● Business offer — Communication & advertising
The AI agent for copywriting & content creation: produce faster, in your brand voice
Brief, first draft, adaptations per channel, reviews: producing content takes a considerable amount of time — without writers and creatives always working on what creates the value. Your AI agent absorbs the first layer of writing, in your editorial voice. Hosted in France — on local inference or an isolated resource — your content, your brands and your client data stay under control. The creative direction approves.
Hosted in France
Content & brands protected
GDPR & AI Act: governed deployment
Human oversight
Updated on
Deployed in a few weeks
Copywriting assistant · hosted in France
From this client brief, write an 800-word SEO blog article on “solar energy for very small businesses”, then adapt it into 3 LinkedIn posts and a newsletter, in our editorial voice.
Article written (812 words) structured with H2/H3, the main keyword and related queries woven in naturally, meta title and meta description proposed.
Adaptations ready: 3 LinkedIn posts (hook + body + call to action), 1 newsletter (A/B subject line + intro + button). Tone taken from your editorial guidelines — for review.
⛓ Source · your editorial guidelines + the client brief + knowledge base
Adaptations ready: 3 LinkedIn posts (hook + body + call to action), 1 newsletter (A/B subject line + intro + button). Tone taken from your editorial guidelines — for review.
⛓ Source · your editorial guidelines + the client brief + knowledge base
LinkedIn hook no. 2 is too corporate. Can you make it more direct?
Rewritten in 3 more direct variants, free of jargon, with a first line that opens a concrete tension for the owner of a very small business. I have kept the same angle and the same promise, and flagged the variant most aligned with your best-performing posts.
You choose; I send nothing.
✎ Action · variants ready to approve — the creative decides
You choose; I send nothing.
✎ Action · variants ready to approve — the creative decides
Local inference · no data outside the EU
Content 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 direction decidesThe agent assists, never publishes alone
✦ In brief
For a communication team or an agency, a Blue Lemon Agent agent speeds up content production — articles, social media posts, web pages, newsletters, advertising scripts, adaptations per channel — while respecting your editorial guidelines and your brand voice. It runs on local inference or is hosted in France: your content, your client briefs and your advertiser data are never exposed to a foreign service, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity. The time saved on the first draft is reinvested in creativity and strategy. Live within a few weeks.
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 matters to communication — and why it worries people
Demand for content is exploding: multi-channel, multi-format, multi-client. But quality writing takes time, and entrusting your briefs and your content to a consumer tool often means exposing the brand's most strategic material.
! The issue
Teams are caught between a demand for volume — ever more posts, articles and adaptations — and a demand for quality and brand consistency. Yet most consumer AI tools amount to entrusting confidential briefs, launch strategies, advertiser data and unpublished content to a third party, often hosted outside Europe and subject to the Cloud Act. To say nothing of the risk of generic content, disconnected from the brand voice.
✓ Our answer
AI is only of interest in communication if it is sovereign, confidential and set to your editorial line. Local inference or an isolated resource hosted in France, an agent trained on your guidelines and your best content, systematic human oversight, decisions reserved to the creatives: the time saved on the first draft is never paid for in lost confidentiality or in uniformity. The aim is not to replace the writer, but to give them back time for the idea.
The decisive point
Confidentiality of content & advertiser data: sovereignty & ownership
A communication team handles a brand's most strategic material: launches, positioning, client data. Here is how the architecture of our agents protects it, project by project.
Local inference
The agent can run on a machine belonging to the team: no brief and no unpublished content 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
Architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
One isolated resource per client
No pooling: a strictly dedicated environment, essential when an agency handles competing brands.
Set to your brand voice
The agent is fed with your editorial guidelines and your best content: it writes in your tone, not in generic prose.
AI Act: governed deployment
An agent strictly in support; no content published automatically; traceability and human oversight from end to end.
What depends on the architecture chosen These points are not general guarantees: they are settled deployment by deployment, in the quotation.
- The applicable location is that of the architecture set out in the quotation and verified before commissioning.
- Local execution is announced only for the configuration explicitly described and accepted in the quotation.
- The applicable isolation depends on the deployment mode set out in the quotation; no dedicated isolation is presumed.
- The events logged, their content, their retention period and who may access them are defined for the deployment chosen.
Your content remains your property: the agent produces texts that your teams correct and publish. For the most sensitive data, SecNumCloud and HDS options are available depending on your requirements. 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 companySextant Contenus — independent editorial studio
- Sector
- Editorial production and content creation — articles, web pages, newsletters, social media, ad scripts
- Headcount
- 11 staff, including 4 copywriters, 2 in social media, 1 part-time search specialist, 2 in account management and 1 editorial director
- Client base
- 9 advertisers on annual contracts — B2B software, energy, healthcare, vocational training
- Volume
- 40 articles, 260 social posts, 12 newsletters and 6 web pages a month; 4 launches a year
- Tools in place
- The advertisers' CMS, a social scheduler, an e-mail platform, an office suite and a 9-year document store — the agent plugs into them, nothing is replaced
- Who decides
- The editorial director approves everything that goes live; project managers set the schedules; the advertiser signs off any figure about its own products
- Room for improvement
- A first draft of an article and its derivatives takes 60 % of the time of a production; deriving a post across networks, 37 % of the time of a post; answering a recurring advertiser request, 10 % of a project manager's time. 61 % of briefs arrive with no target query and no list of derivatives
Sextant Contenus isn't looking to write more, but to spend less time on the first layer and give that time back to the angle, the idea and the advice. The agent runs on local inference on a machine at the studio and plugs into the document store, the advertisers' CMS, the social scheduler and the e-mail platform: it prepares, the editorial director approves, nobody else publishes. 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.
Copywriting assistant · hosted in France
I read nine years of your archives last night — 3,120 published pieces and 214 briefs — and I found where your rounds of rework come from. It isn't the writing.
A brief is the written order for a piece of content: subject, angle, audience, format and what it must achieve.
The gap I measured, and it is the gap that decides the gain: 61 % of your briefs state neither the target query nor the list of derivatives expected. Those briefs produce 2.4 rounds of rework per piece; the others, 0.8. In other words, your writers aren't rewriting because they write badly: they are rewriting because the order was incomplete.
What I did without waiting: I filled in a standard brief for your 12 live productions, with everything I can establish from your own sources — advertiser, editorial voice that applies, format, target length, target query, planned derivatives, content already published on the subject — and I left highlighted what I cannot establish: across those 12 briefs, the measurable objective is missing 7 times and the approver on the advertiser's side 4 times.
