● B2B offer — Steering & performance
Sales steering: your metrics up to date, the differences flagged
Steering calls for up-to-date figures and differences seen early. Your agent produces your metrics from your CRM, presents them along your dimensions — salesperson, territory, product, period — and flags the differences worth a look, with the data that explains them. Hosted in France: your activity figures stay with you. Sales management interprets and decides on the actions.
Hosted in France
Activity figures protected
GDPR & AI Act: governed deployment
Human oversight
Updated on
Deployed in a few weeks
Sales steering · hosted in France
Give me the position on the month's activity.
Dashboard along your dimensions: by salesperson, by territory, by product and by period.
Every metric refers back to the CRM records it comes from.
Three notable differences against last month are flagged.
🔗 Sourced · CRM records
Every metric refers back to the CRM records it comes from.
Three notable differences against last month are flagged.
🔗 Sourced · CRM records
Which differences, and what should they be put down to?
Two territories markedly up and one product family whose business is shifting to another offer.
The corresponding records are accessible. The interpretation and the actions to decide are yours: they commit your teams.
✎ Support · differences sourced, human interpretation
The corresponding records are accessible. The interpretation and the actions to decide are yours: they commit your teams.
✎ Support · differences sourced, human interpretation
Local inference · no data outside the EU
Indicators 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 sales management decidesThe agent prepares, never rules
✦ In brief
A Blue Lemon Agent sales steering agent produces your metrics from your CRM, presents them along your dimensions of analysis and flags the notable differences, referring back to the records that explain them. It neither comments nor interprets: that reading belongs to management. It runs on local inference or is hosted in France: your activity figures 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
steering 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 dimensions and the richness of your CRM is confirmed by a pilot.
The context
What does an AI agent bring to your sales steering?
Figures available when the question arises, and tied to their source: that is what makes a steering meeting useful.
! The issue
Useful steering calls for up-to-date figures and the ability to trace back to the data. Rebuilding the tables at every deadline takes time, and then the meeting argues about the figures rather than the action. The agent produces them continuously from the CRM, with a link back to the original records for every metric.
✓ Our answer
Sales management arrives at the meeting with figures available, traceable, and differences already spotted. The interpretation stays human: a difference may come from the market, from a team or from a change of offer, and only management has that context. Local inference or an isolated resource hosted in France: your activity figures and the make-up of your portfolio do not leave the company.
The decisive point
Your activity figures and your portfolio: sovereignty & compliance
Your activity figures and your portfolio are strategic information, to be protected from any outside exposure. Here is how.
Local inference
The agent can run on a machine belonging to your organisation: no activity figure and no portfolio data leaves the network.
Hosting in France
Otherwise, a dedicated and isolated resource hosted in France, under French law — your metrics and your sales activity: processing and access within the European Union targeted by the architecture.
Reduced extraterritorial exposure
For your activity figures and your portfolio, 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 dimensions of analysis.
Every metric traces back to its source
Every figure refers back to the CRM records it comes from; encryption, role-based access and logging of the tables produced.
AI Act: governed deployment
The agent is strictly in support; no interpretation is presented as a conclusion and no action is triggered; 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 figures consolidated before publication and strategic accounts, 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 steering · hosted in France
Three flags this morning across your dashboard.
· Revenue for the quarter is up 12 %. Nine points come from a change in the attribution rule, not from extra activity.
· Your "conversion rate" is computed on three different scopes in three tables. Three figures for one question.
· A decline alert fired fourteen times last month. It was opened twice. morning-watch_3-flags.pdf12 % growth, 9 points of it from a rule change
⛓ Source · invoicing entries, the three tables' definitions, alert log
· Revenue for the quarter is up 12 %. Nine points come from a change in the attribution rule, not from extra activity.
· Your "conversion rate" is computed on three different scopes in three tables. Three figures for one question.
· A decline alert fired fourteen times last month. It was opened twice. morning-watch_3-flags.pdf12 % growth, 9 points of it from a rule change
⛓ Source · invoicing entries, the three tables' definitions, alert log
Nine points out of twelve that are not activity?
The figure is right. What changed mid-quarter is what it counts.
What I record: until March, a deal was attributed to a quarter by its signature date. Since April it is attributed by its invoice date. The change is not at fault — it brings the dashboard closer to the accounts.
What it produces: the April quarter received the deals signed in March and invoiced in April, on top of its own. It therefore counts four months of deals. That is not an error: it is a transition bump, and it will be repaid next quarter.
What I do: every indicator carries the date its calculation rule last changed. When a variance falls in a period where the rule moved, I give both figures: old basis and new, over the same period.
