An orchestration agent, scoped with you before it is priced. This agent coordinates several specialised agents. Its orchestration follows your actual workflows — which is why it is scoped with you rather than bought off the shelf. We establish the scope together, then the quotation commits it.
The journeys described below form the scope that this review refines and the quotation commits. The specialised agents it coordinates can be ordered today.
Request a quote
● Business offer — Communication & advertising
The AI agent for communication agencies: produce more content, without diluting the creative work
Copywriting, social media variants, briefs, SEO/SEA, campaign reporting: an enormous share of the teams' time goes into production and formatting, not into the idea. Your AI agent absorbs this repetitive work and gives you back time to create. Hosted in France — on local inference or an isolated resource — your content, your brands and your advertisers' data stay with you. The creative direction keeps control.
Hosted in France
Content & advertiser data protected
GDPR & AI Act: governed deployment
Human oversight
Updated on
Deployed in a few weeks
Communication assistant · hosted in France
We are launching the back-to-school campaign for the Aubel brand. From the brief and the brand guidelines, prepare 1 LinkedIn post, 3 Instagram stories and 5 email subject lines, premium in tone but accessible.
Brief and brand guidelines read. I have produced the variants in keeping with the Aubel tone of voice (formal address, no emojis, "craftsmanship" vocabulary). Delivered: 1 LinkedIn post (2 variants), 3 stories framed around the product benefit, 5 email subject lines of which 2 can be A/B tested.
Everything is written from your key messages, ready to read over — the art director and the account manager approve before publication.
⛓ Source · the client brief + the brand guidelines + the brand's history
Everything is written from your key messages, ready to read over — the art director and the account manager approve before publication.
⛓ Source · the client brief + the brand guidelines + the brand's history
Good. And on rights, can we publish the visuals you are proposing?
A point for you to settle: the copy is original and usable. For the visuals, I only produce variants from your templates and the client's licensed image library — no unsourced third-party visual.
I am attaching the list of assets used and their rights status, so you keep the audit trail before it goes out.
✎ Action · rights note ready to read over — the studio approves
I am attaching the list of assets used and their rights status, so you keep the audit trail before it goes out.
✎ Action · rights note ready to read over — the studio approves
Local inference · no data outside the EU
Data 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 creative team decidesThe agent assists, never publishes alone
✦ In brief
In a communication or advertising agency, a Blue Lemon Agent agent speeds up the repetitive tasks — copywriting, social media variants, visual creation, SEO/SEA, monitoring and campaign reporting — and frees up time for strategy and creative work. It runs on local inference or is hosted in France: your content, your brands and your advertisers' data are never exposed to a foreign service, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity. Production becomes industrial, the creative value stays human. Live within a few weeks Your teams write to it from Microsoft Teams, Slack or their email, and your advertisers reach it on WhatsApp Business, your website chat or email — with no account to create and nothing to install. These connections are included in every plan, at no extra cost, within the number of connections your level includes. For each advertiser, the agent works on the accounts the brand owns — LinkedIn pages, Instagram professional accounts, Facebook pages — and on the reviews on its Google Business Profile or on Trustpilot, with one queue and one approval circuit per brand. These platforms belong to third parties and their interfaces change without notice; nothing is published in an advertiser's name without their approval, and access fees are passed on at actual cost, with no margin, outside the subscription. Trustpilot cannot be ordered directly: that connection requires your own Trustpilot for Business account with access to the API module, your Business Unit ID and the matching rights; it is priced on quote, after a feasibility study, with Trustpilot's own costs at your charge.
100%
hosted in France in the target architecture
0
transfer outside the EU in the target architecture
9
uses ready to deploy on this scope
0
decision taken without human approval
Reference points describing our offer, not results measured at a client. The scale of the gain is confirmed by a pilot on your own scope.
The context
Why AI is shaking up agencies — and why they hesitate
Advertisers expect more content, on more channels, faster and on a constrained budget. But mass production threatens what makes an agency valuable — the idea — and the data brands entrust to it is among the most sensitive to expose.
! The issue
The agency is caught between clients who want omnichannel volume (posts, articles, newsletters, variants, reporting) and creative teams whose time is swallowed by production and formatting. Yet most consumer AI tools amount to entrusting confidential briefs, media plans, unpublished content and advertisers' data to a third party, often hosted outside Europe and subject to the Cloud Act — and, on top of that, generated content whose origin and rights are unclear.
✓ Our answer
AI is only of interest to an agency if it is sovereign and confidential by design, and trained on your guidelines, your key messages and your content — not on a public model that learns from your briefs. Local inference or an isolated resource hosted in France, systematic human oversight, approval reserved for the creative direction: time gained on production is never paid for in lost confidentiality or a diluted brand. The aim is not to replace the copywriter or the art director, but to give them back headspace for the idea.
The decisive point
Confidentiality of content & advertiser data: sovereignty & compliance
An agency handles its clients' data and unpublished content. Here is how the architecture of our agents protects it, brand by brand.
Local inference
The agent can run on a machine at the agency: no brief and no creative work leaves the network, nothing passes through a cloud.
Hosting in France
Otherwise, a dedicated and isolated resource, hosted in France under French law — your content and advertiser data: processing and access within the European Union targeted by the architecture.
Reduced extraterritorial exposure
Exposure of your content and advertiser data to the Cloud Act and FISA 702 is reduced by design; location alone does not guarantee immunity.
A resource isolated per agency
No pooling: your guidelines, your briefs and your client data stay in an environment strictly dedicated to your agency.
Encryption & controlled access
Encryption in transit and at rest, role-based access (RBAC), strong authentication and logging.
AI Act: governed deployment
The agent is strictly in support; nothing is published automatically; traceability of sources 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.
- The encryption mechanisms in transit and at rest, their components and key management are those documented for the architecture chosen.
- 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.
The agent is trained on your guidelines and content, without feeding a public model. What you and your advertisers produce stays your property — with an audit trail of the assets and their rights at every step. 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 companyTrame & Cie — independent communication agency
- Sector
- Full-service communication — publishing, social media, search and paid campaigns
- Headcount
- 18 staff, including 6 in creative (2 art directors, 2 copywriters), 4 in account management and 2 in social
- Client base
- 14 advertisers on annual contracts — industry, healthcare, local government
- Volume
- 40 campaigns a year, around 300 posts a month, €34,000 of media budget entrusted every month
- Tools in place
- Creative suite, social scheduler, ad accounts, CRM and a 12-year shared drive — the agent plugs into them, nothing is replaced
- Who decides
- The creative director approves everything that ships; project managers set the schedules; the account team arbitrates media budgets
- Room for improvement
- 60 % of the teams' time goes into production and formatting; client reporting is put together by hand the night before; an inbound enquiry waits 3.2 days for a first reply
Trame & Cie isn't looking to produce worse, but to produce the same faster and give time back to the idea. The agent runs on local inference on a machine at the agency and plugs into the drive, the social scheduler, the ad accounts and the CRM: it prepares, the agency approves. 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.