What that gives you this week: a writer who opens a complete brief writes once. Across the 12 productions, the measured gap is 19 rounds of rework avoided — a little over a day of writing.
What I suggest: the editorial director reads the standard brief, we freeze it, and I pre-fill it on every new order. The boxes I cannot fill stay visible: that is their entire purpose. standard-brief_12-live-productions.pdf61 % of briefs incomplete · 19 rounds of rework avoided
⛓ Sourced · studio document store, 3,120 published pieces over 9 years, 214 briefs, successive versions of the productions
A brief is the written order for a piece of content: subject, angle, audience, format and what it must achieve.
The gap I measured, and it is the gap that decides the gain: 61 % of your briefs state neither the target query nor the list of derivatives expected. Those briefs produce 2.4 rounds of rework per piece; the others, 0.8. In other words, your writers aren't rewriting because they write badly: they are rewriting because the order was incomplete.
What I did without waiting: I filled in a standard brief for your 12 live productions, with everything I can establish from your own sources — advertiser, editorial voice that applies, format, target length, target query, planned derivatives, content already published on the subject — and I left highlighted what I cannot establish: across those 12 briefs, the measurable objective is missing 7 times and the approver on the advertiser's side 4 times.
What that gives you this week: a writer who opens a complete brief writes once. Across the 12 productions, the measured gap is 19 rounds of rework avoided — a little over a day of writing.
What I suggest: the editorial director reads the standard brief, we freeze it, and I pre-fill it on every new order. The boxes I cannot fill stay visible: that is their entire purpose. standard-brief_12-live-productions.pdf61 % of briefs incomplete · 19 rounds of rework avoided
⛓ Sourced · studio document store, 3,120 published pieces over 9 years, 214 briefs, successive versions of the productions
We're looking for the Novaris editorial voice, the one from before they changed their name. Nobody can find it, and we need it to rework some old articles.
Here it is, with the three documents that go with it — and I'll tell you which one governs.
· Novaris editorial voice v2, filed 03/09/2024, the last version before the October name change. That is the one you want.
· The name-change brief, which explains what moved in the vocabulary — useful if you have to justify a difference to the advertiser.
· The transition glossary listing the 22 replaced terms, old and new side by side.
· And 14 articles still online that use the old brand name. Those call for a decision: rework them, date them, or leave them.
How I find it when its file name says nothing — it is called LE_v2_def_OK.docx: I read the content, not the name. The vocabulary specific to the brand, the filing date and the internal cross-references are enough to identify it.
What that search was costing you: your own timesheets carry 26 hours of “looking for a document” this quarter, more than three working days spread across eleven people. The same document now comes back in one sentence.
What I suggest: that I sweep recent filings once a month and tag the advertiser, the kind of document and its version. Of the 3,120 pieces, 940 have no usable file name — and those are the ones you look for on a Friday evening. Indexing renames nothing: your current paths keep working. studio-memory_what-was-recovered.pdf940 files with no usable name, indexed
⛓ Sourced · studio document store, file contents rather than file names, quarterly timesheets
· Novaris editorial voice v2, filed 03/09/2024, the last version before the October name change. That is the one you want.
· The name-change brief, which explains what moved in the vocabulary — useful if you have to justify a difference to the advertiser.
· The transition glossary listing the 22 replaced terms, old and new side by side.
· And 14 articles still online that use the old brand name. Those call for a decision: rework them, date them, or leave them.
How I find it when its file name says nothing — it is called LE_v2_def_OK.docx: I read the content, not the name. The vocabulary specific to the brand, the filing date and the internal cross-references are enough to identify it.
What that search was costing you: your own timesheets carry 26 hours of “looking for a document” this quarter, more than three working days spread across eleven people. The same document now comes back in one sentence.
What I suggest: that I sweep recent filings once a month and tag the advertiser, the kind of document and its version. Of the 3,120 pieces, 940 have no usable file name — and those are the ones you look for on a Friday evening. Indexing renames nothing: your current paths keep working. studio-memory_what-was-recovered.pdf940 files with no usable name, indexed
⛓ Sourced · studio document store, file contents rather than file names, quarterly timesheets
That leaves the real question: AI text is recognisable from ten metres away. Our advertisers buy our pen, not generic copy.
Which is exactly why I don't write from a generic model but from your material: I have written one editorial voice per advertiser. You had two written down; there are nine this morning.
An editorial voice is the way of writing that belongs to a brand: its vocabulary, its register, its usual formats, what it allows itself and what it rules out. Your writers carry it in their heads; it was written down nowhere.
What each one contains, and all of it can be checked sentence by sentence — because everything is drawn from copy you got signed off: average sentence length, share of active voice, vocabulary specific to the brand, opening and closing formulas, heading structure, and the words the brand never uses — the most useful trait of all, and the one no brand guide thinks to write down.
· Helianthe Énergies: 16 words per sentence, a figure at the head of each paragraph, “revolutionary” and “essential” absent from the last 40 approved pieces.
· Clairval Santé: formal address throughout, no promise of results — a constraint of their field, held without anyone having to remind them.
· Institut Peyrelade: informal on social, formal on web pages. Two tones for one advertiser, and it is deliberate.
What that changes on a first draft: it arrives ALREADY in the right tone. Your writers stop re-tuning a voice and start arbitrating an angle — and the arbitration is what the advertiser pays for.
What I suggest: the editorial director reads the nine editorial voices, and I apply them from the next order on. Each can be corrected one word at a time: it is your material, I have only set it down clearly. editorial-voices_9-advertisers.pdfWhat each brand says, and never says
⛓ Sourced · 3,120 approved pieces over 9 years, corrections made to the last 46 first drafts
An editorial voice is the way of writing that belongs to a brand: its vocabulary, its register, its usual formats, what it allows itself and what it rules out. Your writers carry it in their heads; it was written down nowhere.
What each one contains, and all of it can be checked sentence by sentence — because everything is drawn from copy you got signed off: average sentence length, share of active voice, vocabulary specific to the brand, opening and closing formulas, heading structure, and the words the brand never uses — the most useful trait of all, and the one no brand guide thinks to write down.
· Helianthe Énergies: 16 words per sentence, a figure at the head of each paragraph, “revolutionary” and “essential” absent from the last 40 approved pieces.