What it avoids: a decision taken on 12 % — a hire, a raised target — that turns around next quarter when the bump reverses. On a constant rule, real growth is 3 %. 12-percent_9-of-it-rule-change.pdfA transition bump, which will be repaid next quarter
⛓ Source · two attribution rules, real growth 3 % on a constant rule
What I record: until March, a deal was attributed to a quarter by its signature date. Since April it is attributed by its invoice date. The change is not at fault — it brings the dashboard closer to the accounts.
What it produces: the April quarter received the deals signed in March and invoiced in April, on top of its own. It therefore counts four months of deals. That is not an error: it is a transition bump, and it will be repaid next quarter.
What I do: every indicator carries the date its calculation rule last changed. When a variance falls in a period where the rule moved, I give both figures: old basis and new, over the same period.
What it avoids: a decision taken on 12 % — a hire, a raised target — that turns around next quarter when the bump reverses. On a constant rule, real growth is 3 %. 12-percent_9-of-it-rule-change.pdfA transition bump, which will be repaid next quarter
⛓ Source · two attribution rules, real growth 3 % on a constant rule
What exactly is the agent for?
It holds the definitions, publishes the variances and says what moved. It does not say why.
What it does: computes indicators on a single definition, flags variances against the comparable period, separates what comes from activity from what comes from a rule change, tracks its own forecasts, and counts the alerts nobody opens.
Routing follows who can decide: an activity variance goes to whoever steers the scope concerned; a divergent definition between two tables to whoever owns the indicator, not to both authors; an alert that has become useless to whoever created it; a gap between my forecast and reality to nobody: it is published.
With a monthly summary: variances by scope, rule changes and their costed effect, the error of my own forecasts, alerts fired and alerts read.
Three things to know. I give you everything needed to establish a cause — breakdown, upstream indicators, what is stable — and I name the cause as soon as it is in the data, as with this morning's 9 points of rule change. I produce per-person indicators if you ask for them, with the information notice and works council file. And your past forecasts stay displayed as given: that is precisely what makes the next one credible.
✎ Framework · cause named when established · individual indicators equipped · history intact
What it does: computes indicators on a single definition, flags variances against the comparable period, separates what comes from activity from what comes from a rule change, tracks its own forecasts, and counts the alerts nobody opens.
Routing follows who can decide: an activity variance goes to whoever steers the scope concerned; a divergent definition between two tables to whoever owns the indicator, not to both authors; an alert that has become useless to whoever created it; a gap between my forecast and reality to nobody: it is published.
With a monthly summary: variances by scope, rule changes and their costed effect, the error of my own forecasts, alerts fired and alerts read.
Three things to know. I give you everything needed to establish a cause — breakdown, upstream indicators, what is stable — and I name the cause as soon as it is in the data, as with this morning's 9 points of rule change. I produce per-person indicators if you ask for them, with the information notice and works council file. And your past forecasts stay displayed as given: that is precisely what makes the next one credible.
✎ Framework · cause named when established · individual indicators equipped · history intact
How can one indicator give three figures?
Because it carries the same name in three tables and counts three different things.
What I record: the "conversion rate" is 18 %, 24 % or 31 % depending which table is open. All three calculations are correct. The first counts signed quotes against all inbound contacts; the second against qualified contacts only; the third excludes deals lost to a prospect's silence.
What it produces in a meeting: a discussion about which figure is the right one, lasting twenty minutes and ending with the most favourable being chosen, without the actual subject — why conversion is down — ever being reached.
What I do: one indicator, one definition, one owner. All three calculations stay available, but under three distinct names, each with its scope written out in full beneath the figure.
What is left to management, and it fits in one meeting: saying which of the three is your steering indicator. The three answer different questions, and that choice is a steering one — all three definitions are written, costed over the last eight quarters and set side by side. What is not negotiable, on the other hand: that they share a name.
What it gave: of 34 indicators, 11 had at least two definitions. After the rework, 41 remain, all distinctly named. 11-indicators_3-definitions.pdfTwenty minutes of meeting, and the most favourable figure chosen
⛓ Source · 34 indicators, 11 with multiple definitions, 41 after separation
What I record: the "conversion rate" is 18 %, 24 % or 31 % depending which table is open. All three calculations are correct. The first counts signed quotes against all inbound contacts; the second against qualified contacts only; the third excludes deals lost to a prospect's silence.
What it produces in a meeting: a discussion about which figure is the right one, lasting twenty minutes and ending with the most favourable being chosen, without the actual subject — why conversion is down — ever being reached.
What I do: one indicator, one definition, one owner. All three calculations stay available, but under three distinct names, each with its scope written out in full beneath the figure.
