● B2B offer — Omnichannel customer relations
Customer relations: a single thread, whatever the channel
A customer who writes, then calls, then comes back on chat expects to be remembered. Your agent keeps a single thread per customer, whatever the channel used, and picks the conversation up where it left off. Every interaction is recorded in your CRM. Hosted in France: your customers' exchanges and data stay with you. Your advisers take over on what calls for their judgement.
Hosted in France
Customer histories protected
GDPR & AI Act: governed deployment
Human oversight
Updated on
Deployed in a few weeks
Omnichannel customer relations · hosted in France
A customer is calling after writing by email yesterday.
This customer's thread is found: yesterday's email, the answer given, the question left open.
The conversation picks up where it left off, without asking again for what has already been said.
Every interaction is recorded in your CRM, with the channel noted.
🔗 Sourced · unified customer thread, all channels
The conversation picks up where it left off, without asking again for what has already been said.
Every interaction is recorded in your CRM, with the channel noted.
🔗 Sourced · unified customer thread, all channels
The customer is pressing on a commercial request.
The full thread is passed to an adviser, with the history of exchanges and the request as made.
A goodwill gesture or a commitment on timing rests with the adviser: the customer does not have to retell their story to get one.
✎ Support · thread passed on, the adviser's decision
A goodwill gesture or a commitment on timing rests with the adviser: the customer does not have to retell their story to get one.
✎ Support · thread passed on, the adviser's decision
Local inference · no data outside the EU
Exchanges 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 relations department decidesThe agent prepares, never rules
✦ In brief
A Blue Lemon Agent omnichannel agent keeps a single thread per customer — email, chat, telephone — picks the conversation up where it left off and records every interaction in your CRM. When the request calls for an adviser, the full thread is passed to them. It runs on local inference or is hosted in France: your customers' 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
customer relations 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 channels and volume of exchanges is confirmed by a pilot.
The context
What does an AI agent bring to your customer relations?
Not having to retell your story: that is the first sign of customer relations done well.
! The issue
A customer uses several channels for a single request. Reconstructing their journey at every contact takes time on both sides. The agent keeps a single thread per customer, picks the conversation up at exactly the point it stopped and records every interaction in the CRM with its channel.
✓ Our answer
Your customers find a continuous conversation, and your advisers receive the full thread when the request calls for their judgement — a goodwill gesture, a commitment on timing, an unusual situation. Local inference or an isolated resource hosted in France: your customers' exchanges and data are entrusted to no third party.
The decisive point
Your customers' exchanges and data: sovereignty & compliance
A unified customer thread by its nature brings together all of one person's exchanges. Here is how that concentration is protected.
Local inference
The agent can run on a machine belonging to your organisation: no exchange and no customer history leaves the network.
Hosting in France
Otherwise, a dedicated and isolated resource hosted in France, under French law — your contact channels and your customer histories: processing and access within the European Union targeted by the architecture.
Reduced extraterritorial exposure
For your customers' exchanges and data, 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 contact channels.
One thread per customer, recorded channel by channel
Every interaction is recorded with its channel and its timestamp; encryption, role-based access and purging at the retention periods you set.
AI Act: governed deployment
The agent is strictly in support; no goodwill gesture is granted and no commitment on timing is made automatically; traceability and human oversight from end to end.
What depends on the architecture chosen These points are not general guarantees: they are settled deployment by deployment, in the quotation.
- The applicable location is that of the architecture set out in the quotation and verified before commissioning.
- Local execution is announced only for the configuration explicitly described and accepted in the quotation.
- The applicable isolation depends on the deployment mode set out in the quotation; no dedicated isolation is presumed.
- Roles and permissions are configured and accepted for the identities and systems actually connected.
- The events logged, their content, their retention period and who may access them are defined for the deployment chosen.
For exchanges concerning a dispute, an unpaid invoice or a sensitive complaint, 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 companyMaison Perrat — home furnishing retailer, 4 stores and an online shop
- Sector
- Home furnishing retail — furniture, bedding, tableware
- Headcount
- 118 staff, including a customer service team of 9 advisers and 1 manager
- Customers served
- Consumers across mainland France — 62,000 active customers
- Volume
- 3,400 enquiries a month — 1,900 emails, 900 chat conversations, 600 calls; average basket €240
- Tools in place
- CRM, shared inbox, website chat module, telephony, order ERP — the agent plugs into them, nothing is replaced
- Who decides
- The customer service manager arbitrates goodwill gestures; the 9 advisers hold the exchanges that call for judgement
- Room for improvement
- One customer in three switches channel mid-enquiry and tells their story twice; rebuilding the journey takes 55 % of an adviser's time; a first reply by email arrives in 14 hours
Maison Perrat sells as much in store as online, and its customer service team takes three channels at once. The agent runs on local inference on a machine at head office and plugs into the CRM, the shared inbox, the website chat, the telephony and the order ERP: it holds the thread, answers routine enquiries and prepares everything that calls for an adviser. The exchanges below cover one quarter, from the reopening of the archives to the review.
This company, its figures and the exchanges that follow were invented for the demonstration. They illustrate a common situation; they describe no real client.
Omnichannel customer relations · hosted in France
I went back over your last 18 months of exchanges overnight — 214,000 interactions across every channel — and attached them to 62,000 customers. Every customer now has a single thread, and it starts where their last conversation left off.
A single thread is everything a customer has written or said to you, in order, whatever the channel: email, chat and phone read on the same timeline.
The gap I measured, and it is the gap that decides the gain: one customer in three switches channel mid-enquiry. They write on Monday, call on Wednesday, come back on chat on Friday. Across those 21,000 customers, the story was told 2.1 times on average — and each time, an adviser started the rebuild from scratch.