Communication assistant · hosted in France
I read twelve years of your deliverables last night — 4,180 files — and drew one tone-of-voice guide per advertiser. You had three written down; there are fourteen this morning.
Tone of voice is the way of writing that belongs to a brand: its vocabulary, its register, what it allows itself and what it rules out. Your creatives carry it in their heads; it was written down nowhere.
The gap I measured, and it is the gap that decides the gain: across your last 34 first drafts, 7 corrections out of 10 were about tone, not substance. Your copywriters aren't reviewing ideas, they are re-tuning a voice — the one job that can be written down once and for all.
What each guide contains, and all of it can be checked line by line: average sentence length, share of active voice, vocabulary specific to the brand, opening and closing formulas, and the words the brand never uses — the most useful trait of all, and the one no brand guide thinks to write down.
· Cerault (industry): 14 words per sentence, never a superlative, “solution” absent from the last 17 approved deliverables.
· Institut Vallois (healthcare): formal address throughout, no promise of results — a constraint of their field, held without anyone having to remind them.
· Town of Sancelles: informal on social, formal in print. Two tones for one advertiser, and it is deliberate.
What that gives you tomorrow: a first draft arrives ALREADY in the right tone. Your creatives stop rewriting and start arbitrating — and that is where your value shows.
What I suggest: the creative director reads the fourteen guides, and I apply them from the next brief on. Every guide can be corrected one word at a time: it is your material, I have only set it down clearly. tone-of-voice-guides_14-advertisers.pdfWhat each brand says, and never says
⛓ Sourced · agency drive, 4,180 approved deliverables over 12 years, successive versions of the last 34 productions
Tone of voice is the way of writing that belongs to a brand: its vocabulary, its register, what it allows itself and what it rules out. Your creatives carry it in their heads; it was written down nowhere.
The gap I measured, and it is the gap that decides the gain: across your last 34 first drafts, 7 corrections out of 10 were about tone, not substance. Your copywriters aren't reviewing ideas, they are re-tuning a voice — the one job that can be written down once and for all.
What each guide contains, and all of it can be checked line by line: average sentence length, share of active voice, vocabulary specific to the brand, opening and closing formulas, and the words the brand never uses — the most useful trait of all, and the one no brand guide thinks to write down.
· Cerault (industry): 14 words per sentence, never a superlative, “solution” absent from the last 17 approved deliverables.
· Institut Vallois (healthcare): formal address throughout, no promise of results — a constraint of their field, held without anyone having to remind them.
· Town of Sancelles: informal on social, formal in print. Two tones for one advertiser, and it is deliberate.
What that gives you tomorrow: a first draft arrives ALREADY in the right tone. Your creatives stop rewriting and start arbitrating — and that is where your value shows.
What I suggest: the creative director reads the fourteen guides, and I apply them from the next brief on. Every guide can be corrected one word at a time: it is your material, I have only set it down clearly. tone-of-voice-guides_14-advertisers.pdfWhat each brand says, and never says
⛓ Sourced · agency drive, 4,180 approved deliverables over 12 years, successive versions of the last 34 productions
We're looking for the Cerault 2024 brand guide, the one before the rebrand. Nobody can find it.
Here it is, with the three documents that go with it — and I'll tell you which one is authoritative.
· Cerault brand guide v2.3, filed 14/03/2024, the last version before the September rebrand. That's the one you want.
· The rebrand brief explaining what changed and why — useful if you have to justify a departure from it to the advertiser.
· The January 2024 awareness study that prompted it.
· And eight v2.3 adaptations still live on materials nobody has updated. Those need a decision: they carry the old logo.
How I find it when the filename says nothing — it's called CER_charte_VF_VRAIE_final2.pdf: I read the content, not the name. The logo, the palette, the filing date and the advertiser are enough to identify it.
What that search was costing you: your own timesheets record 41 hours of “looking for a document” over the quarter, more than a full week of work for five people. The same document now comes back in one sentence.
What I suggest: that I sweep recent uploads once a month and tag the advertiser, the type of document and its version. Of the 4,180 files, 1,260 have no usable name — and those are the ones you hunt for on a Friday evening. Indexing renames nothing: your existing paths keep working. agency-memory_what-was-recovered.pdf1,260 files with no usable name, indexed
⛓ Sourced · agency drive, file contents rather than filenames, timesheets for the quarter
· Cerault brand guide v2.3, filed 14/03/2024, the last version before the September rebrand. That's the one you want.
· The rebrand brief explaining what changed and why — useful if you have to justify a departure from it to the advertiser.
· The January 2024 awareness study that prompted it.
· And eight v2.3 adaptations still live on materials nobody has updated. Those need a decision: they carry the old logo.
How I find it when the filename says nothing — it's called CER_charte_VF_VRAIE_final2.pdf: I read the content, not the name. The logo, the palette, the filing date and the advertiser are enough to identify it.
What that search was costing you: your own timesheets record 41 hours of “looking for a document” over the quarter, more than a full week of work for five people. The same document now comes back in one sentence.
What I suggest: that I sweep recent uploads once a month and tag the advertiser, the type of document and its version. Of the 4,180 files, 1,260 have no usable name — and those are the ones you hunt for on a Friday evening. Indexing renames nothing: your existing paths keep working. agency-memory_what-was-recovered.pdf1,260 files with no usable name, indexed
⛓ Sourced · agency drive, file contents rather than filenames, timesheets for the quarter
All of that assumes you read our briefs, our contracts and our advertisers' data. Where exactly does it all go?
It doesn't leave your walls. I run on local inference on a machine at the agency, and your drive is copied nowhere.
Local inference means the model computes on your own machine: the text of a brief crosses no outside network to be processed. If you'd rather not host a machine, the other route is an isolated resource hosted in France, dedicated to your agency — not shared with another agency, nor with another advertiser.
What that changes, brand by brand:
· Your briefs train no public model. What I learn from Cerault serves Cerault. Nothing you entrust to me surfaces anywhere else.
· Encryption in transit and at rest, and role-based access — rights follow the job: the social intern opens editorial calendars, not media plans or rights agreements.
· A full log: who asked for what, when, and what was approved. That log is what lets you answer an advertiser who asks.
· Hosting in France, under French law, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
And you change no tools: I plug into your creative suite, your social scheduler, your ad accounts, your CRM and your DMS — the document management system, meaning your twelve-year drive. No migration, no extra subscription for your teams.
The figure that makes this commercial rather than technical: 3 of your 14 advertisers — two in healthcare, one local authority — already require hosting in France in their framework contract, and the last public tender you answered scored it as a criterion worth 5 points. You ticked the box without a caveat.
