● B2B offer — Sales & prospecting
Qualifying enquiries: your criteria, applied to every email
A sales enquiry handled the next day is often worth less than one handled within the hour. Your agent reads the incoming messages, draws out the subject, the need expressed and the context, then prioritises them against the criteria you have set — stating what the ranking rests on. Hosted in France: your prospects' enquiries stay with you. The salesperson decides who to call back, and when.
Hosted in France
Prospects never exposed
GDPR & AI Act: governed deployment
Human oversight
Updated on
Deployed in a few weeks
Sales qualification · hosted in France
Qualify the enquiries received this morning.
Enquiries qualified against your criteria: nature of the need, sector, size of organisation and any deadline mentioned.
For each, the ranking obtained and the element of the message that supports it.
Two enquiries mention a close deadline: they are put at the top.
✎ Action · reasoned prioritisation, changeable
For each, the ranking obtained and the element of the message that supports it.
Two enquiries mention a close deadline: they are put at the top.
✎ Action · reasoned prioritisation, changeable
The second one — have we had contact before?
An earlier exchange appears in your CRM with the same organisation; I attach it to the enquiry so you have the history in front of you.
Judging the opportunity is yours: it depends on what you know about the account.
🔗 Sourced · history attached from your CRM
Judging the opportunity is yours: it depends on what you know about the account.
🔗 Sourced · history attached from your CRM
Local inference · no data outside the EU
Enquiries 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 salesperson decidesThe agent prepares, never rules
✦ In brief
A Blue Lemon Agent qualification agent reads incoming sales enquiries, draws out the subject and the need expressed, attaches the known history and prioritises them against your criteria — stating the element of the message that supports each ranking. It does not reply to the prospect and makes no commitment. It runs on local inference or is hosted in France: your prospects' enquiries and their contact details 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
sales 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 volume of incoming enquiries is confirmed by a pilot.
The context
What does an AI agent bring to your incoming enquiries?
An enquiry qualified the moment it arrives is an enquiry handled at the right time by the right person. That is often what separates a deal won from a deal lost.
! The issue
The value of a sales enquiry rests heavily on how fast the first reply comes. But you still need to know which one to handle first. The agent reads each message as it arrives, draws out the need, attaches the known history and applies your priority criteria — so the salesperson opens the day on a queue that is already ordered and reasoned.
✓ Our answer
You set the criteria, the agent applies them and gives its reasons: every ranking refers back to the element of the message that produced it, and stays changeable in one move. It never replies to the prospect and makes no sales commitment — the relationship remains your salesperson's. Local inference or an isolated resource hosted in France: your prospects are known to no third party.
The decisive point
Your prospects' enquiries and their contact details: sovereignty & compliance
Your prospects' enquiries contain their plans and their contact details. Here is how the architecture of our agents protects them.
Local inference
The agent can run on a machine belonging to your organisation: no prospect enquiry leaves the network, no contact details pass through a public cloud.
Hosting in France
Otherwise, a dedicated and isolated resource hosted in France, under French law — your incoming enquiries and your customer history: processing and access within the European Union targeted by the architecture.
Reduced extraterritorial exposure
For your prospects' enquiries and their contact details, 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 prospect portfolio.
Reasoned and changeable ranking
Every prioritisation refers back to the element of the message that supports it and stays changeable; encryption, role-based access and logging of every qualification.
AI Act: governed deployment
The agent is strictly in support; no reply is sent to the prospect and no sales commitment is made; 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 enquiries tied to a confidential transaction or a restricted tender, 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
4 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.
Sales qualification · hosted in France
Four flags this morning, and the first is about my own rejects.
· Four hundred and twelve emails were discarded as unqualified. Nobody ever reread them. I had a sample reread, and I will say what it contained.
· Three signals I was using predict nothing. Two of them described people, not their request.
· Twenty-three messages came from existing clients, filed as cold prospects.
· Eleven follow-ups went out on messages that said no. morning-watch_4-flags.pdf412 rejects never reread · 3 worthless signals
⛓ Source · 2,840 incoming emails, qualification log
· Four hundred and twelve emails were discarded as unqualified. Nobody ever reread them. I had a sample reread, and I will say what it contained.
· Three signals I was using predict nothing. Two of them described people, not their request.
· Twenty-three messages came from existing clients, filed as cold prospects.