What that was costing, measured on your own handling times: rebuilding the journey took 55 % of an adviser's time; it now takes 6 %. Across 9 advisers, that is close to 4 posts given back to handling the enquiry itself.
What changes for the customer, and this is the part that shows: the conversation resumes at the exact point it stopped, with what has already been said, what has already been promised and by whom. Of your 3,400 enquiries this month, 1,130 come from a customer who already wrote to you within 30 days.
What I suggest: the customer service manager reads ten rebuilt threads of their own choosing, the ones they know by heart, and the single thread opens to all 9 advisers as soon as they approve it. Your tools do not change: the thread appears inside your CRM, exactly where your advisers already look. single-customer-thread_214000-interactions.pdf62,000 customers, 1 in 3 switches channel
⛓ Sourced · CRM, shared inbox, chat log, telephony log — 18 months, 214,000 interactions
A single thread is everything a customer has written or said to you, in order, whatever the channel: email, chat and phone read on the same timeline.
The gap I measured, and it is the gap that decides the gain: one customer in three switches channel mid-enquiry. They write on Monday, call on Wednesday, come back on chat on Friday. Across those 21,000 customers, the story was told 2.1 times on average — and each time, an adviser started the rebuild from scratch.
What that was costing, measured on your own handling times: rebuilding the journey took 55 % of an adviser's time; it now takes 6 %. Across 9 advisers, that is close to 4 posts given back to handling the enquiry itself.
What changes for the customer, and this is the part that shows: the conversation resumes at the exact point it stopped, with what has already been said, what has already been promised and by whom. Of your 3,400 enquiries this month, 1,130 come from a customer who already wrote to you within 30 days.
What I suggest: the customer service manager reads ten rebuilt threads of their own choosing, the ones they know by heart, and the single thread opens to all 9 advisers as soon as they approve it. Your tools do not change: the thread appears inside your CRM, exactly where your advisers already look. single-customer-thread_214000-interactions.pdf62,000 customers, 1 in 3 switches channel
⛓ Sourced · CRM, shared inbox, chat log, telephony log — 18 months, 214,000 interactions
Take a real case. Someone who wrote and then called — what does the adviser see?
Here is Tuesday's case, a customer who used all three channels in eight days. The adviser opens one page and has everything.
What the thread shows, in order:
· Monday 3rd, 9.14 p.m., email — a sofa delivered with a leg missing, photo attached, order number recognised automatically.
· Tuesday 4th, 9.02 a.m., reply — part identified in the ERP, available within 5 days, dispatch offered.
· Thursday 6th, 2.40 p.m., call — the customer asks whether the fitter can come back. The adviser had read the email and the reply before picking up: the conversation started at the question, not at a recap of the facts.
· Monday 10th, 11.27 a.m., chat — dispatch tracking, given in three seconds straight from the ERP.
What the adviser used to see: three tools, three windows, and an order number to retype. Average time between picking up and the first useful answer was 2 minutes 40; it is now 20 seconds.
The detail that makes the difference, and it is measurable: the commitment made in writing on the Tuesday sits at the top of the thread — “part within 5 days”. Across your last 18 months, 1,340 complaints were about a commitment made on another channel and lost from view. Those are precisely the ones that disappear.
What I suggest: that every delivery commitment made in an exchange appears at the top of the thread with its date, and that I warn you the day before it falls due. Approve the principle and it applies across all three channels. a-real-case_journey-rebuilt-across-three-channels.pdf2 min 40 → 20 seconds to the first useful answer
⛓ Sourced · single customer thread, order ERP, telephony log
What the thread shows, in order:
· Monday 3rd, 9.14 p.m., email — a sofa delivered with a leg missing, photo attached, order number recognised automatically.
· Tuesday 4th, 9.02 a.m., reply — part identified in the ERP, available within 5 days, dispatch offered.
· Thursday 6th, 2.40 p.m., call — the customer asks whether the fitter can come back. The adviser had read the email and the reply before picking up: the conversation started at the question, not at a recap of the facts.
· Monday 10th, 11.27 a.m., chat — dispatch tracking, given in three seconds straight from the ERP.
What the adviser used to see: three tools, three windows, and an order number to retype. Average time between picking up and the first useful answer was 2 minutes 40; it is now 20 seconds.
The detail that makes the difference, and it is measurable: the commitment made in writing on the Tuesday sits at the top of the thread — “part within 5 days”. Across your last 18 months, 1,340 complaints were about a commitment made on another channel and lost from view. Those are precisely the ones that disappear.
What I suggest: that every delivery commitment made in an exchange appears at the top of the thread with its date, and that I warn you the day before it falls due. Approve the principle and it applies across all three channels. a-real-case_journey-rebuilt-across-three-channels.pdf2 min 40 → 20 seconds to the first useful answer
⛓ Sourced · single customer thread, order ERP, telephony log
And our duplicates? We have customers listed three times over in the CRM.
81,400 records for 62,000 real customers: 19,400 duplicates matched, each with the evidence that justifies it and each reversible.
How I match, and what I count as evidence: the same order number on two records · the same phone number and the same surname · the same delivery address and the same loyalty card. Every match carries its evidence, in plain words, and comes undone in one click.
What that gives: 17,900 matches on single, strong evidence, applied. 1,500 on thinner evidence — same name and town, different addresses — which I left side by side with the question asked rather than merged: your manager settles them in batches of fifty, an hour in total.
What the duplicates were costing, measured on your data: 2,300 customers received the same order reminder twice over 18 months, and 418 enquiries were handled twice by two different advisers. At an average handling time of 11 minutes, that is 77 hours of customer service spent twice.