What I suggest: that I keep up to date the technical sheet your advertisers ask for at renewal — hosting, subprocessors, retention periods, who accesses what. It is asked for once a year and takes three days to assemble. technical-framework_where-your-briefs-live.pdfLocal inference, role-based access, processing in the EU targeted
✎ Framework · deployment architecture, access log, framework contracts of the 14 advertisers
Local inference means the model computes on your own machine: the text of a brief crosses no outside network to be processed. If you'd rather not host a machine, the other route is an isolated resource hosted in France, dedicated to your agency — not shared with another agency, nor with another advertiser.
What that changes, brand by brand:
· Your briefs train no public model. What I learn from Cerault serves Cerault. Nothing you entrust to me surfaces anywhere else.
· Encryption in transit and at rest, and role-based access — rights follow the job: the social intern opens editorial calendars, not media plans or rights agreements.
· A full log: who asked for what, when, and what was approved. That log is what lets you answer an advertiser who asks.
· Hosting in France, under French law, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
And you change no tools: I plug into your creative suite, your social scheduler, your ad accounts, your CRM and your DMS — the document management system, meaning your twelve-year drive. No migration, no extra subscription for your teams.
The figure that makes this commercial rather than technical: 3 of your 14 advertisers — two in healthcare, one local authority — already require hosting in France in their framework contract, and the last public tender you answered scored it as a criterion worth 5 points. You ticked the box without a caveat.
What I suggest: that I keep up to date the technical sheet your advertisers ask for at renewal — hosting, subprocessors, retention periods, who accesses what. It is asked for once a year and takes three days to assemble. technical-framework_where-your-briefs-live.pdfLocal inference, role-based access, processing in the EU targeted
✎ Framework · deployment architecture, access log, framework contracts of the 14 advertisers
We have Cerault's brief for their workshop range launch. What can you produce?
Thirty-four pieces ready to review, in Cerault's tone of voice — and I'll tell you the order to approve them in.
The core first: a 1,200-word article carrying the main argument. Everything else flows from it — if the article changes, I propagate the change to the other thirty-three without being asked.
Then the adaptations:
· LinkedIn — 6 posts, expert register, one per benefit, long form.
· Instagram — 8 posts, short sentences, the workshop shown rather than described.
· Newsletter — 2 sends, one announcement, one re-send to non-openers a fortnight later.
· 16 ad headlines, four angles across four formats.
· A six-week editorial calendar — the plan of what goes out, where and on which day — paced on the slots where Cerault's audience responds: Tuesday and Thursday mornings, which carry 2.1 times average engagement across their last twelve months.
The time this moves: a first draft of an article or post used to take 22 % of a campaign's time; it now takes 6 %. The remaining 6 % is your review, and it is what makes the difference between a correct text and a Cerault text.
What I did not write: the range's production figure. The brief states it; no agency source confirms it. I've left it highlighted: have the advertiser confirm it before publication — that's the kind of figure that ends up in the press.
Approval order I suggest: the article, then the headlines — they're what commits media budget. Social can wait until tomorrow. workshop-campaign_34-pieces.pdfOne core, five adaptations, one calendar
⛓ Sourced · advertiser brief, Cerault tone-of-voice guide, 12 months of engagement by time slot
The core first: a 1,200-word article carrying the main argument. Everything else flows from it — if the article changes, I propagate the change to the other thirty-three without being asked.
Then the adaptations:
· LinkedIn — 6 posts, expert register, one per benefit, long form.
· Instagram — 8 posts, short sentences, the workshop shown rather than described.
· Newsletter — 2 sends, one announcement, one re-send to non-openers a fortnight later.
· 16 ad headlines, four angles across four formats.
· A six-week editorial calendar — the plan of what goes out, where and on which day — paced on the slots where Cerault's audience responds: Tuesday and Thursday mornings, which carry 2.1 times average engagement across their last twelve months.
The time this moves: a first draft of an article or post used to take 22 % of a campaign's time; it now takes 6 %. The remaining 6 % is your review, and it is what makes the difference between a correct text and a Cerault text.
What I did not write: the range's production figure. The brief states it; no agency source confirms it. I've left it highlighted: have the advertiser confirm it before publication — that's the kind of figure that ends up in the press.
Approval order I suggest: the article, then the headlines — they're what commits media budget. Social can wait until tomorrow. workshop-campaign_34-pieces.pdfOne core, five adaptations, one calendar
⛓ Sourced · advertiser brief, Cerault tone-of-voice guide, 12 months of engagement by time slot
The 16 headlines — how did you choose them?
I didn't choose them: I built them to be comparable. Four angles × four formats, everything else identical.
Comparable means ready for an A/B test: two versions served at the same time to two halves of the audience, everything else equal, so that the gap you measure comes from the one thing that changed.
The four angles, taken from the brief's own arguments: time saved · durability · French manufacture · cost in use. None of them is my opinion — they are the four the advertiser wrote, in their order.
Why it's comparable: same length, same call to action, same visual. If an angle wins, it's the angle that won, not one headline written better than the others.
What their history already says: across the 2025 Cerault campaigns, “French manufacture” carried 2.3 times the click rate of the other three. That's a reason to test it first, not a reason to skip the test — the product isn't the same.
What I suggest next: let it run five days, then move the budget of the two losing angles onto the two winners, and write four new headlines in the winning angle. You approve the switch, I prepare everything — including the next four headlines, so the following test is ready the day this one ends.
Comparable means ready for an A/B test: two versions served at the same time to two halves of the audience, everything else equal, so that the gap you measure comes from the one thing that changed.
The four angles, taken from the brief's own arguments: time saved · durability · French manufacture · cost in use. None of them is my opinion — they are the four the advertiser wrote, in their order.
Why it's comparable: same length, same call to action, same visual. If an angle wins, it's the angle that won, not one headline written better than the others.
What their history already says: across the 2025 Cerault campaigns, “French manufacture” carried 2.3 times the click rate of the other three. That's a reason to test it first, not a reason to skip the test — the product isn't the same.
What I suggest next: let it run five days, then move the budget of the two losing angles onto the two winners, and write four new headlines in the winning angle. You approve the switch, I prepare everything — including the next four headlines, so the following test is ready the day this one ends.
And will the article actually get found? Cerault keeps asking us for search, and we have nobody on it.
The article is written FOR a specific query rather than optimised afterwards — and the plan is ready before the first line.
Organic search, or SEO, is what lifts you in a search engine's free results; SEA is the paid advertising of the same engine. Both are worked on, but not in the same place.
The main query I'm targeting: “professional workbench”, 2,400 searches a month, clear buying intent, and Cerault appears nowhere in the first three pages.