· Clairval Santé: formal address throughout, no promise of results — a constraint of their field, held without anyone having to remind them.
· Institut Peyrelade: informal on social, formal on web pages. Two tones for one advertiser, and it is deliberate.
What that changes on a first draft: it arrives ALREADY in the right tone. Your writers stop re-tuning a voice and start arbitrating an angle — and the arbitration is what the advertiser pays for.
What I suggest: the editorial director reads the nine editorial voices, and I apply them from the next order on. Each can be corrected one word at a time: it is your material, I have only set it down clearly. editorial-voices_9-advertisers.pdfWhat each brand says, and never says
⛓ Sourced · 3,120 approved pieces over 9 years, corrections made to the last 46 first drafts
We have the Helianthe Énergies brief: an 800-word SEO blog article on “solar power for small businesses”. What can you produce?
The article is written — 812 words — in Helianthe's editorial voice, and I'll give you the order in which to review it.
What I produced: an H2/H3 structure — the heading levels of the article, which give its hierarchy to the reader and to the search engine alike —, four sections each answering a question people actually search for, the main query placed in the title, the standfirst and one subheading, the associated queries in the body without a single forced sentence, and a 57-character meta title and a 149-character meta description already written.
Where each fact comes from, line by line: the technical specifications from Helianthe's product sheet; the self-consumption grant and the applicable VAT rate from the texts in force, with the date of the version cited; the three use cases from site reports the advertiser opened to us.
The only passage I did not fill in, and I leave it highlighted: the brief announces “up to 40 % savings on the bill”. No source you hold establishes that figure — not the site reports, not the technical documentation. This isn't caution: a numerical claim in published content commits the advertiser, and it gets copied from page to page for years. The path is simple and I have prepared it: a one-page letter to Helianthe asking for the meter readings of three comparable installations — with those readings the sentence becomes true, sourced, and far more persuasive than an “up to”.
The time this moves: a first draft of an article and its derivatives took 60 % of the time of a production; it now takes 12 %. The remaining 12 % is your review, and it is what makes the difference between a correct text and a Helianthe text.
Review order I suggest: the highlighted passage first, then the headings — they decide the search performance. The body can wait until tomorrow. helianthe-article_812-words.pdfStructure, sources per paragraph, 1 figure pending
⛓ Sourced · advertiser brief, Helianthe editorial voice, product sheet, 3 site reports, texts in force cited and dated
What I produced: an H2/H3 structure — the heading levels of the article, which give its hierarchy to the reader and to the search engine alike —, four sections each answering a question people actually search for, the main query placed in the title, the standfirst and one subheading, the associated queries in the body without a single forced sentence, and a 57-character meta title and a 149-character meta description already written.
Where each fact comes from, line by line: the technical specifications from Helianthe's product sheet; the self-consumption grant and the applicable VAT rate from the texts in force, with the date of the version cited; the three use cases from site reports the advertiser opened to us.
The only passage I did not fill in, and I leave it highlighted: the brief announces “up to 40 % savings on the bill”. No source you hold establishes that figure — not the site reports, not the technical documentation. This isn't caution: a numerical claim in published content commits the advertiser, and it gets copied from page to page for years. The path is simple and I have prepared it: a one-page letter to Helianthe asking for the meter readings of three comparable installations — with those readings the sentence becomes true, sourced, and far more persuasive than an “up to”.
The time this moves: a first draft of an article and its derivatives took 60 % of the time of a production; it now takes 12 %. The remaining 12 % is your review, and it is what makes the difference between a correct text and a Helianthe text.
Review order I suggest: the highlighted passage first, then the headings — they decide the search performance. The body can wait until tomorrow. helianthe-article_812-words.pdfStructure, sources per paragraph, 1 figure pending
⛓ Sourced · advertiser brief, Helianthe editorial voice, product sheet, 3 site reports, texts in force cited and dated
On search, we only have a part-time specialist. Will the article actually get found?
The article is written FOR a precise query, not optimised afterwards — and the plan was ready before the first line.
Organic search, or SEO, is what lifts you in a search engine's free results; SEA refers to the paid ads on the same engine. Both are worked on, but not in the same place.
The main query I keep: “solar panels for small business”, 1,300 searches a month, clear buying intent, and Helianthe appears nowhere in the first three pages.
Three secondary queries carried by the same article: 880, 720 and 590 monthly searches — 3,490 in total with the main one. 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: the financial support appears in both, since that is what drives clicks everywhere else in this market.
What I rule out, and I say so because this is where six months get lost: 9 heavily searched queries I advise against targeting. On “solar panels”, 74,000 searches a month but the top five results are comparison sites and marketplaces carrying a hundred times Helianthe's authority: targeting that query means writing for page four.
The internal linking — the links between your own pages, which circulate their weight — : the article links to the 3 relevant service pages, and 5 older blog articles link to it, since they already rank.
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 — and your part-time specialist arbitrates instead of collecting. keyword-plan_helianthe.pdf4 queries targeted, 9 ruled out and why
⛓ Sourced · search volumes, results displayed on the 4 target and 9 rejected queries, advertiser's blog
Organic search, or SEO, is what lifts you in a search engine's free results; SEA refers to the paid ads on the same engine. Both are worked on, but not in the same place.
The main query I keep: “solar panels for small business”, 1,300 searches a month, clear buying intent, and Helianthe appears nowhere in the first three pages.
Three secondary queries carried by the same article: 880, 720 and 590 monthly searches — 3,490 in total with the main one. 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: the financial support appears in both, since that is what drives clicks everywhere else in this market.
What I rule out, and I say so because this is where six months get lost: 9 heavily searched queries I advise against targeting. On “solar panels”, 74,000 searches a month but the top five results are comparison sites and marketplaces carrying a hundred times Helianthe's authority: targeting that query means writing for page four.
The internal linking — the links between your own pages, which circulate their weight — : the article links to the 3 relevant service pages, and 5 older blog articles link to it, since they already rank.
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 — and your part-time specialist arbitrates instead of collecting. keyword-plan_helianthe.pdf4 queries targeted, 9 ruled out and why
⛓ Sourced · search volumes, results displayed on the 4 target and 9 rejected queries, advertiser's blog
We also need, from the same brief, a service page and two ad scripts for a short video.
The page and both scripts are written, and they come off the same base as the article — so they don't contradict each other.