What is left to management, and it fits in one meeting: saying which of the three is your steering indicator. The three answer different questions, and that choice is a steering one — all three definitions are written, costed over the last eight quarters and set side by side. What is not negotiable, on the other hand: that they share a name.
What it gave: of 34 indicators, 11 had at least two definitions. After the rework, 41 remain, all distinctly named. 11-indicators_3-definitions.pdfTwenty minutes of meeting, and the most favourable figure chosen
⛓ Source · 34 indicators, 11 with multiple definitions, 41 after separation
Can you forecast the quarter?
I publish a forecast and, beside it, the error of every previous one.
What I produce: a landing forecast, with a range, and the list of deals that swing it from one end to the other. This quarter, four deals account for 60 % of the gap between the bottom and the top of the range.
What I always publish beside it: my average error over the eight previous quarters — 7 % —, and the two quarters where I was out by more than 15 %. Both times, a large deal slipped from one quarter to the next: an event nothing in my data announced.
What stays frozen, and it is what gives the rest its value: forecasts already published. A forecast corrected afterwards is always right, and it is no longer of any use. The ones I gave stay as they were, dated, beside the actual — eight quarters lined up, which is exactly what tells you what today's forecast is worth, and no off-the-shelf dashboard gives you that.
And what I write at the head of every forecast: that it is one, and not a commitment. A range that becomes a target stops being a forecast — deals then move to fall in the right box, and it is the invoicing calendar that adjusts, not the activity. The commitment, if there is one, is made by sales management and carries another name. average-error-7-percent.pdfA forecast corrected afterwards is always right, and no longer of any use
⛓ Source · average error 7 % over eight quarters, two misses beyond 15 %
What I produce: a landing forecast, with a range, and the list of deals that swing it from one end to the other. This quarter, four deals account for 60 % of the gap between the bottom and the top of the range.
What I always publish beside it: my average error over the eight previous quarters — 7 % —, and the two quarters where I was out by more than 15 %. Both times, a large deal slipped from one quarter to the next: an event nothing in my data announced.
What stays frozen, and it is what gives the rest its value: forecasts already published. A forecast corrected afterwards is always right, and it is no longer of any use. The ones I gave stay as they were, dated, beside the actual — eight quarters lined up, which is exactly what tells you what today's forecast is worth, and no off-the-shelf dashboard gives you that.
And what I write at the head of every forecast: that it is one, and not a commitment. A range that becomes a target stops being a forecast — deals then move to fall in the right box, and it is the invoicing calendar that adjusts, not the activity. The commitment, if there is one, is made by sales management and carries another name. average-error-7-percent.pdfA forecast corrected afterwards is always right, and no longer of any use
⛓ Source · average error 7 % over eight quarters, two misses beyond 15 %
Are too many alerts a problem?
It is the same problem as no alerts, with the appearance of surveillance.
What I record: your decline alert fired fourteen times last month, eleven of them on normal variation in a low-volume scope. It was opened twice.
What that produces: an alert that fires too often stops being read, and the day it concerns something real, it arrives in the same queue as the other thirteen. Unread surveillance is more dangerous than none: it creates the feeling that the subject is covered.
What I do: for each alert I measure its open rate and the number of times it preceded an action. An alert never followed by an action is flagged to whoever created it — with its figures, so they decide.
What I hand back to its author, ready to decide: its open rate, the number of times it preceded an action, and the threshold that would have made it readable — for this one, a threshold that would have fired three times instead of fourteen without missing the single real case. Deletion and raising alike carry a signature and a date: raising a threshold without saying so is the most discreet way to make a problem disappear from a dashboard, and an alert changed in silence is no longer an alert.
What it gave: of 22 alerts, 9 were deleted by their authors, 5 had their threshold raised explicitly and dated, and 2 were judged too rare to be reliable — they had never fired in a year. 22-alerts_9-deleted.pdfUnread surveillance creates the feeling that the subject is covered
⛓ Source · 22 alerts, open rate measured, 9 deleted by their authors
What I record: your decline alert fired fourteen times last month, eleven of them on normal variation in a low-volume scope. It was opened twice.
What that produces: an alert that fires too often stops being read, and the day it concerns something real, it arrives in the same queue as the other thirteen. Unread surveillance is more dangerous than none: it creates the feeling that the subject is covered.
What I do: for each alert I measure its open rate and the number of times it preceded an action. An alert never followed by an action is flagged to whoever created it — with its figures, so they decide.