· Eleven follow-ups went out on messages that said no. morning-watch_4-flags.pdf412 rejects never reread · 3 worthless signals
⛓ Source · 2,840 incoming emails, qualification log
What was in the rejects?
Of a hundred messages reread at random, seven were genuine requests. That is my error rate, and it was measured nowhere.
What I record: 412 emails filed as "unqualified" this quarter. Not one was opened by a human. That is how sorting normally works — and it is what makes it unverifiable.
What the rereading showed: of 100 messages drawn at random, 7 were real requests. Three from an existing client, two partnership approaches, two from prospects whose message was very short.
Why that figure matters most: a qualification rate is easy to measure — you can see what gets through. The reject error rate can only be measured if somebody rereads what was thrown away, and nobody has any reason to. It is the one indicator an automatic sort never produces on its own.
What I now do: a sample of the rejects is reread every week, and the false-negative rate appears in the summary, beside the volume handled.
What becomes of a discarded message: it is kept, dated and rereadable. A deleted reject is an error rate nobody will ever measure — it is that keeping which made the 7 % visible. 412-rejects_7-of-100.pdfThe reject error rate exists only if somebody rereads
⛓ Source · 412 rejects, 100 reread, 7 genuine requests
What I record: 412 emails filed as "unqualified" this quarter. Not one was opened by a human. That is how sorting normally works — and it is what makes it unverifiable.
What the rereading showed: of 100 messages drawn at random, 7 were real requests. Three from an existing client, two partnership approaches, two from prospects whose message was very short.
Why that figure matters most: a qualification rate is easy to measure — you can see what gets through. The reject error rate can only be measured if somebody rereads what was thrown away, and nobody has any reason to. It is the one indicator an automatic sort never produces on its own.
What I now do: a sample of the rejects is reread every week, and the false-negative rate appears in the summary, beside the volume handled.
What becomes of a discarded message: it is kept, dated and rereadable. A deleted reject is an error rate nobody will ever measure — it is that keeping which made the 7 % visible. 412-rejects_7-of-100.pdfThe reject error rate exists only if somebody rereads
⛓ Source · 412 rejects, 100 reread, 7 genuine requests
How do you qualify, and who do you tell?
On what the message asks for. Never on what its sender appears to be.
What I look at: the object of the request, its precision, whether a need is expressed, any mention of a deadline or budget where present, and whether the sender is already known to your systems.
Routing follows who can answer: a qualified request goes to the salesperson for that area, with the full message; a message from an existing client to their usual contact, never into the prospect flow; a message I cannot classify to a human, not to the rejects; an explicit refusal is taken out of every sequence, immediately.
With a weekly summary: volume qualified, false-negative rate measured on a sample, unclassifiable messages, and refusals detected.
What this morning already gives back: 7 real requests per 100 rejects re-read — across 412 set aside, that is some thirty pieces of business already in your building that nobody had opened. From tomorrow: 23 existing clients sent back to their usual contact instead of the prospect flow, 11 chasers that no longer go out to people who said no, 3 worthless signals dropped from the sort, and a qualified request reaching the right salesperson within the hour rather than the next day. The access is yours: I read your mailboxes and your systems, opened by role, logged, withdrawn with a word — and I go looking for nothing outside them about whoever writes to you. I qualify the request, not the person: every classification carries the sentence of the message behind it, so it can be re-read, challenged and corrected — and my error rate is measured every week on a sample, which no sorting produces by itself. As for replying, that is a mandate, not a prohibition: you set the cases, the template and the cap, you date it, you withdraw it with a word — and the acknowledgement goes out in minutes instead of two days. The next step is ready: the quarter's 412 rejects are kept and can be re-sorted on your new criteria; give me the go-ahead and I will hand back the list of recoverable requests.
✎ Framework · no scoring of people · no outside research on the sender
What I look at: the object of the request, its precision, whether a need is expressed, any mention of a deadline or budget where present, and whether the sender is already known to your systems.
Routing follows who can answer: a qualified request goes to the salesperson for that area, with the full message; a message from an existing client to their usual contact, never into the prospect flow; a message I cannot classify to a human, not to the rejects; an explicit refusal is taken out of every sequence, immediately.
With a weekly summary: volume qualified, false-negative rate measured on a sample, unclassifiable messages, and refusals detected.