Three secondary queries carried by the same article: 1,900, 2,600 and 1,400 monthly searches — 8,300 in total. They are handled in the subheadings, without forcing a word into a sentence.
The meta tags are written — the title and description shown in the search result, before anyone clicks: a 58-character title, a 152-character description, with the French manufacture argument in both, since that is what makes people click everywhere else.
What I rule out, and I say so because this is where six months get lost: 12 high-volume keywords I advise against targeting. On “workbench”, the top three results are marketplaces with a hundred times your authority: targeting that query means writing for page four.
Internal links: the article points to the 6 relevant product pages, and 4 older Cerault blog posts point to it — they already rank, so they pass on part of their weight.
What I suggest: I measure the position at 30 and 60 days and tell you which of the four queries move. If only one takes off, we'll know why the other three didn't. keyword-plan_cerault-workshop.pdf4 queries targeted, 12 ruled out and why
⛓ Sourced · search volumes, results displayed for the 4 queries, advertiser's blog
Organic search, or SEO, is what lifts you in a search engine's free results; SEA is the paid advertising of the same engine. Both are worked on, but not in the same place.
The main query I'm targeting: “professional workbench”, 2,400 searches a month, clear buying intent, and Cerault appears nowhere in the first three pages.
Three secondary queries carried by the same article: 1,900, 2,600 and 1,400 monthly searches — 8,300 in total. They are handled in the subheadings, without forcing a word into a sentence.
The meta tags are written — the title and description shown in the search result, before anyone clicks: a 58-character title, a 152-character description, with the French manufacture argument in both, since that is what makes people click everywhere else.
What I rule out, and I say so because this is where six months get lost: 12 high-volume keywords I advise against targeting. On “workbench”, the top three results are marketplaces with a hundred times your authority: targeting that query means writing for page four.
Internal links: the article points to the 6 relevant product pages, and 4 older Cerault blog posts point to it — they already rank, so they pass on part of their weight.
What I suggest: I measure the position at 30 and 60 days and tell you which of the four queries move. If only one takes off, we'll know why the other three didn't. keyword-plan_cerault-workshop.pdf4 queries targeted, 12 ruled out and why
⛓ Sourced · search volumes, results displayed for the 4 queries, advertiser's blog
We need the campaign in 41 formats, all channels plus outdoor. That's two days of studio time.
The 41 formats are ready to review, and every one comes out of your templates — not out of an invention of mine.
How they're produced: I start from your master templates — the reference layout files every other size is derived from — and the Cerault v3.1 brand guide. Each format keeps the grid, the margins, the typefaces and the palette. What I adapt is the crop, the order of blocks and the type size — never the brand guide.
The 41: 12 social formats, 9 banners, 6 outdoor formats, 8 email adaptations, 6 press formats.
The four I set aside, and why: in 4×3 outdoor, the headline falls inside the 25 mm safe margin — the margin the site imposes, inside which nothing may sit or it gets trimmed at print. I didn't shorten it to make it fit: I've prepared two possible crops, and creative decides in thirty seconds.
The time it gives you back: multi-channel adaptations used to take 60 % of a campaign's time. They now take 18 % — the rest is your review, and that doesn't delegate. Across 40 campaigns a year, it's your studio that gets the two days back, not your margin that pays for them. visual-adaptations_41-formats.pdfWhat is adapted, what never is
⛓ Sourced · agency master templates, Cerault brand guide v3.1, technical safe margins
How they're produced: I start from your master templates — the reference layout files every other size is derived from — and the Cerault v3.1 brand guide. Each format keeps the grid, the margins, the typefaces and the palette. What I adapt is the crop, the order of blocks and the type size — never the brand guide.
The 41: 12 social formats, 9 banners, 6 outdoor formats, 8 email adaptations, 6 press formats.
The four I set aside, and why: in 4×3 outdoor, the headline falls inside the 25 mm safe margin — the margin the site imposes, inside which nothing may sit or it gets trimmed at print. I didn't shorten it to make it fit: I've prepared two possible crops, and creative decides in thirty seconds.
The time it gives you back: multi-channel adaptations used to take 60 % of a campaign's time. They now take 18 % — the rest is your review, and that doesn't delegate. Across 40 campaigns a year, it's your studio that gets the two days back, not your margin that pays for them. visual-adaptations_41-formats.pdfWhat is adapted, what never is
⛓ Sourced · agency master templates, Cerault brand guide v3.1, technical safe margins
And the photography? We got caught out once by an image that was lying around.
Every visual in the campaign carries its licence and its expiry date — and three expire before the campaign ends.
A rights grant is the photographer's written permission. It is bounded by media, territory and duration: once the duration is up, the image has to come off the materials, even if the file is still on the drive.
What I check before adapting anything: the image's origin, the licence type, the scope granted, and the expiry date, compared with the planned run.
The three cases:
· Two photos from the 2023 workshop shoot: three-year grant, expiring 12/11/2026, and the campaign runs into January. An extension has to be requested from the photographer. I've drafted the letter, with the exact scope and the media concerned — it goes out the moment you sign it.
· One stock image whose licence excludes large-format outdoor. The campaign plans six. Two ways through: upgrade the licence, or take one of the four alternatives from your own library that I've pulled out — same crop, same dominant colour, unlimited rights and no cost.
What I use in no circumstances: a visual whose origin I can't name. That isn't caution, it's copyright — and an image with no provenance costs more than a day of studio time.
What I suggest: that I watch rights expiry across all 14 advertisers. There are 26 in the next six months, 9 of them on materials still live. You set the warning period, I bring them to you with the letter already written. image-rights_26-expiries.pdf3 grants to extend before January
⛓ Sourced · agency image library, rights agreements, stock licences
A rights grant is the photographer's written permission. It is bounded by media, territory and duration: once the duration is up, the image has to come off the materials, even if the file is still on the drive.
What I check before adapting anything: the image's origin, the licence type, the scope granted, and the expiry date, compared with the planned run.
The three cases:
· Two photos from the 2023 workshop shoot: three-year grant, expiring 12/11/2026, and the campaign runs into January. An extension has to be requested from the photographer. I've drafted the letter, with the exact scope and the media concerned — it goes out the moment you sign it.
· One stock image whose licence excludes large-format outdoor. The campaign plans six. Two ways through: upgrade the licence, or take one of the four alternatives from your own library that I've pulled out — same crop, same dominant colour, unlimited rights and no cost.
What I use in no circumstances: a visual whose origin I can't name. That isn't caution, it's copyright — and an image with no provenance costs more than a day of studio time.
What I suggest: that I watch rights expiry across all 14 advertisers. There are 26 in the next six months, 9 of them on materials still live. You set the warning period, I bring them to you with the letter already written. image-rights_26-expiries.pdf3 grants to extend before January
⛓ Sourced · agency image library, rights agreements, stock licences
And if we ask you for an image that exists nowhere? A shot we've never taken.