What matching brings beyond the service desk: 4,700 customers looked like one-time buyers when in fact they have ordered three to five times. Those are your best customers, and they were invisible. On an average basket of €240, that segment accounts for €3.4 m of orders already placed.
What I suggest: that matching runs continuously on newly created records, and that I hand you the 4,700 loyal customers found again each month so your marketing team can treat them as such. The list is ready. record-matching_19400-duplicates.pdf17,900 applied, 1,500 put as a question, all reversible
⛓ Sourced · CRM (81,400 records), order ERP, reminder log, average handling times
How I match, and what I count as evidence: the same order number on two records · the same phone number and the same surname · the same delivery address and the same loyalty card. Every match carries its evidence, in plain words, and comes undone in one click.
What that gives: 17,900 matches on single, strong evidence, applied. 1,500 on thinner evidence — same name and town, different addresses — which I left side by side with the question asked rather than merged: your manager settles them in batches of fifty, an hour in total.
What the duplicates were costing, measured on your data: 2,300 customers received the same order reminder twice over 18 months, and 418 enquiries were handled twice by two different advisers. At an average handling time of 11 minutes, that is 77 hours of customer service spent twice.
What matching brings beyond the service desk: 4,700 customers looked like one-time buyers when in fact they have ordered three to five times. Those are your best customers, and they were invisible. On an average basket of €240, that segment accounts for €3.4 m of orders already placed.
What I suggest: that matching runs continuously on newly created records, and that I hand you the 4,700 loyal customers found again each month so your marketing team can treat them as such. The list is ready. record-matching_19400-duplicates.pdf17,900 applied, 1,500 put as a question, all reversible
⛓ Sourced · CRM (81,400 records), order ERP, reminder log, average handling times
Our three channels don't talk to each other. How do you hold all three at once?
All three are plugged in, they write into the same thread, and switching channel no longer starts anything from scratch. Here is what each one produced last month.
· Email — 1,900 enquiries. First reply in 4 minutes instead of 14 hours, at any hour. Of those 1,900, 612 arrive between 7 p.m. and 8 a.m.: they used to run a night late, they now get an answer the same evening.
· Chat — 900 conversations. Immediate reply, with the customer's order already loaded if they are signed in. The mid-conversation drop-off rate went from 23 % to 7 %: people drop off when they wait.
· Phone — 600 calls. I take the call, identify the customer from their number, settle what is routine and hand an adviser what calls for judgement, with the thread on screen before they even pick up.
The point that matters most, and it is the one your customers notice: when a customer moves from chat to phone, the conversation carries on. That is what omnichannel means: not being present on several channels, but holding a single conversation across them.
What that shifts: routine replies took 50 % of your advisers' time; they now take 8 %.
What I suggest: open them in this order — email first, where the gain in speed is most visible; chat a fortnight later; phone last, once the first two have fed the answer base. Each channel opens and closes on a single setting, without touching the other two. three-channels_one-conversation.pdf14 h → 4 min on email, chat drop-off 23 % → 7 %
⛓ Sourced · shared inbox, chat log, telephony log, first-reply times over 18 months
· Email — 1,900 enquiries. First reply in 4 minutes instead of 14 hours, at any hour. Of those 1,900, 612 arrive between 7 p.m. and 8 a.m.: they used to run a night late, they now get an answer the same evening.
· Chat — 900 conversations. Immediate reply, with the customer's order already loaded if they are signed in. The mid-conversation drop-off rate went from 23 % to 7 %: people drop off when they wait.
· Phone — 600 calls. I take the call, identify the customer from their number, settle what is routine and hand an adviser what calls for judgement, with the thread on screen before they even pick up.
The point that matters most, and it is the one your customers notice: when a customer moves from chat to phone, the conversation carries on. That is what omnichannel means: not being present on several channels, but holding a single conversation across them.
What that shifts: routine replies took 50 % of your advisers' time; they now take 8 %.
What I suggest: open them in this order — email first, where the gain in speed is most visible; chat a fortnight later; phone last, once the first two have fed the answer base. Each channel opens and closes on a single setting, without touching the other two. three-channels_one-conversation.pdf14 h → 4 min on email, chat drop-off 23 % → 7 %
⛓ Sourced · shared inbox, chat log, telephony log, first-reply times over 18 months
Our customers must know who they are talking to. We don't want to mislead anyone.
They know from the first word, and it is written into the welcome message on every channel — not into a setting someone could forget.
What appears, channel by channel: on chat, a welcome line stating that they are talking to an automated assistant of Maison Perrat · by email, the same statement at the top of the first reply plus the customer service signature · on the phone, the welcome sentence, in the caller's language.
Why that statement 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 systematically, and it is logged.
And moving to a human takes one word: “an adviser”, “a person”, “someone” — all three trigger the handover, on all three channels, without the customer repeating anything. Over the past month 174 customers asked for it, and all 174 got an adviser with the full thread in front of them.
What the statement produced, since that was your concern: satisfaction on exchanges announced as automated is 4.3 out of 5, against 4.4 for exchanges held by an adviser. What counts for your customers is the answer and the speed.
What I suggest: that you approve the wording of the three welcome messages — they are written, in the house tone — and that the handover word stays on screen permanently on chat. A customer who knows they can step out of an automated exchange stays in it more willingly. ai-disclosure_and-handover-to-an-adviser.pdf174 handovers asked for, 174 granted with the thread
✎ Framework · welcome messages on the three channels, handover request log, satisfaction scores
What appears, channel by channel: on chat, a welcome line stating that they are talking to an automated assistant of Maison Perrat · by email, the same statement at the top of the first reply plus the customer service signature · on the phone, the welcome sentence, in the caller's language.
Why that statement 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 systematically, and it is logged.