What this morning already gives back: 7 real requests per 100 rejects re-read — across 412 set aside, that is some thirty pieces of business already in your building that nobody had opened. From tomorrow: 23 existing clients sent back to their usual contact instead of the prospect flow, 11 chasers that no longer go out to people who said no, 3 worthless signals dropped from the sort, and a qualified request reaching the right salesperson within the hour rather than the next day. The access is yours: I read your mailboxes and your systems, opened by role, logged, withdrawn with a word — and I go looking for nothing outside them about whoever writes to you. I qualify the request, not the person: every classification carries the sentence of the message behind it, so it can be re-read, challenged and corrected — and my error rate is measured every week on a sample, which no sorting produces by itself. As for replying, that is a mandate, not a prohibition: you set the cases, the template and the cap, you date it, you withdraw it with a word — and the acknowledgement goes out in minutes instead of two days. The next step is ready: the quarter's 412 rejects are kept and can be re-sorted on your new criteria; give me the go-ahead and I will hand back the list of recoverable requests.
✎ Framework · no scoring of people · no outside research on the sender
Which signals did you remove, and why?
The email domain, the quality of the writing, the time of sending. All three worked a little. All three described people.
The domain: a personal address rather than a company one. A real but weak correlation — and it discards freelancers, very small outfits, and people writing from home in the evening.
The quality of the writing: mistakes, awkward phrasing, a very short message. It is the most effective and the least acceptable of the three. It discards non-native speakers, people in a hurry, those writing from a phone, and those not at ease in writing — none of whom is a bad prospect.
The time of sending: night-time messages converted less. They mostly describe people who work at night, or in another time zone.
What it cost: I checked. Without those three signals my false-positive rate rises by 4 points — four requests in a hundred passed to a salesperson for nothing.
What it avoided: of the 412 rejects, 129 were rejects only because of one of those three signals. Four points of noise against a hundred and twenty-nine messages discarded for what their author appears to be: the arithmetic is not close. 3-signals-removed_129-rejects-avoided.pdfWriting quality discards people, not bad prospects
⛓ Source · 3 signals removed, +4 points of false positives, 129 rejects avoided
The domain: a personal address rather than a company one. A real but weak correlation — and it discards freelancers, very small outfits, and people writing from home in the evening.
The quality of the writing: mistakes, awkward phrasing, a very short message. It is the most effective and the least acceptable of the three. It discards non-native speakers, people in a hurry, those writing from a phone, and those not at ease in writing — none of whom is a bad prospect.
The time of sending: night-time messages converted less. They mostly describe people who work at night, or in another time zone.
What it cost: I checked. Without those three signals my false-positive rate rises by 4 points — four requests in a hundred passed to a salesperson for nothing.
What it avoided: of the 412 rejects, 129 were rejects only because of one of those three signals. Four points of noise against a hundred and twenty-nine messages discarded for what their author appears to be: the arithmetic is not close. 3-signals-removed_129-rejects-avoided.pdfWriting quality discards people, not bad prospects
⛓ Source · 3 signals removed, +4 points of false positives, 129 rejects avoided
And what do you look at instead?
What is written, and only what is written.
The four elements kept: a need expressed — the person says what they are looking for — ; enough precision for a salesperson to reply with something other than a question; a mention of a deadline or a constraint where there is one; and the link with your systems — client, past contact, neither.
What I do when those are missing: I do not file it as a reject. I file it as "imprecise request", and those messages go to a human with a note of what is missing. Of 2,840 emails, 318 fall in that category.
Why that category rather than a reject: because an imprecise message is not a bad prospect, it is a first sentence. Of those 318, 94 led to an exchange, and 21 to a deal.
What I score, and what I leave alone: I score the request — its nature, its precision, the deadline it mentions, the organisation's history — not the requester. So the salesperson receives a nature of need and the four elements that establish it, not a forecast about somebody they have not read yet. What that changes, measured across the 2,840: the 21 deals that came out of "imprecise requests" all came from a category a propensity score would have left at the bottom of the queue. 4-elements_318-imprecise-requests.pdfAn imprecise message is a first sentence, not a bad prospect
⛓ Source · 318 imprecise requests, 94 exchanges, 21 deals
The four elements kept: a need expressed — the person says what they are looking for — ; enough precision for a salesperson to reply with something other than a question; a mention of a deadline or a constraint where there is one; and the link with your systems — client, past contact, neither.