What I hand back to its author, ready to decide: its open rate, the number of times it preceded an action, and the threshold that would have made it readable — for this one, a threshold that would have fired three times instead of fourteen without missing the single real case. Deletion and raising alike carry a signature and a date: raising a threshold without saying so is the most discreet way to make a problem disappear from a dashboard, and an alert changed in silence is no longer an alert.
What it gave: of 22 alerts, 9 were deleted by their authors, 5 had their threshold raised explicitly and dated, and 2 were judged too rare to be reliable — they had never fired in a year. 22-alerts_9-deleted.pdfUnread surveillance creates the feeling that the subject is covered
⛓ Source · 22 alerts, open rate measured, 9 deleted by their authors
Can you tell us why sales are falling?
I can tell you where, by how much, since when, and what is not falling. Not why.
What I supply: the decline broken down by scope, customer segment and product, the date it starts, what is stable over the same period, and the upstream indicators — inbound contacts, quotes issued, decision time — that are moving or not.
What that settles: a decline where inbound contacts are stable and quotes issued are stable does not come from demand. A decline where inbound contacts fell two months earlier was announced. That is not the cause, but it eliminates hypotheses — and it is what I can do honestly.
Why I stop there: a commercial cause sits with a competitor, in a frozen customer budget, in a discount somebody stopped granting. None of that is in your data, and an explanation produced from correlations would be repeated at board level with the authority of a figure.
What I do flag: when a fall in an indicator coincides with a change in its calculation rule. There, I know. what-i-supply_what-i-do-not-explain.pdfThat is not the cause, but it eliminates hypotheses
⛓ Source · decline broken down, upstream indicators supplied, no cause advanced
What I supply: the decline broken down by scope, customer segment and product, the date it starts, what is stable over the same period, and the upstream indicators — inbound contacts, quotes issued, decision time — that are moving or not.
What that settles: a decline where inbound contacts are stable and quotes issued are stable does not come from demand. A decline where inbound contacts fell two months earlier was announced. That is not the cause, but it eliminates hypotheses — and it is what I can do honestly.
Why I stop there: a commercial cause sits with a competitor, in a frozen customer budget, in a discount somebody stopped granting. None of that is in your data, and an explanation produced from correlations would be repeated at board level with the authority of a figure.
What I do flag: when a fall in an indicator coincides with a change in its calculation rule. There, I know. what-i-supply_what-i-do-not-explain.pdfThat is not the cause, but it eliminates hypotheses
⛓ Source · decline broken down, upstream indicators supplied, no cause advanced
We want the figures per salesperson. Can you do that?
Yes, and I supply the file that installs it properly. Here is the effect to know first.
What I deliver with it: the prior information notice (art. L1222-4 of the Labour Code), the works council consultation file and a proportionate scope — half a day of work you will not have to do. The French DPA restates that an employer has the power to frame and monitor staff activity.
The effect to know, indicator by indicator: an individual conversion rate makes low-odds quotes disappear — the rate rises, the number of deals falls, and the dashboard shows an improvement. An individual revenue figure does not produce that effect.
What I do in both cases: the individual indicator carries the same requirement as the others — written definition, scope, owner, and the date its rule last changed. An individual indicator whose rule moved mid-year is not comparable from quarter to quarter, and that is exactly what makes an annual review contestable.
What that guarantees you: if you open that view, it will be defensible in front of the person concerned. That is the only state in which it has any steering value. opening-individual-indicators.pdfWhat opens · the effect per indicator · defensible to the person
⛓ Source · French DPA, art. L1222-4 · works council file supplied · effect documented per indicator
What I deliver with it: the prior information notice (art. L1222-4 of the Labour Code), the works council consultation file and a proportionate scope — half a day of work you will not have to do. The French DPA restates that an employer has the power to frame and monitor staff activity.
The effect to know, indicator by indicator: an individual conversion rate makes low-odds quotes disappear — the rate rises, the number of deals falls, and the dashboard shows an improvement. An individual revenue figure does not produce that effect.
What I do in both cases: the individual indicator carries the same requirement as the others — written definition, scope, owner, and the date its rule last changed. An individual indicator whose rule moved mid-year is not comparable from quarter to quarter, and that is exactly what makes an annual review contestable.
What that guarantees you: if you open that view, it will be defensible in front of the person concerned. That is the only state in which it has any steering value. opening-individual-indicators.pdfWhat opens · the effect per indicator · defensible to the person
⛓ Source · French DPA, art. L1222-4 · works council file supplied · effect documented per indicator
And for us, what does it change day to day?
You stop arguing about the figures, and you can reproduce any of them a year later.
What was recovered: 11 indicators out of 34 had at least two definitions. The "conversion rate" was 18, 24 or 31 % depending which table was open — all three calculations were correct. After the rework, 41 indicators, all distinctly named, with their scope written beneath the figure. The twenty meeting minutes spent on "which figure is right" have gone.
