installer enquiries handled the same day. Across your nine current markets, that figure went from 61 to 340 a quarter, and 96 of those enquiries produced an order, worth €214,000.
What I suggest: open Czechia at the same time as the monthly review of specifics, so the new market manager walks into a meeting that already exists. The only launch that goes off the rails is the one handled separately. launch_poland-compared-with-czechia.pdf88 % from month 1, 0 temporary staff, 640 hours freed
⛓ Sourced · records of the Polish launch, register of specifics, CRM (reseller enquiries)
And if a Czech customer asks something we have never had anywhere?
They get a useful answer within the minute, a date that is kept for the rest, and your market manager gets the file done. Here is the exact protocol, the one that produced 214 new specifics this quarter.
What goes to the customer immediately: what the reference article says and what holds everywhere · what the other nine markets have documented on the same point, presented as a company practice and not as a rule of their country · and the dated commitment: local answer before noon tomorrow.
What goes to the market manager in the same second: the question restated · the reference passage · the eight other markets' answers side by side · the precise source to consult, which I name · and the answer drafted in their language, ready to confirm. Observed time: 40 minutes.
What happens next, and this is what makes the arrangement cumulative: their answer becomes a dated, signed, citable specific — and it serves the next customer within seconds. Of the 214 new specifics this quarter, 168 already had their equivalent in another market: only the local confirmation was missing.
What I have already prepared for Czechia, without waiting: the 30 most-asked questions across your nine markets, translated, each with the Czech specific to be confirmed. Your future market manager starts with 30 answers to validate, not with a blank page.
And for the documents that come with a launch: manuals, packaging, terms of sale — your document translation agent works from the same single source, and your enterprise knowledge agent makes all of it searchable from one place. Three agents, one source. new-question_the-protocol-and-the-30-already-ready.pdfCustomer answer within the minute, specific in 40 minutes
⛓ Sourced · work-up protocol, register of specifics across the 9 markets, most-asked questions
What goes to the customer immediately: what the reference article says and what holds everywhere · what the other nine markets have documented on the same point, presented as a company practice and not as a rule of their country · and the dated commitment: local answer before noon tomorrow.
What goes to the market manager in the same second: the question restated · the reference passage · the eight other markets' answers side by side · the precise source to consult, which I name · and the answer drafted in their language, ready to confirm. Observed time: 40 minutes.
What happens next, and this is what makes the arrangement cumulative: their answer becomes a dated, signed, citable specific — and it serves the next customer within seconds. Of the 214 new specifics this quarter, 168 already had their equivalent in another market: only the local confirmation was missing.
What I have already prepared for Czechia, without waiting: the 30 most-asked questions across your nine markets, translated, each with the Czech specific to be confirmed. Your future market manager starts with 30 answers to validate, not with a blank page.
And for the documents that come with a launch: manuals, packaging, terms of sale — your document translation agent works from the same single source, and your enterprise knowledge agent makes all of it searchable from one place. Three agents, one source. new-question_the-protocol-and-the-30-already-ready.pdfCustomer answer within the minute, specific in 40 minutes
⛓ Sourced · work-up protocol, register of specifics across the 9 markets, most-asked questions
One quarter. What has it given us?
The equivalent of 4.3 posts returned out of 16, €214,000 of reseller orders over the quarter, and zero substantive gaps across your nine languages.
What moved on the three measures you were tracking:
· Keeping the languages consistent: 60 % → 7 % of a support writer's time.
· Routine questions: 50 % → 8 %.
· Checking local specifics: 35 % → 6 %.
What those hours became, according to your own records: 340 reseller and installer enquiries handled the same day this quarter, against 61. 96 produced an order, worth €214,000. It is the one figure on this list that reads straight off your revenue, and it comes from time given back, not from one more tool.