What I do when those are missing: I do not file it as a reject. I file it as "imprecise request", and those messages go to a human with a note of what is missing. Of 2,840 emails, 318 fall in that category.
Why that category rather than a reject: because an imprecise message is not a bad prospect, it is a first sentence. Of those 318, 94 led to an exchange, and 21 to a deal.
What I score, and what I leave alone: I score the request — its nature, its precision, the deadline it mentions, the organisation's history — not the requester. So the salesperson receives a nature of need and the four elements that establish it, not a forecast about somebody they have not read yet. What that changes, measured across the 2,840: the 21 deals that came out of "imprecise requests" all came from a category a propensity score would have left at the bottom of the queue. 4-elements_318-imprecise-requests.pdfAn imprecise message is a first sentence, not a bad prospect
⛓ Source · 318 imprecise requests, 94 exchanges, 21 deals
And the order my salespeople work through them in — where does that come from?
Prioritisation follows your criteria, written by your sales management, and there are four — all drawn from the message, none from the sender.
· The nature of the need — a request for a quote comes ahead of a request for a brochure;
· the sector stated in the message or inferable from the need described;
· the size where the message gives it — volumes, number of sites, headcount quoted;
· the deadline mentioned — "for September" is not "whenever suits you".
Every message carries its rank and the sentence from the message that produced it: "rank 2 — quote request, 4 sites, 'delivery wanted before 15/09'". A rank that quotes the message can be challenged in three seconds, and that is what made your salespeople accept the queue.
What order of arrival was costing, measured over the previous quarter: across 1,240 qualified requests, the 118 carrying a deadline under 30 days were handled in 2 days 4 hours on average. They are handled in 3 hours today, and 19 deals were signed on that shift alone.
The criteria are adjustable, and I show the effect first: raise the weight on deadline and I run the rule back over twelve months and tell you how many requests change rank and how many won deals would have moved up or down.
The rule that is not adjustable: no criterion describes the sender. Not their domain, not their language, not the hour they wrote. Those three worked a little, and they filtered out people rather than poor requests.
✎ Framework · prioritisation by your criteria — four of them, arithmetic visible
· The nature of the need — a request for a quote comes ahead of a request for a brochure;
· the sector stated in the message or inferable from the need described;
· the size where the message gives it — volumes, number of sites, headcount quoted;
· the deadline mentioned — "for September" is not "whenever suits you".
Every message carries its rank and the sentence from the message that produced it: "rank 2 — quote request, 4 sites, 'delivery wanted before 15/09'". A rank that quotes the message can be challenged in three seconds, and that is what made your salespeople accept the queue.
What order of arrival was costing, measured over the previous quarter: across 1,240 qualified requests, the 118 carrying a deadline under 30 days were handled in 2 days 4 hours on average. They are handled in 3 hours today, and 19 deals were signed on that shift alone.
The criteria are adjustable, and I show the effect first: raise the weight on deadline and I run the rule back over twelve months and tell you how many requests change rank and how many won deals would have moved up or down.
The rule that is not adjustable: no criterion describes the sender. Not their domain, not their language, not the hour they wrote. Those three worked a little, and they filtered out people rather than poor requests.
✎ Framework · prioritisation by your criteria — four of them, arithmetic visible
Twenty-three clients treated as cold prospects?
Yes, and one of them was describing an incident on their live contract.
What I record: 23 messages came from addresses already present in your systems as active clients. They entered the prospecting flow because the message did not say so — and because nothing matched the sender's address against the client file.
What that produced: an average reply time of 4 days, against a few hours for an identified client. Three received a presentation of the very service they already use.
What I do: matching against your systems is the very first move, before any qualification. An existing client leaves the prospect flow and goes to their usual contact, whatever the message says.
What I put at the head of the message, and it is what prevents the blunder: "message from an address attached to client X, sender not identified". A known company address does not mean the sender is the usual contact — it may be another department, with a genuinely new request, and the contact opens the message knowing it.
What it changed: the 23 cases became 0. Matching is done on 100% of incoming messages. 23-clients_4-day-delay.pdfA known address does not say who is writing
⛓ Source · 23 clients in the prospect flow, 4 days against a few hours
What I record: 23 messages came from addresses already present in your systems as active clients. They entered the prospecting flow because the message did not say so — and because nothing matched the sender's address against the client file.
What that produced: an average reply time of 4 days, against a few hours for an identified client. Three received a presentation of the very service they already use.