The service page doesn't repeat the article: it changes its function. Headings oriented towards a decision, benefit before feature, a table of three installation packages, objections handled in the order they come up in meetings — I took them from your site reports —, and two distinct calls to action: a quote for those who are ready, a simulator for the others. Today Helianthe's page offers only one, and 74 % of visitors are not ready to ask for a quote on a first visit.
The two ad scripts, timed for the edit: one of 20 seconds, one of 30, with the duration stated for each shot, a voice-over written to be read aloud, and the legal notice placed on screen where it must be.
· Script A — the financial support: opens on the amount, closes on the process.
· Script B — the bill: opens on the cost line, closes on the same process.
The two are comparable: same duration, same call to action, same closing shot. If one angle wins, it is the angle that wins, not a better-written script.
And six hooks — three per angle — ready for the paid ads, each under the character limit of its format.
What I suggest next: run both scripts for seven days on equal budget, then shift the loser's budget onto the winner and write two new scripts in the winning angle. You approve the switch, I prepare everything — including the next two scripts, so the following test is ready the day this one ends.
The service page doesn't repeat the article: it changes its function. Headings oriented towards a decision, benefit before feature, a table of three installation packages, objections handled in the order they come up in meetings — I took them from your site reports —, and two distinct calls to action: a quote for those who are ready, a simulator for the others. Today Helianthe's page offers only one, and 74 % of visitors are not ready to ask for a quote on a first visit.
The two ad scripts, timed for the edit: one of 20 seconds, one of 30, with the duration stated for each shot, a voice-over written to be read aloud, and the legal notice placed on screen where it must be.
· Script A — the financial support: opens on the amount, closes on the process.
· Script B — the bill: opens on the cost line, closes on the same process.
The two are comparable: same duration, same call to action, same closing shot. If one angle wins, it is the angle that wins, not a better-written script.
And six hooks — three per angle — ready for the paid ads, each under the character limit of its format.
What I suggest next: run both scripts for seven days on equal budget, then shift the loser's budget onto the winner and write two new scripts in the winning angle. You approve the switch, I prepare everything — including the next two scripts, so the following test is ready the day this one ends.
Turn the article into 3 LinkedIn posts and a newsletter, in our editorial voice.
The three posts and the newsletter are ready to review, and the editorial calendar is filled in to the end of March.
The editorial calendar is the plan of what gets published, where and on which day.
The 3 LinkedIn posts, one per benefit, all built as hook + body + call to action: the first line opens a concrete tension for the owner of a small business, the body stays under the display limit before “see more”, and the call to action points to the article, not to the quote page — you don't ask for a quote from someone who has just discovered the subject.
The newsletter: a subject line in two versions for an A/B test — an A/B test sends two versions at the same time to two halves of the list, everything else equal, so that the measured difference comes from the one thing that changed —, a four-line standfirst, three blocks and a single button. Your last two newsletters carried four: the click rate splits and no block stands out.
The slots I propose, drawn from Helianthe's last twelve months: Tuesday and Thursday 8:30 am, which carry 2.2 times the average engagement; newsletter on Tuesday 7 am, never on a Friday — 31 % fewer opens on this list.
The time this moves: deriving a post across networks took 37 % of the time of a post; it now takes 8 %.
Nothing is scheduled: the drafts are placed in your tool, and the editorial director schedules with one click what she has approved. I suggest starting with the newsletter: it has the shortest window. derivatives_3-posts-1-newsletter.pdfOne base, four derivatives, one calendar
⛓ Sourced · approved article, Helianthe editorial voice, 12 months of engagement by slot, list open rates
The editorial calendar is the plan of what gets published, where and on which day.
The 3 LinkedIn posts, one per benefit, all built as hook + body + call to action: the first line opens a concrete tension for the owner of a small business, the body stays under the display limit before “see more”, and the call to action points to the article, not to the quote page — you don't ask for a quote from someone who has just discovered the subject.
The newsletter: a subject line in two versions for an A/B test — an A/B test sends two versions at the same time to two halves of the list, everything else equal, so that the measured difference comes from the one thing that changed —, a four-line standfirst, three blocks and a single button. Your last two newsletters carried four: the click rate splits and no block stands out.
The slots I propose, drawn from Helianthe's last twelve months: Tuesday and Thursday 8:30 am, which carry 2.2 times the average engagement; newsletter on Tuesday 7 am, never on a Friday — 31 % fewer opens on this list.
The time this moves: deriving a post across networks took 37 % of the time of a post; it now takes 8 %.
Nothing is scheduled: the drafts are placed in your tool, and the editorial director schedules with one click what she has approved. I suggest starting with the newsletter: it has the shortest window. derivatives_3-posts-1-newsletter.pdfOne base, four derivatives, one calendar
⛓ Sourced · approved article, Helianthe editorial voice, 12 months of engagement by slot, list open rates
LinkedIn hook number 2 is too corporate. Can you make it more direct?
Three more direct variants, same angle, same promise — and I'll tell you which one looks like what already works for you.
What I changed, and nothing else: the first sentence opens on a situation the owner recognises, the jargon is gone — “optimising your energy mix” became “your electricity bill” —, and the promise moved from line six to line two.
What I did not touch: the angle, the promise, the call to action and the length. Three variants that differ only in their opening stay comparable; three variants that differ everywhere can only be settled on taste.
The one I flag, with the figure that justifies it: variant B. Across your 20 best-performing LinkedIn posts for Helianthe, 14 open on a closed question of fewer than 12 words — that is the shape of variant B. This is not my opinion, it is your own history.
What I suggest, and it beats a straight choice: publish B now, keep A for the same week next month on a comparable subject. Two posts are enough to know whether it's the shape that carries or the subject.
You choose; I publish nothing. All three variants sit in the draft, and the editorial director switches to whichever she wants.
✎ Framework · variants ready to approve — 20 best-performing posts analysed, the creative decides
What I changed, and nothing else: the first sentence opens on a situation the owner recognises, the jargon is gone — “optimising your energy mix” became “your electricity bill” —, and the promise moved from line six to line two.
What I did not touch: the angle, the promise, the call to action and the length. Three variants that differ only in their opening stay comparable; three variants that differ everywhere can only be settled on taste.
The one I flag, with the figure that justifies it: variant B. Across your 20 best-performing LinkedIn posts for Helianthe, 14 open on a closed question of fewer than 12 words — that is the shape of variant B. This is not my opinion, it is your own history.