I produce it, and your art director decides whether it stays. Three routes, from cheapest to longest.
· Re-cropping and retouching your own library — the shot often exists, badly framed. Of the six missing shots on this campaign, two were already yours.
· A composite built from your own elements: product cut out, setting from your own shoot, lighting matched. Your rights, your material.
· A synthetic image, when neither of the first two routes yields the shot. I did that for the four remaining shots: 40 minutes instead of a day's shoot. The art director kept two, rejected one and had the last one re-shot — which is exactly the division of labour you want.
What I mark, every time: every synthetic image goes out with its provenance recorded in the file's metadata and a line in the asset note. Your own internal rule of 12/03/2026 requires it, and three of your advertisers have written it into their framework contract — the rule comes from their requirement, not from me.
What it opens up commercially: testing a visual idea no longer costs a production. You can put six routes on the table in a client presentation where you used to show two — and the shoot is commissioned on the one the advertiser chose, not on the one you had to guess.
· Re-cropping and retouching your own library — the shot often exists, badly framed. Of the six missing shots on this campaign, two were already yours.
· A composite built from your own elements: product cut out, setting from your own shoot, lighting matched. Your rights, your material.
· A synthetic image, when neither of the first two routes yields the shot. I did that for the four remaining shots: 40 minutes instead of a day's shoot. The art director kept two, rejected one and had the last one re-shot — which is exactly the division of labour you want.
What I mark, every time: every synthetic image goes out with its provenance recorded in the file's metadata and a line in the asset note. Your own internal rule of 12/03/2026 requires it, and three of your advertisers have written it into their framework contract — the rule comes from their requirement, not from me.
What it opens up commercially: testing a visual idea no longer costs a production. You can put six routes on the table in a client presentation where you used to show two — and the shoot is commissioned on the one the advertiser chose, not on the one you had to guess.
The monthly Cerault meeting is at 9 tomorrow. Nothing is prepared.
The pack is ready, with the reading and the three decisions to get taken. It only needs your review.
What the campaign did: 412,000 impressions, 1.9 % click rate against 1.1 % on the previous campaign, 34 quote requests against 22. Cost per request falls from €74 to €41 — the budget spent divided by the number of requests received, the one figure your advertiser compares month to month.
What it means, and I'm saying it before I'm asked: the “French manufacture” angle carries 68 % of clicks on 25 % of the budget. Across 412,000 impressions the gap is too clear to be chance — it was the angle tested first, and the test settled it.
On organic search: the launch article sits 4th for “professional workbench”, the 2,400-searches-a-month query we targeted. It was nowhere in the first three pages eight weeks ago. Three tags to rewrite and it moves past two competitors — I've written them.
The three decisions for tomorrow: move budget onto the winning angle · raise December's media budget from €1,400 to €2,400, which at a constant cost per request would mean 58 requests instead of 34 · open the campaign to tradespeople, an audience Cerault has never targeted and which already accounts for 19 % of the requests received.
The time this moves: this pack used to be 40 % of a campaign's time, assembled by hand the night before. It now takes 8 %, and it is ready before you ask for it.
What I did not put in the pack: a revenue projection. I have neither their conversion rate nor their average order value. Ask Cerault for those two figures tomorrow and the projection is in next month's pack — until then I'd rather hand you a shorter pack than a decorative one. client-pack_cerault_november.pdf3 decisions, each with its figure
⛓ Sourced · ad accounts, search console, quote request forms
What the campaign did: 412,000 impressions, 1.9 % click rate against 1.1 % on the previous campaign, 34 quote requests against 22. Cost per request falls from €74 to €41 — the budget spent divided by the number of requests received, the one figure your advertiser compares month to month.
What it means, and I'm saying it before I'm asked: the “French manufacture” angle carries 68 % of clicks on 25 % of the budget. Across 412,000 impressions the gap is too clear to be chance — it was the angle tested first, and the test settled it.
On organic search: the launch article sits 4th for “professional workbench”, the 2,400-searches-a-month query we targeted. It was nowhere in the first three pages eight weeks ago. Three tags to rewrite and it moves past two competitors — I've written them.
The three decisions for tomorrow: move budget onto the winning angle · raise December's media budget from €1,400 to €2,400, which at a constant cost per request would mean 58 requests instead of 34 · open the campaign to tradespeople, an audience Cerault has never targeted and which already accounts for 19 % of the requests received.
The time this moves: this pack used to be 40 % of a campaign's time, assembled by hand the night before. It now takes 8 %, and it is ready before you ask for it.
What I did not put in the pack: a revenue projection. I have neither their conversion rate nor their average order value. Ask Cerault for those two figures tomorrow and the projection is in next month's pack — until then I'd rather hand you a shorter pack than a decorative one. client-pack_cerault_november.pdf3 decisions, each with its figure
⛓ Sourced · ad accounts, search console, quote request forms
Opening up to tradespeople — what does that actually mean inside the ad accounts?
One more audience, three ads written for it, and a test budget I suggest capping at €400. It is all built, none of it is live.
An audience, in an ad account, is the population the ad is shown to: trade, area, interests, pages already visited.
What I've built:
· The “wood and metal trades” audience, France-wide, crossed with visitors to the 6 product pages — it's the crossing that makes it profitable, not the trade on its own.
· Three ads in Cerault's tone of voice, on the “French manufacture” angle since that is the one that wins, but with the benefit rewritten for a tradesperson: service life, not workshop compliance.
· Nine negative keywords — the searches the ad must never show up on: “second-hand”, “free”, “build your own”, “DIY”… They cost you €340 in wasted clicks last month, and nobody had excluded them.
· Exclusion of existing Cerault customers, from the list the advertiser gives us. Paying to be seen by your own customers is the most common leak in an ad account.
The threshold I suggest: €400 over three weeks, stopping if the cost per request goes above €60 — beyond that, the trades audience is less profitable than the one you already have, and it is better to know in three weeks than in three months.
What I advise against, with the figures: opening the “DIY consumers” audience at the same time. Across requests received, it accounts for 6 % of volume and 0 signed quotes in two years. It would eat the test budget without settling anything.
The decision stays with the account team: approve the audience and the threshold, I switch it on tomorrow morning and bring you the first reading at day 3. paid-campaign_trades-audience.pdf3 ads, 9 negative keywords, stop threshold
⛓ Sourced · ad accounts, search terms that triggered the ads, history of requests by type
An audience, in an ad account, is the population the ad is shown to: trade, area, interests, pages already visited.
What I've built:
· The “wood and metal trades” audience, France-wide, crossed with visitors to the 6 product pages — it's the crossing that makes it profitable, not the trade on its own.