The four measures I always report together: first-reply time 4 minutes across the nine markets · first-contact resolution 88 %, against 64 % · answers judged clumsy 0.6 % · substantive gaps between languages: 0, against 148 at take-on. None of the four reads without the other three — a fast answer that says something different from the version next door costs more than a slow one.
Compared with what: with yourselves first — the previous quarter, on the same 7,400 monthly enquiries. Then with the sector, as an indication: in multilingual consumer-goods support, first-contact resolution is commonly seen between 70 and 85 %. Your 88 % puts you above that range, and the single source accounts for most of it.
What I suggest for the coming quarter: extend the same source to manuals, packaging and terms of sale, and launch Czechia. The two prepare in parallel and need only one appointment. quarterly-review_what-support-regained.pdf4.3 posts returned, 0 substantive gaps, €214,000
⛓ Sourced · ticketing tool, CRM, satisfaction surveys, activity records across the 9 markets
What moved on the three measures you were tracking:
· Keeping the languages consistent: 60 % → 7 % of a support writer's time.
· Routine questions: 50 % → 8 %.
· Checking local specifics: 35 % → 6 %.
What those hours became, according to your own records: 340 reseller and installer enquiries handled the same day this quarter, against 61. 96 produced an order, worth €214,000. It is the one figure on this list that reads straight off your revenue, and it comes from time given back, not from one more tool.
The four measures I always report together: first-reply time 4 minutes across the nine markets · first-contact resolution 88 %, against 64 % · answers judged clumsy 0.6 % · substantive gaps between languages: 0, against 148 at take-on. None of the four reads without the other three — a fast answer that says something different from the version next door costs more than a slow one.
Compared with what: with yourselves first — the previous quarter, on the same 7,400 monthly enquiries. Then with the sector, as an indication: in multilingual consumer-goods support, first-contact resolution is commonly seen between 70 and 85 %. Your 88 % puts you above that range, and the single source accounts for most of it.
What I suggest for the coming quarter: extend the same source to manuals, packaging and terms of sale, and launch Czechia. The two prepare in parallel and need only one appointment. quarterly-review_what-support-regained.pdf4.3 posts returned, 0 substantive gaps, €214,000
⛓ Sourced · ticketing tool, CRM, satisfaction surveys, activity records across the 9 markets
Our customers' exchanges across nine countries — where do they go? And do they move between markets?
Nothing leaves your walls, and nothing moves between markets. I run on local inference on a machine at head office, and your nine country folders stay partitioned.
Local inference means the model computes on your machine: the text of a customer email crosses no outside network to be processed. If you would rather not host a machine, the other route is an isolated resource hosted in France, dedicated to Aldevine — no pooling with another manufacturer.
Partitioning by market, since that is your question: a customer's personal data in one market is not readable from another. Role-based access — rights follow the job AND the country: your Swedish market manager opens Swedish files, not Polish ones. What does move between markets is the reference articles and the register of specifics — company content, never people. The log shows it line by line, market by market.
How the rest is held: encryption in transit and at rest · a full log of who consulted which file and when · retention periods you set market by market, with automatic purge at the deadline · hosting in France, under French law, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity, including where a US provider hosts in Europe.
And your exchanges feed no outside model: what I learn from your customers serves your customers. That is a written undertaking, not a setting.
What it opens commercially: two of your distributors wrote the absence of transfer outside the European Union into their renewal contract, and your largest German reseller scored that point out of 5 in its latest supplier questionnaire. You ticked the box without reservation.
What I suggest: that I keep the processing record up to date, market by market — hosting, subprocessors, retention periods, who has access to what. It gets asked for once a year per country, and it used to take three days to find each time. technical-framework_partitioning-by-market.pdfLocal inference, 9 markets partitioned, processing in the EU targeted
✎ Framework · deployment architecture, access log by market, supplier questionnaires
Local inference means the model computes on your machine: the text of a customer email crosses no outside network to be processed. If you would rather not host a machine, the other route is an isolated resource hosted in France, dedicated to Aldevine — no pooling with another manufacturer.