What I suggest, and it beats a straight choice: publish B now, keep A for the same week next month on a comparable subject. Two posts are enough to know whether it's the shape that carries or the subject.
You choose; I publish nothing. All three variants sit in the draft, and the editorial director switches to whichever she wants.
✎ Framework · variants ready to approve — 20 best-performing posts analysed, the creative decides
And the visuals? We outsource to a designer and it blocks us for two days every time.
The 18 formats are ready to review, and every one of them comes from Helianthe's art direction — not from an invention of mine.
How they are produced: I start from your master templates — the reference layout files from which every size is derived — and from the Helianthe brand guide. Every format keeps the grid, the margins, the typefaces and the palette. What I adapt is the crop, the order of the blocks and the type size — never the brand guide.
The 18: 6 social formats, 4 ad banners, 3 newsletter headers, 3 article illustrations, 2 video thumbnails.
The two I set aside, and why: on the vertical video thumbnail, the hook falls under the area where the network's interface overlays its own buttons. I did not shorten it to make it fit: I prepared two possible crops, and your editorial director settles it in thirty seconds.
On rights, because that is where it gets expensive: every visual carries its origin and its licence expiry date. A rights assignment is the photographer's written permission, bounded by media, territory and duration. Two photographs from the 2023 site shoot expire on 12/11/2026, before the end of the planned run — I have drafted the extension request, and it goes out as soon as a human signs it. And I use no visual whose origin I cannot name: that isn't caution, it is copyright.
What it gives you: the two outsourced days become a review. Across 40 productions a month, your margin is what gets them back — and you can show six routes in a client meeting where you used to show two.
What I suggest: that I watch the licence expiries across all 9 advertisers. There are 11 in the next six months, 4 of them on content still online. You set the warning period, I bring them to you with the letter already written. visual-derivatives_18-formats.pdfWhat is adapted, what never is, 11 expiries
⛓ Sourced · master templates, Helianthe brand guide, assignment contracts and image library licences
How they are produced: I start from your master templates — the reference layout files from which every size is derived — and from the Helianthe brand guide. Every format keeps the grid, the margins, the typefaces and the palette. What I adapt is the crop, the order of the blocks and the type size — never the brand guide.
The 18: 6 social formats, 4 ad banners, 3 newsletter headers, 3 article illustrations, 2 video thumbnails.
The two I set aside, and why: on the vertical video thumbnail, the hook falls under the area where the network's interface overlays its own buttons. I did not shorten it to make it fit: I prepared two possible crops, and your editorial director settles it in thirty seconds.
On rights, because that is where it gets expensive: every visual carries its origin and its licence expiry date. A rights assignment is the photographer's written permission, bounded by media, territory and duration. Two photographs from the 2023 site shoot expire on 12/11/2026, before the end of the planned run — I have drafted the extension request, and it goes out as soon as a human signs it. And I use no visual whose origin I cannot name: that isn't caution, it is copyright.
What it gives you: the two outsourced days become a review. Across 40 productions a month, your margin is what gets them back — and you can show six routes in a client meeting where you used to show two.
What I suggest: that I watch the licence expiries across all 9 advertisers. There are 11 in the next six months, 4 of them on content still online. You set the warning period, I bring them to you with the letter already written. visual-derivatives_18-formats.pdfWhat is adapted, what never is, 11 expiries
⛓ Sourced · master templates, Helianthe brand guide, assignment contracts and image library licences
The monthly Helianthe review is tomorrow at 9. We haven't prepared anything.
The pack is ready, with the reading of it and the three decisions to get taken. It is only waiting for your review.
What the content did, eight weeks after going live: the article received 4,200 visits and holds 6th place on “solar panels for small business”, the 1,300-searches-a-month query we targeted — it did not appear in the first three pages in January. The newsletter opened at 34 % against 22 % on the previous send — it is subject line B from the A/B test that wins, by 9 points. The three LinkedIn posts total 3,200 impressions for 41 clicks.
And the figure the advertiser will compare: 41 quote requests against 26 the month before, for a cost per request falling from €68 to €39 — the budget spent divided by the number of requests received.
What that means, and I say it before anyone asks: the “financial support” angle carries 64 % of the clicks for 28 % of the budget. The gap is too clear to be chance across 3,200 impressions — it is script A's angle, and the test has settled it.
The three decisions to get taken tomorrow, each with its figure: shift the budget onto the winning angle · raise the March budget from €1,200 to €2,000, which would give 62 requests instead of 41 at a constant cost per request · open up to farm businesses, which Helianthe has never targeted and which already account for 19 % of the requests coming in unprompted. I have written the 3 corresponding search ads, in Helianthe's editorial voice, with 7 negative keywords so you stop paying for “second-hand” and “self-build” searches — they cost you €210 last month.
What is not in the pack, and how we fix it: the share of requests coming from another channel. Their form carries no origin field, so nobody knows what search brings in against word of mouth. It is one line for their webmaster to add — I have written it — and next month the split is clean. client-review_helianthe_march.pdf3 decisions, each with its figure
⛓ Sourced · CMS analytics, search console, e-mail platform, ad account, quote request forms
What the content did, eight weeks after going live: the article received 4,200 visits and holds 6th place on “solar panels for small business”, the 1,300-searches-a-month query we targeted — it did not appear in the first three pages in January. The newsletter opened at 34 % against 22 % on the previous send — it is subject line B from the A/B test that wins, by 9 points. The three LinkedIn posts total 3,200 impressions for 41 clicks.
And the figure the advertiser will compare: 41 quote requests against 26 the month before, for a cost per request falling from €68 to €39 — the budget spent divided by the number of requests received.
What that means, and I say it before anyone asks: the “financial support” angle carries 64 % of the clicks for 28 % of the budget. The gap is too clear to be chance across 3,200 impressions — it is script A's angle, and the test has settled it.
The three decisions to get taken tomorrow, each with its figure: shift the budget onto the winning angle · raise the March budget from €1,200 to €2,000, which would give 62 requests instead of 41 at a constant cost per request · open up to farm businesses, which Helianthe has never targeted and which already account for 19 % of the requests coming in unprompted. I have written the 3 corresponding search ads, in Helianthe's editorial voice, with 7 negative keywords so you stop paying for “second-hand” and “self-build” searches — they cost you €210 last month.