And moving to a human takes one word: “an adviser”, “a person”, “someone” — all three trigger the handover, on all three channels, without the customer repeating anything. Over the past month 174 customers asked for it, and all 174 got an adviser with the full thread in front of them.
What the statement produced, since that was your concern: satisfaction on exchanges announced as automated is 4.3 out of 5, against 4.4 for exchanges held by an adviser. What counts for your customers is the answer and the speed.
What I suggest: that you approve the wording of the three welcome messages — they are written, in the house tone — and that the handover word stays on screen permanently on chat. A customer who knows they can step out of an automated exchange stays in it more willingly. ai-disclosure_and-handover-to-an-adviser.pdf174 handovers asked for, 174 granted with the thread
✎ Framework · welcome messages on the three channels, handover request log, satisfaction scores
And on substance, what exactly do you answer?
2,720 enquiries out of 3,400 handled end to end last month, or 80 %. Here is which ones, and where each answer comes from.
· 1,190 order and delivery tracking requests — answered live from the ERP, with the tracking number and the estimated date. It is the first reason customers contact any retailer, and the easiest one to make instant.
· 640 product questions — dimensions, materials, care, compatibility — taken from your product sheets, with the link to the page.
· 410 invoice, credit note and duplicate requests — document regenerated and sent within the minute.
· 290 address or delivery slot changes — applied in the ERP while the order has not shipped, and confirmed to the customer with the new date.
· 190 returns and exchanges within the framework you set — return label issued, deadline restated, refund triggered on receipt.
First-contact resolution — the share of enquiries settled without the customer having to come back — goes from 61 % to 84 %.
The commercial figure behind all this: 2,720 instant replies pushed 118 pre-purchase questions to an adviser the same day, instead of the next. On an average basket of €240, those conversations produced 61 orders.
What I suggest next: extend the same immediacy to quote requests on orders above €1,500 — I build the quote from the ERP, the adviser reads it over and sends it. You receive 90 a month and they currently go out in 2 days. 2720-enquiries-handled_by-reason.pdf80 % end to end, resolution 61 % → 84 %
⛓ Sourced · order ERP, product sheets, returns policy, contact reasons for the month
· 1,190 order and delivery tracking requests — answered live from the ERP, with the tracking number and the estimated date. It is the first reason customers contact any retailer, and the easiest one to make instant.
· 640 product questions — dimensions, materials, care, compatibility — taken from your product sheets, with the link to the page.
· 410 invoice, credit note and duplicate requests — document regenerated and sent within the minute.
· 290 address or delivery slot changes — applied in the ERP while the order has not shipped, and confirmed to the customer with the new date.
· 190 returns and exchanges within the framework you set — return label issued, deadline restated, refund triggered on receipt.
First-contact resolution — the share of enquiries settled without the customer having to come back — goes from 61 % to 84 %.
The commercial figure behind all this: 2,720 instant replies pushed 118 pre-purchase questions to an adviser the same day, instead of the next. On an average basket of €240, those conversations produced 61 orders.
What I suggest next: extend the same immediacy to quote requests on orders above €1,500 — I build the quote from the ERP, the adviser reads it over and sends it. You receive 90 a month and they currently go out in 2 days. 2720-enquiries-handled_by-reason.pdf80 % end to end, resolution 61 % → 84 %
⛓ Sourced · order ERP, product sheets, returns policy, contact reasons for the month
CRM entry is the thing our advisers do least well. And you can see why.
It is done for them, to the second, on all 3,400 interactions of the month. Nothing is retyped and nothing is missed.
What is recorded on every interaction: the channel (email, chat or phone), the timestamp — the exact date and time, to the second, of every message and every reply —, the reason under your own classification, the order concerned, a three-line summary of the exchange, the commitment made if there is one, and who handled it — the agent or the named adviser.
What that shifts: CRM entry took 30 % of your advisers' time; it now takes 4 %. The remaining 4 % is them reading over the summary on the cases they held themselves.
What complete entry makes possible, and it did not exist before: your contact reasons are now filled in on 100 % of interactions against 46 % before. First measured consequence: the reason “leg or fixings missing on delivery” accounts for 214 enquiries a month, or 6.3 % of your entire customer service load. It was invisible, because it sat inside “other”.
What I did with that finding: I cross-referenced those 214 enquiries with the product references involved. Three sofa references carry 61 % of them, all shipped from the same warehouse. Your logistics manager has had the detail by reference since Monday.
What I suggest: a monthly statement of the top ten reasons with their handling cost, and automatic cross-referencing with the product reference and the warehouse. A reason you can see is a reason you can fix at source, and it then costs once instead of 214 times a month. crm-in-sync_channel-timestamp-reason.pdfReasons filled in 46 % → 100 %
⛓ Sourced · CRM, reason classification, order ERP, product and warehouse master data
What is recorded on every interaction: the channel (email, chat or phone), the timestamp — the exact date and time, to the second, of every message and every reply —, the reason under your own classification, the order concerned, a three-line summary of the exchange, the commitment made if there is one, and who handled it — the agent or the named adviser.
What that shifts: CRM entry took 30 % of your advisers' time; it now takes 4 %. The remaining 4 % is them reading over the summary on the cases they held themselves.
What complete entry makes possible, and it did not exist before: your contact reasons are now filled in on 100 % of interactions against 46 % before. First measured consequence: the reason “leg or fixings missing on delivery” accounts for 214 enquiries a month, or 6.3 % of your entire customer service load. It was invisible, because it sat inside “other”.
What I did with that finding: I cross-referenced those 214 enquiries with the product references involved. Three sofa references carry 61 % of them, all shipped from the same warehouse. Your logistics manager has had the detail by reference since Monday.