· Three ads in Cerault's tone of voice, on the “French manufacture” angle since that is the one that wins, but with the benefit rewritten for a tradesperson: service life, not workshop compliance.
· Nine negative keywords — the searches the ad must never show up on: “second-hand”, “free”, “build your own”, “DIY”… They cost you €340 in wasted clicks last month, and nobody had excluded them.
· Exclusion of existing Cerault customers, from the list the advertiser gives us. Paying to be seen by your own customers is the most common leak in an ad account.
The threshold I suggest: €400 over three weeks, stopping if the cost per request goes above €60 — beyond that, the trades audience is less profitable than the one you already have, and it is better to know in three weeks than in three months.
What I advise against, with the figures: opening the “DIY consumers” audience at the same time. Across requests received, it accounts for 6 % of volume and 0 signed quotes in two years. It would eat the test budget without settling anything.
The decision stays with the account team: approve the audience and the threshold, I switch it on tomorrow morning and bring you the first reading at day 3. paid-campaign_trades-audience.pdf3 ads, 9 negative keywords, stop threshold
⛓ Sourced · ad accounts, search terms that triggered the ads, history of requests by type
There are 60 comments under the posts. We never get to them in time.
All 60 are sorted, 55 have a reply drafted in Cerault's tone of voice, and 5 go to a human.
· 31 product questions — dimensions, lead times, warranty. Answers taken from the advertiser's spec sheet, with a link to the matching page.
· 13 compliments or shares: a short thank-you, no sales follow-up. Answering a compliment with an offer is the surest way to lose it.
· 11 price enquiries: Cerault doesn't publish prices. Directed to the quote form, which is also what the advertiser wants — and 3 of those 11 became quote requests this week.
· 5 go to you: two complaints about delivery times, one technical compliance question, two reasoned negative comments. An automatic reply to a well-founded complaint does more damage than silence — those deserve your project manager, not a template.
The delay, which is the whole subject: your replies used to go out on average the next day at 4 p.m. The 55 replies are ready 20 minutes after the post goes live and are waiting on your go-ahead — that is the window in which a comment is still read by the people who saw it.
What I suggest: that you approve the three low-stakes families once and for all — product questions, compliments, price enquiries. They then go out within 20 minutes, and you only review the cases that belong to you. You can withdraw that approval in one word, at any time. comments_60-sorted.pdf55 replies ready, 5 that must stay human
⛓ Sourced · comment threads, advertiser spec sheet, quote forms
· 31 product questions — dimensions, lead times, warranty. Answers taken from the advertiser's spec sheet, with a link to the matching page.
· 13 compliments or shares: a short thank-you, no sales follow-up. Answering a compliment with an offer is the surest way to lose it.
· 11 price enquiries: Cerault doesn't publish prices. Directed to the quote form, which is also what the advertiser wants — and 3 of those 11 became quote requests this week.
· 5 go to you: two complaints about delivery times, one technical compliance question, two reasoned negative comments. An automatic reply to a well-founded complaint does more damage than silence — those deserve your project manager, not a template.
The delay, which is the whole subject: your replies used to go out on average the next day at 4 p.m. The 55 replies are ready 20 minutes after the post goes live and are waiting on your go-ahead — that is the window in which a comment is still read by the people who saw it.
What I suggest: that you approve the three low-stakes families once and for all — product questions, compliments, price enquiries. They then go out within 20 minutes, and you only review the cases that belong to you. You can withdraw that approval in one word, at any time. comments_60-sorted.pdf55 replies ready, 5 that must stay human
⛓ Sourced · comment threads, advertiser spec sheet, quote forms
After a quarter, what has it actually given us?
Nine person-days a month — nine days of one person's work, half a full-time role — and they went where you put them, not where I found them.
On the three measures you tracked:
· Social adaptations of a campaign: 60 % → 18 % of the time. The biggest gain, and the most mechanical.
· Performance reporting: 40 % → 8 %. The pack builds itself the night before.
· First draft of an article or post: 22 % → 6 %.
What those hours became, from your own timesheets: +31 % on “concept” and “recommendation”, and two pitch responses you would have turned down for lack of time. You won one of them. That is the only figure on this list that shows up in your revenue, and it comes from time given back, not from one more tool.
The figure that doesn't flatter me: of 1,240 pieces prepared, 214 were rewritten by more than half — 17 %. Three advertisers account for 140 of those rewrites. They are the ones whose tone-of-voice guide is thinnest, for lack of older deliverables: Cerault has 340 files behind it, those three have fewer than 30.
What I'm doing about it: give me three approved texts per advertiser and I'll rebuild their guides. On the two advertisers where you have already done it, rewriting fell from 31 % to 9 % in six weeks — it is the same correction, applied to the last three.
And the measure you didn't have: zero output published without human approval over the quarter, across 1,240 logged approvals. That isn't a setting, it's how I'm wired in. quarter-review_what-the-agency-won-back.pdf9 person-days a month, 17 % rewritten
⛓ Sourced · agency timesheets, approval log, successive versions of deliverables
On the three measures you tracked:
· Social adaptations of a campaign: 60 % → 18 % of the time. The biggest gain, and the most mechanical.
· Performance reporting: 40 % → 8 %. The pack builds itself the night before.
· First draft of an article or post: 22 % → 6 %.
What those hours became, from your own timesheets: +31 % on “concept” and “recommendation”, and two pitch responses you would have turned down for lack of time. You won one of them. That is the only figure on this list that shows up in your revenue, and it comes from time given back, not from one more tool.
The figure that doesn't flatter me: of 1,240 pieces prepared, 214 were rewritten by more than half — 17 %. Three advertisers account for 140 of those rewrites. They are the ones whose tone-of-voice guide is thinnest, for lack of older deliverables: Cerault has 340 files behind it, those three have fewer than 30.
What I'm doing about it: give me three approved texts per advertiser and I'll rebuild their guides. On the two advertisers where you have already done it, rewriting fell from 31 % to 9 % in six weeks — it is the same correction, applied to the last three.
And the measure you didn't have: zero output published without human approval over the quarter, across 1,240 logged approvals. That isn't a setting, it's how I'm wired in. quarter-review_what-the-agency-won-back.pdf9 person-days a month, 17 % rewritten
⛓ Sourced · agency timesheets, approval log, successive versions of deliverables
Our advertisers chase us constantly to know where their deliverables stand. The account team takes the hit.
Seventy “where are we?” messages a month. I answer 59 of them within ten minutes with the real state of the schedule; 11 go to your project manager.
What I answer, and where it comes from: the stage in progress, who has it, the expected return date. I read the production schedule, not a sales promise — if a deliverable has slipped, the advertiser hears it from you before discovering it.