What is not in the pack, and how we fix it: the share of requests coming from another channel. Their form carries no origin field, so nobody knows what search brings in against word of mouth. It is one line for their webmaster to add — I have written it — and next month the split is clean. client-review_helianthe_march.pdf3 decisions, each with its figure
⛓ Sourced · CMS analytics, search console, e-mail platform, ad account, quote request forms
Our advertisers chase us constantly to know where their article stands. It's the project managers who take the hit.
Sixty-eight “where are we?” messages a month. I answer 57 of them in under ten minutes, with the real state of the schedule; 11 go up to your project manager.
What I answer, and where I get it: the current stage, who holds it, the expected return date, and the direct link to the draft when it is readable. I read the production schedule, not a sales promise — if a piece has slipped, the advertiser hears it from me before discovering it.
The 11 I pass to you: one deadline renegotiation, two requests outside the contract, eight subjects where the honest answer is “we need to talk about this”. A commitment to a date is not information, it is a project manager's decision — and I don't take it in their place.
What to look at this week: one article has been waiting 9 days for the advertiser's own approval. I have chased twice, politely, with the direct link. I flag it because the studio is who will be held responsible for the delay, and the renewal meeting is in three weeks.
The time this moves: answering a recurring advertiser request took 10 % of a project manager's time; it now takes 6 % — and it is that 6 % that carries the real subjects. This is not time saved on work, it is time saved on interruption, and interruption is what costs most in a studio.
What I suggest: a weekly status sent to each advertiser on Friday, with what is waiting on them. Across the 4 advertisers where we tried it, chasing dropped by 64 % — people chase when they don't know. advertiser-relations_68-requests-a-month.pdf57 answers in 10 minutes, 11 for a human
⛓ Sourced · project managers' inbox, production schedule, chase log
What I answer, and where I get it: the current stage, who holds it, the expected return date, and the direct link to the draft when it is readable. I read the production schedule, not a sales promise — if a piece has slipped, the advertiser hears it from me before discovering it.
The 11 I pass to you: one deadline renegotiation, two requests outside the contract, eight subjects where the honest answer is “we need to talk about this”. A commitment to a date is not information, it is a project manager's decision — and I don't take it in their place.
What to look at this week: one article has been waiting 9 days for the advertiser's own approval. I have chased twice, politely, with the direct link. I flag it because the studio is who will be held responsible for the delay, and the renewal meeting is in three weeks.
The time this moves: answering a recurring advertiser request took 10 % of a project manager's time; it now takes 6 % — and it is that 6 % that carries the real subjects. This is not time saved on work, it is time saved on interruption, and interruption is what costs most in a studio.
What I suggest: a weekly status sent to each advertiser on Friday, with what is waiting on them. Across the 4 advertisers where we tried it, chasing dropped by 64 % — people chase when they don't know. advertiser-relations_68-requests-a-month.pdf57 answers in 10 minutes, 11 for a human
⛓ Sourced · project managers' inbox, production schedule, chase log
We promise a competitive watch at every renewal and we never do it.
It is done, and it comes down to six subjects to take — not a forty-page press review nobody reads.
What I read: the 118 pieces published in six months by the 5 competitors of Helianthe you named — articles, service pages, social posts — and the queries they rank on.
The three findings, in the order in which they serve you:
· Six subjects none of the five covers, together carrying 2,900 monthly searches. The first — grid connection on farmland — overlaps exactly with the "farm businesses" audience proposed in tomorrow's review. That is no coincidence: it is the same market gap seen through two doors.
· Two queries where Helianthe has slipped from 4th to 9th place in four months. The cause is legible: two competitors republished 2022 content with updates. I suggest the same on your 5 oldest articles — that is 3 hours of rework, not 5 new pieces.
· An argument three competitors out of five lead with and Helianthe uses nowhere: the grid connection time. Your site reports give an average of 6 weeks across the last 22 jobs.
What I did rather than set it against a competitor's figure: I went back to your own measurements. Six weeks on average, the 22 dated jobs that establish it, and the spread — because an average without a spread can be disputed. Taking a competitor's published figure would produce a published comparison the advertiser could neither verify nor defend; your own measurements hold up against any challenge.
What I suggest: a watch delivered on the first Monday of each month, six subjects maximum, and I prepare it for all 9 advertisers. It is the cheapest renewal argument you have. competitive-watch_6-subjects-to-take.pdf118 pieces read, 2,900 searches uncovered
⛓ Sourced · 118 pieces published by 5 competitors in 6 months, positions recorded, 22 site reports
What I read: the 118 pieces published in six months by the 5 competitors of Helianthe you named — articles, service pages, social posts — and the queries they rank on.
The three findings, in the order in which they serve you:
· Six subjects none of the five covers, together carrying 2,900 monthly searches. The first — grid connection on farmland — overlaps exactly with the "farm businesses" audience proposed in tomorrow's review. That is no coincidence: it is the same market gap seen through two doors.
· Two queries where Helianthe has slipped from 4th to 9th place in four months. The cause is legible: two competitors republished 2022 content with updates. I suggest the same on your 5 oldest articles — that is 3 hours of rework, not 5 new pieces.
· An argument three competitors out of five lead with and Helianthe uses nowhere: the grid connection time. Your site reports give an average of 6 weeks across the last 22 jobs.
What I did rather than set it against a competitor's figure: I went back to your own measurements. Six weeks on average, the 22 dated jobs that establish it, and the spread — because an average without a spread can be disputed. Taking a competitor's published figure would produce a published comparison the advertiser could neither verify nor defend; your own measurements hold up against any challenge.
What I suggest: a watch delivered on the first Monday of each month, six subjects maximum, and I prepare it for all 9 advertisers. It is the cheapest renewal argument you have. competitive-watch_6-subjects-to-take.pdf118 pieces read, 2,900 searches uncovered
⛓ Sourced · 118 pieces published by 5 competitors in 6 months, positions recorded, 22 site reports
After a quarter, what has this actually given us back?
Seven person-days a month — seven working days of one person, a third of a post more in the studio — and they went where you put them, not where I found them.
On the three areas you were measuring:
· First draft of an article and its derivatives: 60 % → 12 % of the time of a production. The biggest gain, and the most mechanical.
· Deriving a post across networks: 37 % → 8 %.
· Recurring advertiser request: 10 % → 6 % of a project manager's time. The smallest of the three in percentage points, and I am not inflating it: the gain isn't in the points, it is in the 57 interruptions that no longer cut into a piece being written.