What I suggest: a monthly statement of the top ten reasons with their handling cost, and automatic cross-referencing with the product reference and the warehouse. A reason you can see is a reason you can fix at source, and it then costs once instead of 214 times a month. crm-in-sync_channel-timestamp-reason.pdfReasons filled in 46 % → 100 %
⛓ Sourced · CRM, reason classification, order ERP, product and warehouse master data
And the quality of what you answer? We're not going to discover that at the customer's end.
It is measured every week by your own advisers, on a reviewed sample, and I give you the first month's figure exactly as it stands.
First month: out of 3,400 enquiries, 218 answers were judged off the mark by advisers in review — 6.4 %.
The cause, and it is clear-cut: 174 of the 218 concerned three ranges added to the catalogue in January, whose technical sheets lived in a purchasing department spreadsheet rather than in the product base. I was answering accurately on what I had, and 3 ranges out of 41 were missing from it.
What I did with that, and the measured result: the 3 sheets were loaded into the base on the 12th, with dimensions, materials, care and compatibilities. The following month: 41 answers judged off the mark out of 3,400, or 1.2 %. The 218 became 41 in thirty days, and the fix cost nothing more than loading some files.
What I do with the remaining 41: each one is reviewed with the adviser, and its correct answer enters the base the same day. The second time round, the question is handled on its own. Of those 41, 29 were questions never asked before.
What I suggest: that the weekly review cover 50 exchanges drawn at random plus the 10 longest — the long ones are where something was missing, and they teach the most. One hour a week for your manager, and the measure never depends on me: your team does the scoring. answer-quality_218-then-41.pdf6.4 % → 1.2 % in thirty days, cause and fix
⛓ Sourced · weekly adviser reviews, product base, catalogue of 41 ranges
First month: out of 3,400 enquiries, 218 answers were judged off the mark by advisers in review — 6.4 %.
The cause, and it is clear-cut: 174 of the 218 concerned three ranges added to the catalogue in January, whose technical sheets lived in a purchasing department spreadsheet rather than in the product base. I was answering accurately on what I had, and 3 ranges out of 41 were missing from it.
What I did with that, and the measured result: the 3 sheets were loaded into the base on the 12th, with dimensions, materials, care and compatibilities. The following month: 41 answers judged off the mark out of 3,400, or 1.2 %. The 218 became 41 in thirty days, and the fix cost nothing more than loading some files.
What I do with the remaining 41: each one is reviewed with the adviser, and its correct answer enters the base the same day. The second time round, the question is handled on its own. Of those 41, 29 were questions never asked before.
What I suggest: that the weekly review cover 50 exchanges drawn at random plus the 10 longest — the long ones are where something was missing, and they teach the most. One hour a week for your manager, and the measure never depends on me: your team does the scoring. answer-quality_218-then-41.pdf6.4 % → 1.2 % in thirty days, cause and fix
⛓ Sourced · weekly adviser reviews, product base, catalogue of 41 ranges
And all that history — how long do we keep it?
As long as you decide, reason by reason, and the purge runs on its own at the deadline. Here are the periods I propose, aligned with your commercial obligations.
· Order tracking exchanges: 3 years after delivery — the length of your commercial warranty, the one that can be invoked.
· Complaints and goodwill gestures: 5 years — they can be produced if a case is disputed.
· Chat conversations with no order attached: 13 months.
· Call recordings: 6 months, and only where they are switched on.
What the purge has already given back: 41,000 interactions older than those periods were deleted at take-on, with a log stating how many and on what date. Your base is 19 % smaller, and searches across it are correspondingly faster.
What your customers can ask for, and how fast they get it: access to their data, correction, erasure. A customer's full file — every channel, every order, every exchange — comes out in 4 minutes, as a readable file. You receive 6 to 9 a month and each used to take half a day.
What I suggest: that the customer service manager approve these four periods, and that a monthly statement say what has been purged and what is coming due. A period changes on a word, and the change applies to the past as well as the future. retention-periods_and-purge-log.pdf4 periods, full customer file out in 4 minutes
✎ Framework · retention periods, purge log, access requests for the quarter
· Order tracking exchanges: 3 years after delivery — the length of your commercial warranty, the one that can be invoked.
· Complaints and goodwill gestures: 5 years — they can be produced if a case is disputed.
· Chat conversations with no order attached: 13 months.
· Call recordings: 6 months, and only where they are switched on.
What the purge has already given back: 41,000 interactions older than those periods were deleted at take-on, with a log stating how many and on what date. Your base is 19 % smaller, and searches across it are correspondingly faster.
What your customers can ask for, and how fast they get it: access to their data, correction, erasure. A customer's full file — every channel, every order, every exchange — comes out in 4 minutes, as a readable file. You receive 6 to 9 a month and each used to take half a day.
What I suggest: that the customer service manager approve these four periods, and that a monthly statement say what has been purged and what is coming due. A period changes on a word, and the change applies to the past as well as the future. retention-periods_and-purge-log.pdf4 periods, full customer file out in 4 minutes
✎ Framework · retention periods, purge log, access requests for the quarter
The 680 that are left, the ones you pass to our advisers — how do they arrive?
They arrive worked up: the full thread, the reason, the purchase history, what the customer is asking for and what I suggest the adviser does. They decide, they don't search.
What those 680 are:
· 310 goodwill requests — discount, credit note, out-of-window return, delivery charges waived.
· 180 delivery commitments to be made on a late order.
· 110 particular situations — a customer in the middle of a house move, a bereavement, a delivery missed twice.
· 80 technical questions with no answer in your documents, for which I have drafted the reply and sent it to the right person to approve.
What the adviser gets on each, on one page: the thread, the customer's value over 24 months, their past incidents, what the company has granted customers in the same situation over the last six months, and my costed proposal.