The 11 I pass on: one deadline renegotiation, two requests outside the contract's scope, eight subjects where the honest answer is “we need to talk”. Committing to a date isn't information, it's a project manager's decision.
What deserves a close look this week: one deliverable has been waiting on the advertiser's own approval for 11 days. I've chased it twice, politely, with the direct link. I'm flagging it because you are the one who will be held responsible for the delay, and the renewal meeting is in three weeks.
The time it gives the account team back: these chases used to take them 6 hours a week, broken into thirty interruptions. They have one hour left — the 11 genuine subjects. This isn't time saved on work, it's time saved on interruption, and in an agency that is the expensive kind.
What I suggest: a weekly status sent to each advertiser on Friday, with what is waiting on them. Across the six advertisers where we tried it, chasing dropped by 70 % — people chase when they don't know. advertiser-relations_70-requests-a-month.pdf59 answers in 10 minutes, 11 for a human
⛓ Sourced · account team inbox, production schedule, chaser log
What I answer, and where it comes from: the stage in progress, who has it, the expected return date. I read the production schedule, not a sales promise — if a deliverable has slipped, the advertiser hears it from you before discovering it.
The 11 I pass on: one deadline renegotiation, two requests outside the contract's scope, eight subjects where the honest answer is “we need to talk”. Committing to a date isn't information, it's a project manager's decision.
What deserves a close look this week: one deliverable has been waiting on the advertiser's own approval for 11 days. I've chased it twice, politely, with the direct link. I'm flagging it because you are the one who will be held responsible for the delay, and the renewal meeting is in three weeks.
The time it gives the account team back: these chases used to take them 6 hours a week, broken into thirty interruptions. They have one hour left — the 11 genuine subjects. This isn't time saved on work, it's time saved on interruption, and in an agency that is the expensive kind.
What I suggest: a weekly status sent to each advertiser on Friday, with what is waiting on them. Across the six advertisers where we tried it, chasing dropped by 70 % — people chase when they don't know. advertiser-relations_70-requests-a-month.pdf59 answers in 10 minutes, 11 for a human
⛓ Sourced · account team inbox, production schedule, chaser log
We'd also like this to find us clients, not just serve them.
Twelve accounts to approach, drawn from what has worked for you — plus the sorting of inbound enquiries, which has already cost you four deals.
Where the twelve come from: I compared your 9 most profitable advertisers with the 5 accounts lost over two years. Three traits recur: industry or healthcare (8 of 9 against 1 of 5), 50 to 300 staff (7 of 9 against 2 of 5), and a communication manager already in post (9 of 9 against 1 of 5) — without one, the cycle doubles and the margin falls.
What I prepare for each: a dated, verifiable fact — a recent campaign, an announced hire, a website rebuild —, an approach angle drawn from comparable work you've done, and two citable references from the same sector. You review, you send.
Now the inbound enquiries, and this is the most profitable subject of all: you get about twenty a month, and the average time to first reply was 3.2 days. Of the five deals lost at first contact, four went to a competitor who replied the same day.
What the last month produced: every enquiry is qualified on arrival — stated budget, sector, deadline —, the acknowledgement goes out within the hour with two useful questions, and when the criteria are met I offer three slots in the account team's calendar. 20 enquiries, 12 qualified, 7 meetings booked, two of them on accounts that match the profile of the twelve. The meeting happens between you and the prospect; all I do is open it.
And the argument that closes them, keep it for the end of your meetings: all of this lives in France, on your machine or on our servers, with the deployment objective of processing and access operated within the European Union. Your advertisers' briefs and your creative work do not leave. For an agency that isn't a legal argument — it's the one your advertisers will ask you for at the next renewal, and you will be the only name on the list who can show it in writing. new-business_12-accounts-and-inbound-triage.pdf20 enquiries, 12 qualified, 7 meetings booked
⛓ Sourced · profitability by advertiser, history of lost accounts, inbound mailbox, account team calendar
Where the twelve come from: I compared your 9 most profitable advertisers with the 5 accounts lost over two years. Three traits recur: industry or healthcare (8 of 9 against 1 of 5), 50 to 300 staff (7 of 9 against 2 of 5), and a communication manager already in post (9 of 9 against 1 of 5) — without one, the cycle doubles and the margin falls.
What I prepare for each: a dated, verifiable fact — a recent campaign, an announced hire, a website rebuild —, an approach angle drawn from comparable work you've done, and two citable references from the same sector. You review, you send.
Now the inbound enquiries, and this is the most profitable subject of all: you get about twenty a month, and the average time to first reply was 3.2 days. Of the five deals lost at first contact, four went to a competitor who replied the same day.
What the last month produced: every enquiry is qualified on arrival — stated budget, sector, deadline —, the acknowledgement goes out within the hour with two useful questions, and when the criteria are met I offer three slots in the account team's calendar. 20 enquiries, 12 qualified, 7 meetings booked, two of them on accounts that match the profile of the twelve. The meeting happens between you and the prospect; all I do is open it.
And the argument that closes them, keep it for the end of your meetings: all of this lives in France, on your machine or on our servers, with the deployment objective of processing and access operated within the European Union. Your advertisers' briefs and your creative work do not leave. For an agency that isn't a legal argument — it's the one your advertisers will ask you for at the next renewal, and you will be the only name on the list who can show it in writing. new-business_12-accounts-and-inbound-triage.pdf20 enquiries, 12 qualified, 7 meetings booked
⛓ Sourced · profitability by advertiser, history of lost accounts, inbound mailbox, account team calendar
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
The uses of AI in a communication agency
Each use corresponds to an agent we deploy. All of them work in support, subject to approval by your teams.
Included in your agent
The 3 capabilities essential to this promise are included, at no extra cost.
Scoping the brief & the brand guidelines
Reading the advertiser's brief and its guidelines, drawing from them the key messages, the tone and the prohibitions, and settling the frame before any production.
Brand-specific request queues
Requests are grouped in the relevant advertiser's queue, with its brand guidelines and history, ready for handling and approval.
Traceability of sources, templates & usage rights
The list of assets used and their rights status goes with every output; only your templates and the client's licensed image library are adapted.
Controls and safeguards
These 2 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 oversight, costs, quality, incidents and safe stop
Human validation before publication, with versions retained
What the agent must be connected to
These 3 connections are required for the agent to work. They concern your information system and are scoped during the audit.
Handoff contracts, least-privilege permissions and shared context
Link content to performance, costs and campaign objectives
Connect each brand’s channels and accounts: Teams, Slack or email to receive requests, and the accounts the advertiser owns, to work on them account by account. Rights delegated and opened by the client, access segregation to be established advertiser by advertiser, with no autonomous publishing by the agent.
Other needs our agents cover
Each card says where the matching agent stands: available, on quote, or still being architected.
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.