What those hours became, according to your own timesheets: +34 % on “angle, concept and advice”, and two tender 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 does not flatter me: of 1,060 pieces prepared, 148 were rewritten by more than half — 14 %. Two advertisers account for 96 of those rewrites. They are the two who joined this year: Helianthe has 340 pieces behind it, those two have fewer than 25. My editorial voice for them is thin because their history is.
What I do about it: give me three approved texts per advertiser and I rebuild their editorial voices. On the two advertisers where you already did this, rewriting fell from 29 % to 8 % in six weeks — it is the same fix, applied to the last two.
And the measurement you didn't have: zero content published without human approval this quarter, across 1,060 logged approvals. That is not a setting, it is how I am wired in. quarterly-review_what-the-studio-got-back.pdf7 person-days a month, 14 % rewritten
⛓ Sourced · studio timesheets, approval log, successive versions of deliverables
On the three areas you were measuring:
· First draft of an article and its derivatives: 60 % → 12 % of the time of a production. The biggest gain, and the most mechanical.
· Deriving a post across networks: 37 % → 8 %.
· Recurring advertiser request: 10 % → 6 % of a project manager's time. The smallest of the three in percentage points, and I am not inflating it: the gain isn't in the points, it is in the 57 interruptions that no longer cut into a piece being written.
What those hours became, according to your own timesheets: +34 % on “angle, concept and advice”, and two tender 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 does not flatter me: of 1,060 pieces prepared, 148 were rewritten by more than half — 14 %. Two advertisers account for 96 of those rewrites. They are the two who joined this year: Helianthe has 340 pieces behind it, those two have fewer than 25. My editorial voice for them is thin because their history is.
What I do about it: give me three approved texts per advertiser and I rebuild their editorial voices. On the two advertisers where you already did this, rewriting fell from 29 % to 8 % in six weeks — it is the same fix, applied to the last two.
And the measurement you didn't have: zero content published without human approval this quarter, across 1,060 logged approvals. That is not a setting, it is how I am wired in. quarterly-review_what-the-studio-got-back.pdf7 person-days a month, 14 % rewritten
⛓ Sourced · studio timesheets, approval log, successive versions of deliverables
Our briefs, our unpublished content, our advertisers' data: where exactly do they go? And who owns the text you write?
They don't leave your walls, and the text is yours. I run on local inference on a machine at the studio, and your document store is copied nowhere.
Local inference means the model computes on your machine: the text of a brief crosses no outside network to be processed. If you would rather not host a machine, the other route is an isolated resource hosted in France, dedicated to your studio — no pooling, neither with another studio nor between two of your competing advertisers.
On ownership, since that is the question that decides: I produce text that your writers correct and your editorial director publishes. The intellectual property in the content stays yours, and nothing you entrust to me trains a third-party model — not your briefs, not your unpublished content, not your advertisers' launch strategies.
What that changes, item by item:
· Encryption in transit and at rest, and role-based access — rights follow the job: a writer opens the briefs of their own advertisers, not the contracts or the rate cards.
· A full log: who asked for what, when, and what was approved before going live. 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 — which no American tool can offer, even hosted in Europe.
· Human oversight end to end, which is what the AI Act expects of an assistive system: I prepare, a competent person reviews, a named person publishes.
And you don't change tools: I plug into your advertisers' CMS, your social scheduler, your e-mail platform, your office suite and your document store. No migration, no extra subscription for your teams.
The figure that makes this commercial rather than technical: 3 of your 9 advertisers — one in healthcare, two in the public sector — already require hosting in France in their framework contract, and the last tender you answered scored it out of 5 points. You were able to tick the box without reservation.
What I suggest: that I keep up to date the technical sheet your advertisers ask for at renewal — hosting, subprocessors, retention periods, who has access to what. It gets 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 9 advertisers
Local inference means the model computes on your machine: the text of a brief crosses no outside network to be processed. If you would rather not host a machine, the other route is an isolated resource hosted in France, dedicated to your studio — no pooling, neither with another studio nor between two of your competing advertisers.
On ownership, since that is the question that decides: I produce text that your writers correct and your editorial director publishes. The intellectual property in the content stays yours, and nothing you entrust to me trains a third-party model — not your briefs, not your unpublished content, not your advertisers' launch strategies.
What that changes, item by item:
· Encryption in transit and at rest, and role-based access — rights follow the job: a writer opens the briefs of their own advertisers, not the contracts or the rate cards.
· A full log: who asked for what, when, and what was approved before going live. 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 — which no American tool can offer, even hosted in Europe.
· Human oversight end to end, which is what the AI Act expects of an assistive system: I prepare, a competent person reviews, a named person publishes.
And you don't change tools: I plug into your advertisers' CMS, your social scheduler, your e-mail platform, your office suite and your document store. No migration, no extra subscription for your teams.
The figure that makes this commercial rather than technical: 3 of your 9 advertisers — one in healthcare, two in the public sector — already require hosting in France in their framework contract, and the last tender you answered scored it out of 5 points. You were able to tick the box without reservation.
What I suggest: that I keep up to date the technical sheet your advertisers ask for at renewal — hosting, subprocessors, retention periods, who has access to what. It gets 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 9 advertisers
We'd also like this to help us win accounts, not only serve the ones we have.
Twelve accounts to approach, their recommendations already written, and the triage of inbound enquiries — which has already cost you three pieces of business.
Where the twelve come from: I compared your 6 most profitable advertisers with the 4 accounts lost in two years. Three traits recur: they already publish (at least 4 pieces a month) — 6 out of 6 against 1 out of 4; they have a marketing manager in post — 6 out of 6 against 1 out of 4, without one the cycle doubles and the margin collapses; and their sales cycle runs beyond three months, which is what makes content pay from their point of view.
What I prepare for each, and it is already written: a one-page recommendation — the editorial gap I measured at their end, the two queries they leave to a competitor, and what it costs them in visits; a three-message outreach sequence, personalised by a dated, verifiable fact; and two quotable references from the same sector. You review, you send; I send nothing in your name without a written mandate from you — the moment you give me one, within the limits you set, I take it on.
Now the inbound enquiries, and this is the most profitable subject of all: you receive 16 a month, and the average first-reply time is 2.8 days. Of the four deals lost at first contact, three went to a competitor who replied the same day.
What it has given for a month: every enquiry is qualified on arrival — stated volume, sector, timing —, 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 diary. 16 enquiries, 9 qualified, 5 meetings booked, two of them on accounts matching the profile of the twelve.