What I do: matching against your systems is the very first move, before any qualification. An existing client leaves the prospect flow and goes to their usual contact, whatever the message says.
What I put at the head of the message, and it is what prevents the blunder: "message from an address attached to client X, sender not identified". A known company address does not mean the sender is the usual contact — it may be another department, with a genuinely new request, and the contact opens the message knowing it.
What it changed: the 23 cases became 0. Matching is done on 100% of incoming messages. 23-clients_4-day-delay.pdfA known address does not say who is writing
⛓ Source · 23 clients in the prospect flow, 4 days against a few hours
And the follow-ups sent on a "no"?
Eleven times, a sequence carried on after a refusal written in plain words.
What I record: 11 messages contained an explicit refusal — "thank you, this is not current for us", "do not contact me again" — and the follow-up sequence ran its course. In two cases as far as the third follow-up.
Why it happens: because a reply to a sequence is treated as proof of interest — the message was opened, there was a reply, the engagement indicator rises. The content itself is not read.
What I do: I detect the refusal, I take the person out of every sequence immediately, and I flag it — because it is the only case where I act without being asked.
What I also do, and it matters as much: I record the refusal durably, so that a later campaign does not put them back on a list. A refusal forgotten six months later is a refusal not respected.
What I record, word for word: the refusal, its date, and the exact proposal it concerned. They said no at a precise moment, to a precise offer — and it is that form which holds: not interested in a score would attribute to them a disposition they never expressed, and a score, unlike a sentence, cannot be contested. 11-refusals_2-to-the-third.pdfA refusal forgotten six months later is a refusal not respected
⛓ Source · 11 explicit refusals, 2 as far as the third follow-up
What I record: 11 messages contained an explicit refusal — "thank you, this is not current for us", "do not contact me again" — and the follow-up sequence ran its course. In two cases as far as the third follow-up.
Why it happens: because a reply to a sequence is treated as proof of interest — the message was opened, there was a reply, the engagement indicator rises. The content itself is not read.
What I do: I detect the refusal, I take the person out of every sequence immediately, and I flag it — because it is the only case where I act without being asked.
What I also do, and it matters as much: I record the refusal durably, so that a later campaign does not put them back on a list. A refusal forgotten six months later is a refusal not respected.
What I record, word for word: the refusal, its date, and the exact proposal it concerned. They said no at a precise moment, to a precise offer — and it is that form which holds: not interested in a score would attribute to them a disposition they never expressed, and a score, unlike a sentence, cannot be contested. 11-refusals_2-to-the-third.pdfA refusal forgotten six months later is a refusal not respected
⛓ Source · 11 explicit refusals, 2 as far as the third follow-up
How do you know a message comes from somebody we already deal with?
By attaching the history: every incoming message is matched against your CRM before being qualified, and the match is to the organisation, not to the person.
What the match rests on, in this order: the exact address if it is already in the CRM, the organisation named in the message or signature, the mail domain matched against existing accounts, and a contract or order reference where the message quotes one. Every match shows which of the four applied — a silent match never gets corrected.
Across the quarter's 1,240 requests: 318 attach to an organisation already known. What the salesperson receives with the message: the previous exchanges and their dates, quotes issued and what became of them, live contracts, and the last person in the firm who wrote to that organisation. A salesperson who calls back without knowing a colleague wrote three days earlier loses the deal for their own company.
What attaching the history caught, and it was the costliest point: the 23 existing customers treated as cold prospects — one of whom was describing an incident on a live contract. An existing customer now leaves the sales flow and goes to the team that handles their contract, with the history attached; they no longer receive a prospecting sequence.
The figure that does not suit me: 14 matches turned out to be wrong over the quarter, all by mail domain — two separate subsidiaries of one group, and two sole traders on the same hosting provider. Domain alone no longer attaches: it proposes, and a second clue is required — name quoted or contract reference. 1 error across the following 402 matches, and the 14 organisations were corrected in the CRM.
⛓ Sourced · CRM — accounts, contacts, exchanges and live contracts
What the match rests on, in this order: the exact address if it is already in the CRM, the organisation named in the message or signature, the mail domain matched against existing accounts, and a contract or order reference where the message quotes one. Every match shows which of the four applied — a silent match never gets corrected.