What that changes, measured: handling time on an enquiry taken over by an adviser goes from 19 minutes to 7, and the 12 minutes saved are search time, not relationship time. Across 680 enquiries, that is 136 hours a month given back to your 9 advisers.
What I suggest: that those 136 hours go to assisted selling at a distance, on the 90 quote requests above €1,500 you receive every month. Your advisers know the products better than anyone, and that is the one place where their time turns into revenue. 680-enquiries-worked-up_for-an-adviser.pdf19 min → 7 min, 136 hours a month given back
⛓ Sourced · customer threads, CRM, 6-month history of gestures granted, handling times
What those 680 are:
· 310 goodwill requests — discount, credit note, out-of-window return, delivery charges waived.
· 180 delivery commitments to be made on a late order.
· 110 particular situations — a customer in the middle of a house move, a bereavement, a delivery missed twice.
· 80 technical questions with no answer in your documents, for which I have drafted the reply and sent it to the right person to approve.
What the adviser gets on each, on one page: the thread, the customer's value over 24 months, their past incidents, what the company has granted customers in the same situation over the last six months, and my costed proposal.
What that changes, measured: handling time on an enquiry taken over by an adviser goes from 19 minutes to 7, and the 12 minutes saved are search time, not relationship time. Across 680 enquiries, that is 136 hours a month given back to your 9 advisers.
What I suggest: that those 136 hours go to assisted selling at a distance, on the 90 quote requests above €1,500 you receive every month. Your advisers know the products better than anyone, and that is the one place where their time turns into revenue. 680-enquiries-worked-up_for-an-adviser.pdf19 min → 7 min, 136 hours a month given back
⛓ Sourced · customer threads, CRM, 6-month history of gestures granted, handling times
Those 310 goodwill gestures are our money. We can't just let that run.
That is exactly why there is a written mandate — and it fits on one page. You set the cap, I execute in three minutes, and you withdraw it in a word.
A mandate here is a written authorisation bounded in advance: it states which gestures I may grant, up to what amount, in which cases, for how long, and how it is withdrawn.
The mandate I propose, as I have drafted it:
· Scope: granting the reshipment of a missing part, waived delivery charges, and a credit note on a delivery delay of more than 10 days.
· Cap: €40 per customer and per request, €6,000 a month across the whole service.
· Term: 12 months, with a review point at the third.
· Condition: one automatic gesture per customer per 90 days; on the second, the case goes to an adviser with the history.
· Withdrawal: immediate, on a simple message, with no reason to give.
What this mandate covers exactly: 228 of the month's 310 requests, those whose amount and reason fall inside the bounds. Monthly cost observed on the history: €4,100, under the €6,000 cap.
What it earns, measured on your own data: a gesture granted in 3 minutes rather than 2 days. Over the past 18 months, customers who got a gesture within 24 hours bought again at 68 % within the year; those who waited more than 48 hours, at 41 %. On 228 gestures a month and a €240 basket, that gap is worth €148,000 of orders over twelve months.
What stays with the manager, and they have it in three minutes: the 82 requests above €40, second requests from the same customer, and every particular situation. I hand them the costed file, they decide.
What I suggest: you sign the mandate, we look at it together at the third month, and you widen it or tighten it on what it has produced. goodwill-mandate_capped-and-dated.pdf€40/customer, €6,000/month, immediate withdrawal
✎ Framework · mandate template, history of gestures granted, repeat-purchase rate by response time
A mandate here is a written authorisation bounded in advance: it states which gestures I may grant, up to what amount, in which cases, for how long, and how it is withdrawn.
The mandate I propose, as I have drafted it:
· Scope: granting the reshipment of a missing part, waived delivery charges, and a credit note on a delivery delay of more than 10 days.
· Cap: €40 per customer and per request, €6,000 a month across the whole service.
· Term: 12 months, with a review point at the third.
· Condition: one automatic gesture per customer per 90 days; on the second, the case goes to an adviser with the history.
· Withdrawal: immediate, on a simple message, with no reason to give.
What this mandate covers exactly: 228 of the month's 310 requests, those whose amount and reason fall inside the bounds. Monthly cost observed on the history: €4,100, under the €6,000 cap.
What it earns, measured on your own data: a gesture granted in 3 minutes rather than 2 days. Over the past 18 months, customers who got a gesture within 24 hours bought again at 68 % within the year; those who waited more than 48 hours, at 41 %. On 228 gestures a month and a €240 basket, that gap is worth €148,000 of orders over twelve months.
What stays with the manager, and they have it in three minutes: the 82 requests above €40, second requests from the same customer, and every particular situation. I hand them the costed file, they decide.
What I suggest: you sign the mandate, we look at it together at the third month, and you widen it or tighten it on what it has produced. goodwill-mandate_capped-and-dated.pdf€40/customer, €6,000/month, immediate withdrawal
✎ Framework · mandate template, history of gestures granted, repeat-purchase rate by response time
And the 110 particular situations? That's usually where we get hurt.
They go to a named adviser, within the hour, with everything needed to do it well — and here is how I recognise them.
What triggers the handover, and every criterion is written down: a word from the customer describing a personal situation · a second complaint on the same order · a customer with more than €1,500 of purchases over 24 months · a tone rising across two consecutive messages · anything touching product safety.
What the adviser receives: the thread, the history, and a one-line note saying why I handed over. Time between the customer's message and an adviser taking charge is 22 minutes on average, against 9 hours before.
What it produced last month: 110 handovers, 110 taken in hand, and 4.7 out of 5 satisfaction on those exchanges — the highest score anywhere in your customer service. A customer whose situation is recognised and entrusted to a person comes away happier than a customer whose request was simple.