Writing & marketing content assistant
Templates, brand tone, document search.
Writing & marketing content assistant from 444 € excl. VAT / month Discover the agent →Social media assistant (posts)
Templates, calendar, scheduling.
Social media assistant (posts) from 522 € excl. VAT / month Discover the agent →Tender watch assistant
Sources, alerts.
Tender watch assistant from 454 € excl. VAT / month Discover the agent →Advertising campaign agent (optimisation)
Ad platforms.
Advertising campaign agent (optimisation) from 499 € excl. VAT / month Discover the agent →AI copywriting & content creation agent
Articles, social media posts, web pages, newsletters, advertising scripts: a sovereign AI agent, hosted in France, in your editorial voice.
Writing & content creation from 429 € excl. VAT / month Discover the agent →AI community management agent
Editorial planning, multi-network publishing, moderation and replies: a sovereign AI agent, hosted in France, that takes the load off your community managers.
Community management from 610 € excl. VAT / month Discover the agent →AI visual creation agent
Generating and adapting multi-format visuals for communication: a sovereign AI agent, hosted in France, producing to your brand guidelines.
Visual creation & adaptations from 432 € excl. VAT / month Discover the agent →AI search & campaigns agent
SEO optimisation, keyword research, content briefs, preparation of paid search and social campaigns: a sovereign AI agent, hosted in France.
Search & campaigns (SEO/SEA) from 534 € excl. VAT / month Discover the agent →AI agent for monitoring & reporting
Media monitoring, online reputation and performance reporting, automated: a sovereign AI agent, hosted in France, that consolidates your data continuously.
Campaign watch & reporting from 539 € excl. VAT / month Discover the agent →AI knowledge base agent
The information exists in your company — but it is scattered across procedures, contracts, an intranet and the memory of a few people.
Document agent (FAQ, knowledge base) from 678 € excl. VAT / month Discover the agent →Monitoring agent
Following competitors, the trade press and regulatory developments takes a steadiness that everyday work eats into.
Watch agent (competitors, press, regulation) from 479 € excl. VAT / month Discover the agent →
Does your need fall outside this?
Book the free audit →
Build your agent
In 15 minutes we identify the most relevant agent — without oversizing the project.
The gain
How much time can an agency recover?
By automating variants, formatting and reporting, an agency can aim for a reduction by half in production time on repetitive tasks — reinvested in billable strategy and creative work.
Social variants of a campaign (across networks)
Performance reporting for a campaign
A first version of an article or a post
Qualitative, non-contractual comparison: the proportions shown illustrate the shift of the work towards review, they represent no measurement. Every output of the agent is reviewed and approved by a competent person.
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
Three options, one agent
A copywriting and content-creation agent (production, variants, reporting), installed and operated for you. Choose according to how you work.
This agent is priced with you, not online. We are adjusting its scope at the moment, and online subscription stays closed while we do. Tell us what you need: we will come back to you with a price. Request a quote
Our commitment
Four guarantees that matter to an agency
Your content and advertiser data never leaveLocal inference or an isolated resource hosted in France; no brief and no creative work entrusted to a foreign third party.
Data in France, under French lawAs regards your content and advertiser data: hosting in France under French law, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
Your brands stay your brandsThe agent is trained on your guidelines, without feeding a public model; the content produced stays your property.
The creative direction keeps the decisionThe agent produces copy, visuals and summaries that can be checked; nothing is published automatically.
Frequently asked questions
Your questions, our answers
Can AI really produce content in each brand's voice?
Yes. The agent is trained on your editorial guidelines, your key messages and your existing content: it respects each advertiser's tone of voice (vocabulary, register, things to avoid) and produces consistent first versions. Your copywriters move from a production role to an editing and approval role — faster and worth more.
Who owns the content the agent produces?
To you. The copy and variants generated from your briefs and your guidelines stay your property and your clients', according to your contracts. The agent only produces original work or variants of your own templates, and it records the assets used and their rights status so that publication is on safe ground.
Are my advertisers' briefs and data protected?
Yes. Confidential briefs, media plans and your advertisers' data are sensitive: our agents run locally at the agency or are hosted in France, 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 briefs never feed a public model.
Will AI make everything uniform, or replace the creative work?
No. The agent takes on the repetitive production (variants, formatting, reporting) to give time back to the idea. The art direction and the copywriting team approve every deliverable: the agent assists, the human decides. The creative value stays human.
Does the agent help with SEO/SEA and reporting?
Yes. It prepares keywords, tags, ads and audiences from each campaign's objectives, then produces readable performance summaries and recommendations ready to talk through in a client meeting. Your media planners and traffic managers keep the call on budgets.
Do we have to change tools?
No. The agent connects to your existing environment (creative suite, social scheduling tools, advertising platforms, document management, CRM) and complements them, without forcing a migration. We adapt the integration to your stack.
How long does it take to deploy an agent?
A few weeks as a rule, after a free 15-minute audit that identifies the use case with the best return, then a phase of design, integration and testing before going live and training your teams.
Which tools can people use to talk to the agent?
The ones you already use. Internally, your teams write to it from Microsoft Teams, Slack or by email, just as they would to a colleague: nothing to install, nothing to learn. On the advertiser side, the agent answers on WhatsApp Business, your website chat and email — people reach you where they already write, with no account to create. Each advertiser account stays walled off: one channel per brand, one style guide per brand, and the creative director approves before anything goes out. You keep control from a web dashboard. These connectors rely on open standards, including the MCP protocol; they are included in every plan, at no extra cost, within the number of connections your level includes, and the catalogue grows at no extra cost. Only the fees charged by the platforms themselves — WhatsApp Business bills per conversation — are passed on at actual cost, with no margin, outside the subscription.
Does the agent work on our advertisers' LinkedIn, Instagram and Facebook accounts?
Yes, whenever it makes sense for you. The agent works account by account, for each of your advertisers: LinkedIn pages and Instagram professional accounts for the brands whose image lives there, Facebook pages for those speaking to the general public, Google Business Profile and Trustpilot for those that welcome visitors. It flags what has just been posted, ranks it by urgency and drafts a reply in the voice of the brand concerned. For an agency, the argument holds account by account: one queue per brand, one approval circuit per brand, with no mixing between advertisers. Two things need saying plainly. These are third-party platforms: they belong to others, their interfaces and their rules change without notice, and the connection follows those changes. And publishing in your name stays under your control: the agent prepares and proposes, and you decide what goes live. Some of these platforms charge for their own access: those fees are passed on to you at actual cost, with no margin, outside the subscription. The exact set of connected platforms is settled during the design phase.
Related industries
Other sectors that produce content and client relationships
Let's talk
Let's size up the potential at your agency
15 minutes to identify the use case with the best return — hosted in France, supervised, with no commitment.