Across the quarter's 1,240 requests: 318 attach to an organisation already known. What the salesperson receives with the message: the previous exchanges and their dates, quotes issued and what became of them, live contracts, and the last person in the firm who wrote to that organisation. A salesperson who calls back without knowing a colleague wrote three days earlier loses the deal for their own company.
What attaching the history caught, and it was the costliest point: the 23 existing customers treated as cold prospects — one of whom was describing an incident on a live contract. An existing customer now leaves the sales flow and goes to the team that handles their contract, with the history attached; they no longer receive a prospecting sequence.
The figure that does not suit me: 14 matches turned out to be wrong over the quarter, all by mail domain — two separate subsidiaries of one group, and two sole traders on the same hosting provider. Domain alone no longer attaches: it proposes, and a second clue is required — name quoted or contract reference. 1 error across the following 402 matches, and the 14 organisations were corrected in the CRM.
⛓ Sourced · CRM — accounts, contacts, exchanges and live contracts
Can you reply to prospects directly?
Yes on everything that is not commercial — and that covers more messages than you would think.
What I send alone as soon as you open it: the acknowledgement with the real response time, the request for detail when a message is too vague to route, redirection to the right department when it is not a commercial enquiry, and answers to documented factual questions — opening hours, address, service area, documents to provide.
What stays written by somebody: the first properly commercial sentence — the one presenting an offer, a price, an availability. A prospect receiving a commercial reply written by an agent recognises it, and draws a conclusion about what their enquiry is worth to you.
What it changes: your enquiries waited on average eleven hours for any sign at all. They now get an acknowledgement in three minutes that states a true lead time.
What I check before any send, and it is not configurable: that the sender is not already a customer — 23 were, and were treated as cold prospects, one of them describing an incident on their live contract — and that they have not expressed a refusal. Eleven sequences had continued after a "not right now" written in plain words. Article 21 of the GDPR makes that second point absolute. direct-reply_by-nature.pdfWhat goes alone · the commercial sentence · the 2 non-configurable checks
⛓ Source · eleven hours down to three minutes, 23 customers and 11 refusals caught
What I send alone as soon as you open it: the acknowledgement with the real response time, the request for detail when a message is too vague to route, redirection to the right department when it is not a commercial enquiry, and answers to documented factual questions — opening hours, address, service area, documents to provide.
What stays written by somebody: the first properly commercial sentence — the one presenting an offer, a price, an availability. A prospect receiving a commercial reply written by an agent recognises it, and draws a conclusion about what their enquiry is worth to you.
What it changes: your enquiries waited on average eleven hours for any sign at all. They now get an acknowledgement in three minutes that states a true lead time.
What I check before any send, and it is not configurable: that the sender is not already a customer — 23 were, and were treated as cold prospects, one of them describing an incident on their live contract — and that they have not expressed a refusal. Eleven sequences had continued after a "not right now" written in plain words. Article 21 of the GDPR makes that second point absolute. direct-reply_by-nature.pdfWhat goes alone · the commercial sentence · the 2 non-configurable checks
⛓ Source · eleven hours down to three minutes, 23 customers and 11 refusals caught
What do you keep?
The messages and their qualification — and above all my own errors, which were measured nowhere.
What is kept: the message, its qualification and the reason, the match against your systems, the rejects and their re-reading, the refusals expressed with their date, and the time between arrival and the first human reply.
The figure I publish against myself: 412 emails set aside as unqualified, which nobody had ever re-read. Of a hundred drawn at random, seven were genuine enquiries. That is my error rate, and it was measured nowhere — an automatic sort whose rejects nobody re-reads is a sort whose cost nobody knows.
What it changed in how I work: a fixed share of rejects is now re-read every week, and the rate published with it.
What I look at to qualify, and what I removed: I qualify on what the message asks — need expressed, precision, deadline, means of contact. I removed the address's domain name, the quality of the language and the time of sending. All three worked a little. All three described people rather than their enquiry — and a personal address or an awkward sentence says nothing about a need. what-you-keep_qualification.pdf5 items kept · the error rate published
⛓ Source · 7 genuine enquiries per 100 rejects re-read, 3 signals removed
What is kept: the message, its qualification and the reason, the match against your systems, the rejects and their re-reading, the refusals expressed with their date, and the time between arrival and the first human reply.
The figure I publish against myself: 412 emails set aside as unqualified, which nobody had ever re-read. Of a hundred drawn at random, seven were genuine enquiries. That is my error rate, and it was measured nowhere — an automatic sort whose rejects nobody re-reads is a sort whose cost nobody knows.