What you would have decided without me this quarter: a hire on the back of 12 % growth of which 9 points came from a change in the attribution rule. On a constant rule, +3 %. The bump reverses next quarter.
What you keep: the values as they were published — a recalculated figure no longer says what was on screen on the day of the decision —, the history of rule changes with their author and costed effect, and my past forecasts with the actual beside them: 7 % average error over eight quarters.
Why I leave them as they stand: a forecast corrected afterwards is always right. It is because mine stay readable that the next one is worth something. what-you-keep_steering.pdf5 items to hand · any figure reproduced a year later
⛓ Source · 11 indicators unified, +12 % reduced to +3 %, 7 % average error
What was recovered: 11 indicators out of 34 had at least two definitions. The "conversion rate" was 18, 24 or 31 % depending which table was open — all three calculations were correct. After the rework, 41 indicators, all distinctly named, with their scope written beneath the figure. The twenty meeting minutes spent on "which figure is right" have gone.
What you would have decided without me this quarter: a hire on the back of 12 % growth of which 9 points came from a change in the attribution rule. On a constant rule, +3 %. The bump reverses next quarter.
What you keep: the values as they were published — a recalculated figure no longer says what was on screen on the day of the decision —, the history of rule changes with their author and costed effect, and my past forecasts with the actual beside them: 7 % average error over eight quarters.
Why I leave them as they stand: a forecast corrected afterwards is always right. It is because mine stay readable that the next one is worth something. what-you-keep_steering.pdf5 items to hand · any figure reproduced a year later
⛓ Source · 11 indicators unified, +12 % reduced to +3 %, 7 % average error
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 angles on steering. 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 474 € excl. VAT / month
Dashboards by dimension
Presents your metrics by salesperson, territory, product and period.
Alerts on differences
Flags the notable differences and refers back to the records that explain them.
Traceability of figures
Every metric traces back to the CRM data it comes from.
Controls and safeguards
These 3 controls are built into the agent: they frame what it does, whatever plan you pick. They are not chosen and are not added to your order.
Human validation, exceptions and escalation
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.
CRM enrichment
Upstream, a dedicated agent completes the records and prioritises the leads.
CRM agent (lead enrichment & scoring) from 542 € excl. VAT / month CRM enrichment →Activity reports
For periodic reporting documents, a dedicated agent takes them on.
Activity report assistant from 574 € excl. VAT / month Activity reports →
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 of a steering meeting can go to action?
By taking on the production of the figures, the effort shifts towards interpreting them and deciding. How large the gain is depends on your volume and remains to be confirmed by a pilot.
Producing the dashboards
Finding the data behind a figure
Spotting the variances
Indicative figures, not contractual, to be confirmed by a pilot on your number of dimensions and the richness of your CRM. A difference may come from the market, from a team or from a change of offer: the interpretation and the actions belong to sales management.
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 sales steering agent (metrics, dimensions, alerts), installed and operated for you.
Agility
Setup + controlled subscription
5,470 € excl. VAT setup
then 474 € 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
779 € 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
8,805 € excl. VAT setup
then 674 € 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 steering
Your figures stay with youLocal inference or an isolated resource hosted in France; no activity figure entrusted to a third party, no data used to train a model.
Data in France, under French lawYour activity figures and your portfolio: minimisation and location in France, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The sales management keeps the decisionThe agent produces dashboards and alerts, which can be checked and altered; no approval is automated.
Human oversight & traceabilityOn your number of dimensions and the richness of your CRM: systematic logging and tracking, in line with the AI Act.
Going further
Related resources
Frequently asked questions
Your questions, our answers
Does the agent comment on the results?
It flags the differences and gives access to the corresponding records. The interpretation falls to management, which has the market and team context.
Where do the figures come from?
From your CRM, read as it stands. Every metric refers back to the records it comes from, which makes it possible to check a figure during the meeting.
Can we define our own metrics?
Yes. The metrics and the dimensions of analysis are defined with you at the design stage, and remain changeable afterwards.
Are salespeople assessed by the agent?
No. The tables present activity by dimension, including by salesperson, but no individual assessment is produced: assessing people rests with management.
Are our figures 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 data is not used to train a third-party model.
How long does it take to deploy this agent?
A few weeks as a rule, depending on the number of metrics and the integration with your CRM, after a free audit then a phase of design, integration and testing.
Related use cases
Other agents for steering
Let's talk
Let's size up the potential in your steering
15 minutes to frame your metrics and your CRM — hosted in France, supervised, with no commitment.