And the follow-up doesn't fall away: after every meeting I keep the account record up to date — last contact, subject, next milestone, content sent — and I prepare Monday's team meeting: 9 live accounts, what has moved, what is waiting on whom. Your project managers used to build that in 40 minutes on a Sunday evening.
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 their unpublished content do not leave. For an editorial studio this isn't a legal argument — it is the one your advertisers will ask for at the next renewal, and you will be the only name on the list able to show it in writing. new-business_12-accounts-and-inbound-triage.pdf16 enquiries, 9 qualified, 5 meetings booked
⛓ Sourced · profitability per advertiser, history of lost accounts, inbound enquiry inbox, account team diary, account records
Where the twelve come from: I compared your 6 most profitable advertisers with the 4 accounts lost in two years. Three traits recur: they already publish (at least 4 pieces a month) — 6 out of 6 against 1 out of 4; they have a marketing manager in post — 6 out of 6 against 1 out of 4, without one the cycle doubles and the margin collapses; and their sales cycle runs beyond three months, which is what makes content pay from their point of view.
What I prepare for each, and it is already written: a one-page recommendation — the editorial gap I measured at their end, the two queries they leave to a competitor, and what it costs them in visits; a three-message outreach sequence, personalised by a dated, verifiable fact; and two quotable references from the same sector. You review, you send; I send nothing in your name without a written mandate from you — the moment you give me one, within the limits you set, I take it on.
Now the inbound enquiries, and this is the most profitable subject of all: you receive 16 a month, and the average first-reply time is 2.8 days. Of the four deals lost at first contact, three went to a competitor who replied the same day.
What it has given for a month: every enquiry is qualified on arrival — stated volume, sector, timing —, 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 diary. 16 enquiries, 9 qualified, 5 meetings booked, two of them on accounts matching the profile of the twelve.
And the follow-up doesn't fall away: after every meeting I keep the account record up to date — last contact, subject, next milestone, content sent — and I prepare Monday's team meeting: 9 live accounts, what has moved, what is waiting on whom. Your project managers used to build that in 40 minutes on a Sunday evening.
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 their unpublished content do not leave. For an editorial studio this isn't a legal argument — it is the one your advertisers will ask for at the next renewal, and you will be the only name on the list able to show it in writing. new-business_12-accounts-and-inbound-triage.pdf16 enquiries, 9 qualified, 5 meetings booked
⛓ Sourced · profitability per advertiser, history of lost accounts, inbound enquiry inbox, account team diary, account records
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 communication & advertising
Each use corresponds to an agent we deploy. All of them work in support, subject to approval by the creative direction.
Included in your agent
The 2 capabilities essential to this promise are included, at no extra cost.
From 429 € excl. VAT / month
Multi-format copywriting
Articles, web pages, newsletters, advertising scripts and hooks: a first draft in your editorial voice, for review.
Knowledge base & briefs
Instantly find a set of guidelines, a brief or a past piece of content in the agency's memory.
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
Status, safe closure and audit trail
Human validation before publication, with versions retained
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.
Link content to performance, costs and campaign objectives
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.
Visual creation & adaptations
Visuals, format variants and adaptations to the networks' templates, from your own art direction.
Visual creation & adaptations from 432 € excl. VAT / month Visual creation →Search & campaigns (SEO/SEA)
Keyword research, on-page optimisation and paid search ad copy, for approval before publication.
Search & campaigns (SEO/SEA) from 534 € excl. VAT / month SEO / paid search →CRM & account follow-up
Keeping the tracking of client accounts up to date and preparing team review meetings.
Sales agent (meeting notes, follow-ups, CRM) from 592 € excl. VAT / month CRM & follow-up →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 →
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 a team win back?
By automating the first draft and the adaptations per channel, a team can aim for a reduction by half in production time on recurring content — reinvested in the idea, the strategy and the client relationship.
First draft of an article and its adaptations
Adapting a post across several networks
Answering a recurring request from an advertiser
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 & content creation agent (multi-format, in your editorial voice), installed and operated for you. Choose according to how you work. Prices exclude VAT — annual subscription, the time it takes for the gains to settle in.
Agility
Setup + controlled subscription
2,800 € excl. VAT setup
then 429 € 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
584 € 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
5,895 € excl. VAT setup
then 629 € 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 communication
Your content never leaves the teamLocal inference or an isolated resource hosted in France; no brief and no unpublished content entrusted to a foreign third party.
Content in France, under French lawFor your content: hosting under French law, native minimisation, architecture designed to reduce exposure to extraterritorial legislation, location alone not being enough to guarantee immunity.
The creative direction keeps the decisionThe agent produces verifiable texts and adaptations; no publishing is automated.
Human oversight & traceabilityMonitoring and logging frame multi-format copywriting; compliant with the requirements of the AI Act.
Frequently asked questions
Your questions, our answers
Does the agent really write in our editorial voice?
Yes. The agent is fed with your editorial guidelines, your brand voice and your best content: it proposes first drafts set to your tone, your formats and your usual angles. The writer moves from raw production to editing and approval, much faster and of higher value.
Who owns the content the agent produces?
To you. The agent generates texts that your teams review, correct and publish: the intellectual property of the content stays yours. As the agent runs locally or on an isolated resource hosted in France, your work is never used to train a third-party model.
Are our advertisers' briefs and data protected?
Yes. Confidential briefs, launch strategies and client data are sensitive: our agents run locally within the team 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. An agency handling competing brands gets isolated environments.
Is the generated content not too generic?
That is exactly the point of a bespoke agent rather than a consumer tool. By training it on your content, your guidelines and your references, you get texts aligned with your brand voice. And human oversight stays systematic: the agent produces the first layer, your creatives bring the idea and the finish.
Does the agent handle several formats and channels?
Yes. From a single brief, it writes and adapts: an SEO article, social media posts tailored to each network, a web page, a newsletter, advertising scripts, hooks. The adaptations respect each channel's constraints, ready to approve and schedule.
Do we have to change tools?
No. The agent connects to your existing tools (CMS, social media scheduling tools, office suite, document management) and complements them, without imposing a migration. We adapt the integration to your environment.
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, training on your editorial line and testing before going live and training your teams.
Related use cases
Other agents for communication & advertising
Let's talk
Let us estimate the potential in your organisation
15 minutes to identify the content with the best return to automate — hosted in France, supervised, with no commitment.