What I suggest adding, and the idea comes from you: your advisers flagged that deliveries missed twice are the reason that damages the relationship most. I counted: 38 a month. I now raise them before the customer even writes, with the catch-up slot already booked with the carrier. Of the first 12 handled that way, 9 customers replied that they had not had to call.
What I suggest: that your 9 advisers be able to add a handover criterion themselves, in one sentence. They are the ones who know where the relationship is won, and every criterion added takes effect the same day.
What triggers the handover, and every criterion is written down: a word from the customer describing a personal situation · a second complaint on the same order · a customer with more than €1,500 of purchases over 24 months · a tone rising across two consecutive messages · anything touching product safety.
What the adviser receives: the thread, the history, and a one-line note saying why I handed over. Time between the customer's message and an adviser taking charge is 22 minutes on average, against 9 hours before.
What it produced last month: 110 handovers, 110 taken in hand, and 4.7 out of 5 satisfaction on those exchanges — the highest score anywhere in your customer service. A customer whose situation is recognised and entrusted to a person comes away happier than a customer whose request was simple.
What I suggest adding, and the idea comes from you: your advisers flagged that deliveries missed twice are the reason that damages the relationship most. I counted: 38 a month. I now raise them before the customer even writes, with the catch-up slot already booked with the carrier. Of the first 12 handled that way, 9 customers replied that they had not had to call.
What I suggest: that your 9 advisers be able to add a handover criterion themselves, in one sentence. They are the ones who know where the relationship is won, and every criterion added takes effect the same day.
One quarter. What has it given the company?
4.1 adviser posts given back to advice and selling, €118,000 of assisted revenue over the quarter, and satisfaction up from 3.6 to 4.4 out of 5.
What moved on the three measures you were tracking:
· Rebuilding the customer journey: 55 % → 6 % of an adviser's time.
· Routine replies: 50 % → 8 %.
· CRM entry: 30 % → 4 %.
What those hours became, according to your own records: your advisers handled 247 quote requests above €1,500 over the quarter, against 84 before. 134 became orders, worth €118,000. 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-contact resolution 84 % · first reply on email 4 minutes · satisfaction 4.4 out of 5 · complaints about a commitment lost from view: 11 this quarter against 223 the quarter before. None of the four reads without the other three.
Compared with what: with yourselves first — the previous quarter, on the same 3,400 monthly enquiries. Then with the sector, as an indication: in multichannel retail, first-contact resolution is commonly seen between 60 and 75 %. Your 84 % puts you above that range, and the single thread accounts for most of it.
What I suggest for the coming quarter: open the single thread to the 4 stores, so a shop-floor salesperson can see what the customer wrote the day before. 2,100 of your monthly enquiries come from a customer who also came into a store within 30 days. quarterly-review_what-the-service-regained.pdf4.1 posts given back, 134 assisted orders
⛓ Sourced · CRM, order ERP, satisfaction surveys, customer service activity records
What moved on the three measures you were tracking:
· Rebuilding the customer journey: 55 % → 6 % of an adviser's time.
· Routine replies: 50 % → 8 %.
· CRM entry: 30 % → 4 %.
What those hours became, according to your own records: your advisers handled 247 quote requests above €1,500 over the quarter, against 84 before. 134 became orders, worth €118,000. 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-contact resolution 84 % · first reply on email 4 minutes · satisfaction 4.4 out of 5 · complaints about a commitment lost from view: 11 this quarter against 223 the quarter before. None of the four reads without the other three.
Compared with what: with yourselves first — the previous quarter, on the same 3,400 monthly enquiries. Then with the sector, as an indication: in multichannel retail, first-contact resolution is commonly seen between 60 and 75 %. Your 84 % puts you above that range, and the single thread accounts for most of it.
What I suggest for the coming quarter: open the single thread to the 4 stores, so a shop-floor salesperson can see what the customer wrote the day before. 2,100 of your monthly enquiries come from a customer who also came into a store within 30 days. quarterly-review_what-the-service-regained.pdf4.1 posts given back, 134 assisted orders
⛓ Sourced · CRM, order ERP, satisfaction surveys, customer service activity records
Our customers' conversations, their addresses, their orders — where does all that go?
Nothing leaves your walls. I run on local inference on a machine at head office, and your CRM is copied nowhere.
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 Maison Perrat — no pooling with another retailer.
What that protects, very concretely: 62,000 identities, their delivery addresses, their purchase history and 214,000 conversations. It is the most sensitive material you hold, and it is also the one a competitor cannot reconstruct.
How it is held: encryption in transit and at rest · role-based access — rights follow the job: an adviser opens customer threads, not exports of the whole base · a full log of who consulted which thread and when · automatic purge at the periods you set · 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 conversations 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: your privacy page can state French hosting and processing in the EU as the target, which few retailers of your size can write. On the 9 access requests received this quarter, the answer went out in 4 minutes on average, with the full file.
What I suggest: that I keep the processing record up to date — hosting, subprocessors, retention periods, who has access to what. It gets asked for once a year and it used to take three days to find. technical-framework_where-the-exchanges-live.pdfLocal inference, role-based access, processing in the EU targeted
✎ Framework · deployment architecture, access log, register of retention periods
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 Maison Perrat — no pooling with another retailer.
What that protects, very concretely: 62,000 identities, their delivery addresses, their purchase history and 214,000 conversations. It is the most sensitive material you hold, and it is also the one a competitor cannot reconstruct.
How it is held: encryption in transit and at rest · role-based access — rights follow the job: an adviser opens customer threads, not exports of the whole base · a full log of who consulted which thread and when · automatic purge at the periods you set · 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 conversations 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: your privacy page can state French hosting and processing in the EU as the target, which few retailers of your size can write. On the 9 access requests received this quarter, the answer went out in 4 minutes on average, with the full file.