Partitioning by market, since that is your question: a customer's personal data in one market is not readable from another. Role-based access — rights follow the job AND the country: your Swedish market manager opens Swedish files, not Polish ones. What does move between markets is the reference articles and the register of specifics — company content, never people. The log shows it line by line, market by market.
How the rest is held: encryption in transit and at rest · a full log of who consulted which file and when · retention periods you set market by market, with automatic purge at the deadline · hosting in France, under French law, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity, including where a US provider hosts in Europe.
And your exchanges feed no outside model: what I learn from your customers serves your customers. That is a written undertaking, not a setting.
What it opens commercially: two of your distributors wrote the absence of transfer outside the European Union into their renewal contract, and your largest German reseller scored that point out of 5 in its latest supplier questionnaire. You ticked the box without reservation.
What I suggest: that I keep the processing record up to date, market by market — hosting, subprocessors, retention periods, who has access to what. It gets asked for once a year per country, and it used to take three days to find each time. technical-framework_partitioning-by-market.pdfLocal inference, 9 markets partitioned, processing in the EU targeted
✎ Framework · deployment architecture, access log by market, supplier questionnaires
And day to day, what goes out without us asking?
Three actions, named, and each reversible in a word. Everything else goes through a market manager, who receives it worked up in forty minutes instead of three days.
· Answering on a reference article and a validated specific, citing both with their date and validator. And the reverse holds too: if a specific goes under review, I say so to the customer in the answer and offer a call back from the local team — that happened 11 times this quarter, 11 confirmations the same day.
· Propagating a rewrite across the nine languages, specifics reapplied, with a before-and-after to each market manager. Each of them can freeze publication on their market in a word, and has 48 hours to object.
· Handing over to the local team as soon as a customer asks, in the wordings your nine managers wrote themselves.
What goes through a person, and you get it in minutes: any statement about a rule specific to a country that is not yet in the register, any creation or change of a specific, any goodwill gesture. You ask me, I hand you the file worked up — question, source to check, eight markets side by side, answer drafted — and the validation is your act.
What the log keeps: 7,400 exchanges a month, their market, their language, their channel, the reference article cited, the specific applied with its date and validator, and who handled it. It exports by market in one click, and it answers a distributor or a complaint in ten minutes, evidence in hand.
And you keep control wherever you are: a web dashboard, and supervision from your phone — a market manager validates a specific, freezes a publication or closes a channel in one message, from their country and in their language. who-decides-what_three-automatic-actions.pdf3 reversible actions, all the rest validated by a market
✎ Framework · exchange log by market, register of specifics, publication log
· Answering on a reference article and a validated specific, citing both with their date and validator. And the reverse holds too: if a specific goes under review, I say so to the customer in the answer and offer a call back from the local team — that happened 11 times this quarter, 11 confirmations the same day.
· Propagating a rewrite across the nine languages, specifics reapplied, with a before-and-after to each market manager. Each of them can freeze publication on their market in a word, and has 48 hours to object.
· Handing over to the local team as soon as a customer asks, in the wordings your nine managers wrote themselves.
What goes through a person, and you get it in minutes: any statement about a rule specific to a country that is not yet in the register, any creation or change of a specific, any goodwill gesture. You ask me, I hand you the file worked up — question, source to check, eight markets side by side, answer drafted — and the validation is your act.
What the log keeps: 7,400 exchanges a month, their market, their language, their channel, the reference article cited, the specific applied with its date and validator, and who handled it. It exports by market in one click, and it answers a distributor or a complaint in ten minutes, evidence in hand.
And you keep control wherever you are: a web dashboard, and supervision from your phone — a market manager validates a specific, freezes a publication or closes a channel in one message, from their country and in their language. who-decides-what_three-automatic-actions.pdf3 reversible actions, all the rest validated by a market
✎ Framework · exchange log by market, register of specifics, publication 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 markets. 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 664 € excl. VAT / month
Answers by language
Answers in each language chosen, from the same reference source.