What it changed in how I work: a fixed share of rejects is now re-read every week, and the rate published with it.
What I look at to qualify, and what I removed: I qualify on what the message asks — need expressed, precision, deadline, means of contact. I removed the address's domain name, the quality of the language and the time of sending. All three worked a little. All three described people rather than their enquiry — and a personal address or an awkward sentence says nothing about a need. what-you-keep_qualification.pdf5 items kept · the error rate published
⛓ Source · 7 genuine enquiries per 100 rejects re-read, 3 signals removed
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 acts in handling incoming enquiries. 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 504 € excl. VAT / month
Reading and qualifying
Draws out the subject, the need expressed and the context of each enquiry.
Prioritising on your criteria
Ranks by nature of the need, sector, size and any deadline mentioned.
Attaching the history
Finds in your CRM the earlier exchanges with the same organisation.
Controls and safeguards
These 4 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
Prioritise opportunities on explicit criteria
Personalise without breaching consent or prospecting rules
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.
Prepare the next action and write back to the CRM without loss
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.
Qualification and appointment booking
To go as far as setting the appointment, a dedicated agent carries on the chain.
Lead qualification / appointment booking agent from 602 € excl. VAT / month Qualification & appointments →Quote generation
Once the need is framed, a dedicated agent builds the quotation from your catalogue.
Simple quote generation agent from 548 € excl. VAT / month Quotes →
Does your need fall outside this?
Book the free audit →
Build your agent
In 15 minutes we identify the most relevant agent — without oversizing the project.
The gain
How much responsiveness can a sales team gain?
By qualifying and prioritising on arrival, the delay before the first reply to the most promising enquiries comes down. How large the gain is depends on your volume and remains to be confirmed by a pilot.
Reading and sorting the enquiries
Finding the customer history
Choosing which enquiry to handle
Indicative figures, not contractual, to be confirmed by a pilot on your volume of incoming enquiries. The agent never replies to the prospect and makes no commitment: the choice to call back, and what is promised, belong to the salesperson.
How it works
The stages of your AI agent project
1
Audit & scoping
15 minutes to target the use case with the best return.
2
Quote or direct sign-up
A catalogue offer is bought online; a specific need gets a costed quote.
3
Design
We design the agent and its guardrails.
4
Integration & testing
We connect your tools to the agent, which is itself hosted in France.
5
Rollout
Going live and training your team.
6
Operation
Continuous supervision and improvement.
Pricing
One package, one agent
A qualification agent (reading, prioritisation, CRM attachment), installed and operated for you.
Agility
Setup + controlled subscription
4,630 € excl. VAT setup
then 504 € 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
764 € 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
7,965 € excl. VAT setup
then 704 € excl. VAT/month · + hardware from 1,058 € (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 portfolio
Your prospects are known to no third partyLocal inference or an isolated resource hosted in France; no prospect enquiry or contact details entrusted to a third party, no data used to train a model.
Data in France, under French lawYour prospects' enquiries and their contact details: minimisation and location in France, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The salesperson keeps the decisionThe agent produces qualified enquiries, which can be checked and altered; no approval is automated.
Human oversight & traceabilityOn your volume of incoming enquiries: systematic logging and tracking, in line with the AI Act.
Frequently asked questions
Your questions, our answers
Does the agent reply to prospects?
No. It qualifies and prioritises; the first reply remains your salesperson's. An automatic reply to a sales enquiry commits the company on a need it has not yet understood.
What is the prioritisation based on?
On the criteria you define — nature of the need, sector, size, deadline. Every ranking states the element of the message that supports it, and stays changeable in one move.
Does the agent judge how valuable the deal is?
It reports what the message contains and attaches the known history. Judging the opportunity depends on what you know about the account and the market: it falls to the salesperson.
Are our prospects' enquiries 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 prospects are not used to train a third-party model.
Does it connect to our CRM?
Yes, to the common CRMs, to find the history and attach the enquiry. The integration is defined at the design stage, with no migration.
How long does it take to deploy this agent?
A few weeks as a rule, depending on your qualification criteria and the CRM integration, after a free audit then a phase of design, integration and testing.
Related use cases
Other agents for business development
Let's talk
Let's size up the potential in your incoming enquiries
15 minutes to frame your priority criteria — hosted in France, supervised, with no commitment.