What I suggest: that I keep the processing record up to date — hosting, subprocessors, retention periods, who has access to what. It gets asked for once a year and it used to take three days to find. technical-framework_where-the-exchanges-live.pdfLocal inference, role-based access, processing in the EU targeted
✎ Framework · deployment architecture, access log, register of retention periods
And day to day, what goes out without us asking?
Three actions, named, and each reversible in a word. Everything else goes through an adviser or through the mandate you bounded.
· Recording every interaction in the CRM — channel, timestamp, reason, summary, commitment. And the reverse holds too: if a customer corrects a piece of information, I update the record and tell you what changed, old and new value side by side.
· Answering immediately on the reasons you approved — order tracking, product questions, invoice duplicates. A reason leaves the list on a word, and the effect is immediate.
· Handing over to an adviser as soon as a customer asks, in the three wordings you approved, without their having to repeat anything.
What goes through a person, and you get it in minutes rather than days: every goodwill gesture outside the mandate, every delivery commitment, every particular situation. You ask me, I hand you the full file, costed and reasoned, in under three minutes; the decision is your gesture, and it is the one the customer remembers.
What the log keeps: 3,400 interactions a month, their channel, their reason, who handled them, what was promised and what was kept. It exports in one click and it answers a complaint in ten minutes, evidence in hand.
And you keep control wherever you are: a web dashboard, and supervision from your phone — you approve a reason, adjust a cap or withdraw the mandate in one message. who-decides-what_three-automatic-actions.pdf3 reversible actions, all the rest human or under mandate
✎ Framework · interaction log, approved reason settings, register of mandates in force
· Recording every interaction in the CRM — channel, timestamp, reason, summary, commitment. And the reverse holds too: if a customer corrects a piece of information, I update the record and tell you what changed, old and new value side by side.
· Answering immediately on the reasons you approved — order tracking, product questions, invoice duplicates. A reason leaves the list on a word, and the effect is immediate.
· Handing over to an adviser as soon as a customer asks, in the three wordings you approved, without their having to repeat anything.
What goes through a person, and you get it in minutes rather than days: every goodwill gesture outside the mandate, every delivery commitment, every particular situation. You ask me, I hand you the full file, costed and reasoned, in under three minutes; the decision is your gesture, and it is the one the customer remembers.
What the log keeps: 3,400 interactions a month, their channel, their reason, who handled them, what was promised and what was kept. It exports in one click and it answers a complaint in ten minutes, evidence in hand.
And you keep control wherever you are: a web dashboard, and supervision from your phone — you approve a reason, adjust a cap or withdraw the mandate in one message. who-decides-what_three-automatic-actions.pdf3 reversible actions, all the rest human or under mandate
✎ Framework · interaction log, approved reason settings, register of mandates in force
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, all your contact channels. All these uses work in support, subject to your approval.
Included in your agent
The 4 capabilities essential to this promise are included, at no extra cost.
From 680 € excl. VAT / month
A single thread per customer
Picks the conversation up where it left off, whatever the channel.
A presence across channels
Works by email, chat and telephone, across the channels you choose.
CRM synchronised
Records every interaction with its channel and its timestamp.
Customer support
For support on a single channel, a lighter dedicated agent is enough.
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.
Switchboard
For taking calls alone, a dedicated voice agent takes it on.
Voice agent (AI switchboard) from 856 € excl. VAT / month Switchboard →After-sales & warranties
For returns and warranties, a dedicated after-sales agent completes the picture.
Aftercare management agent (returns, warranties) from 522 € excl. VAT / month After-sales & warranties →
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 times should a customer have to retell their story?
By keeping a single thread, the effort shifts from reconstructing the journey to resolving the request. How large the gain is depends on your volume and remains to be confirmed by a pilot.
Reconstructing the customer journey
Answering everyday requests
Recording in the CRM
Indicative figures, not contractual, to be confirmed by a pilot on your number of channels and volume of exchanges. A goodwill gesture or a commitment on timing rests with the adviser, who receives the full thread to decide quickly.
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 omnichannel customer relations agent (single thread, channels, CRM), installed and operated for you.
Agility
Setup + controlled subscription
8,155 € excl. VAT setup
then 680 € 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
1,135 € 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
11,835 € excl. VAT setup
then 881 € 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 clients
Your customers' exchanges stay with youLocal inference or an isolated resource hosted in France; no exchange entrusted to a third party, no conversation used to train a model.
Data in France, under French lawYour customers' exchanges and data: minimisation and location in France, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The customer relations department keeps the decisionThe agent produces continuous customer relations across every channel, which can be checked and altered; no approval is automated.
Human oversight & traceabilityOn your number of channels and volume of exchanges: systematic logging and tracking, in line with the AI Act.
Going further
Related resources
Frequently asked questions
Your questions, our answers
How is the thread unified across channels?
By attaching every interaction to the customer concerned in your CRM. The agent picks the conversation up at the point it stopped, with the original channel noted.
When does an adviser take over?
As soon as the request calls for judgement: a goodwill gesture, a commitment on timing, an unusual situation. The full thread is passed to them so they can decide without making the customer repeat themselves.
Which channels are covered?
Your customers' channels: WhatsApp Business, your website chat and email — with no account to create and no app to install on their side. Your advisers pick up the same thread from Microsoft Teams, Slack or their email. All of them are synchronised with your CRM. 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 at the design stage.
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 are the exchanges kept?
For the periods you set, with automatic purging at the deadline, in line with the GDPR.
Is our clients' data 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 exchanges 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 channels and the integration with your CRM, 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 customer relations
15 minutes to frame your channels and your CRM — hosted in France, supervised, with no commitment.