Local particulars
Applies the lead times, warranties and mandatory notices documented per market.
Several channels
Works on the contact channels chosen for each market.
Controls and safeguards
These 5 controls are built into the agent: they frame what it does, whatever plan you pick. They are not chosen and are not added to your order.
Human validation, exceptions and escalation
Status, safe closure and audit trail
Sources, access rights and handling of questions with no answer
Resolve common requests through to a verifiable closure
Escalate quickly to a human, with the reason and the history
What the agent must be connected to
This connection is required for the agent to work. It concerns your information system and is scoped during the audit.
Carry the full context across channels and into the ticket/CRM
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.
Multilingual support
For two to three languages with no per-market particulars, a lighter offer is enough.
Multilingual support agent (2–3 languages) from 654 € excl. VAT / month Multilingual support →Company knowledge
To unify search across all your sources, a dedicated agent completes the picture.
Enterprise knowledge management agent from 708 € excl. VAT / month Company knowledge →
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 many markets can a team serve without drift?
By drawing on a single source, the effort shifts from maintaining versions to supporting the markets. How large the gain is depends on your volume and remains to be confirmed by a pilot.
Keeping languages consistent
Answers to everyday questions
Checking the local particulars
Indicative figures, not contractual, to be confirmed by a pilot on your number of languages, channels and markets. Anything touching a country's legal obligation is documented before it is said: questions not covered go to the local team.
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
An international support agent (languages, channels, local particulars), installed and operated for you.
Agility
Setup + controlled subscription
6,450 € excl. VAT setup
then 664 € 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
1,024 € 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
9,965 € excl. VAT setup
then 861 € 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 markets
Your customer exchanges stay with youLocal inference or an isolated resource hosted in France; no exchange entrusted to a third party, processing in the EU targeted, no conversation used to train a model.
Data in France, under French lawYour exchanges with customers around the world: minimisation and location in France, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The customer service department keeps the decisionThe agent produces consistent support in several languages, which can be checked and altered; no approval is automated.
Human oversight & traceabilityOn your number of languages, channels and markets: systematic logging and tracking, in line with the AI Act.
Going further
Related resources
Frequently asked questions
Your questions, our answers
How is consistency between languages upheld?
Through a single source: the reference content is maintained once, and each language draws on it. Differences by market exist only where you have documented them.
What does it do with a question about local law?
If the particular is documented, it applies it and cites it. Otherwise the question goes to the local team: a legal obligation is not something to improvise.
How many languages and channels?
This offer covers several languages and the channels your customers already use in each market: WhatsApp Business, your website chat and email. Your local teams follow the exchanges from Microsoft Teams, Slack or their email. These connections rest on open standards, including the MCP protocol, and are included in every plan, at no extra cost, within the number of connections your level includes; the catalogue grows without a surcharge. Only the fees charged by third-party platforms — WhatsApp Business bills per conversation — are passed on to you at cost, without margin, outside the subscription. The exact scope is settled with you according to your markets.
Does our customers' data move between markets?
No. The partitioning by market is defined at the design stage, and every access is logged.
Where are the exchanges 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.
Does the agent state that it is an artificial intelligence?
Yes, from the very first interaction, and this is not a configuration option: since 2 August 2026, Article 50(1) of the European AI Regulation requires that any person interacting with an AI system be informed, unless this is obvious. The announcement is built into the greeting, in the other party’s language, and they can ask for a human at any time.
How long does it take to deploy this agent?
A few months as a rule, depending on the number of languages, channels and local particulars to document, after a free audit then phases of design, integration and testing.
Related use cases
Other agents for customer relations
Let's talk
Let's size up the potential in your international support
15 minutes to frame your markets and your channels — hosted in France, supervised, with no commitment.