AI produces and measures, your team keeps the brand.
We set up an AI-assisted marketing chain in your company, from market research to measuring results. Every piece of content is reviewed by a person before it goes out, and every month we measure what it brings.
- Pillar · Marketing and sales
- Uses · 4 functions, 16 uses
- Method · 4Ps 3.0
- Facts · checked on 18 November 2026
- A published page commits the brand. Every piece of content prepared by AI goes through a person who answers for it, and every published figure has its source.
- We publish fewer pages, and each one gives the reader something.
Search now answers before the links.
Google launched AI Overviews and AI Mode in France, which answer the question before the list of links.
of searches with an AI summary ended with a click on a link, against 15% without one (900 US adults, March 2025).
Pew Research Center, July 2025
weekly users for ChatGPT in February 2026, according to OpenAI.
faster on writing tasks, with quality rated 18% higher, for 453 professionals helped by ChatGPT.
Since 2 August 2026, realistic images and videos created by AI must be labelled, and a chatbot must say it is an AI.
European Commission, July 2026
Sixteen uses of AI, from market research to measurement.
Understand the market
- Research
- Competition
Competitive watch
Every month, the agent records your competitors' offers, public prices, messages and customer reviews. You receive a summary of what has changed.
Public sources only, with no collection on platforms that forbid it.
- Research
- Customers
Your customers' voice
The agent reads reviews and support tickets to bring out your customers' needs, in their own words.
Names and contact details are removed before the analysis.
- Research
- Simulation
Personas and studies
The agent prepares the personas and questionnaires, then tests a first version on simulated respondents before the real study.
A study comparing simulated and real answers finds estimates that are often wrong, sometimes even in the opposite direction. Its authors see it as a complement to a real study.
Brand, Israeli and Ngwe, revised April 2026
A persona is checked with real customers before it guides a campaign.
- Offer
- Simulation
The simulated buyer panel
Twenty to a hundred buyer profiles react to your offer, price included, and say whether they would buy. The agent groups the reasons for saying yes or no.
A simulated buyer never pays. The panel rules out the weak versions of the offer, and the proof is still a customer who pays.
Produce the content
- Content
- Writing
Articles and newsletters
Starting from your team's notes and expertise, the agent prepares the outline and the first draft of each text.
A Harvard and BCG study of 758 consultants shows clear gains on tasks within AI's reach, and weaker answers on a task beyond it.
Dell'Acqua et al., HBS and BCG, 2023
An expert from your team reviews every published text and answers for its accuracy.
- Content
- Social
Social media
The agent writes the scripts for short videos and carousels, following an editorial calendar approved every month.
In French influencer marketing, an image retouched or created by AI carries the label “Images retouchées” or “Images virtuelles”.
French Ministry of the Economy, influencer marketing
- Content
- Creative
Visuals and videos
The agent produces the visuals, then adapts each creation to the formats of each network.
A realistic image or video created by AI is labelled as such, and no real person appears in it without their consent.
European Commission, July 2026
- Content
- Languages
Adaptations and translations
Each piece of content is adapted by channel and by language, with each audience's vocabulary.
A native speaker reviews each translation before it is published.
Get found
- Visibility
- SEO
Search optimisation
The agent audits the site and spots the missing pages. It prepares their first draft, with the internal links.
Google says no special optimisation is needed to appear in its AI Overviews, and that the basics of SEO still apply.
Google Search Central, December 2025
Google penalises pages produced at scale to manipulate its rankings, whether by a machine or by hand.
- Visibility
- AI
Citations by assistants
We track what AI assistants say about your brand, and the pages they cite. Your pages gain clear, sourced facts.
Nobody controls an assistant's answer. We track the citations obtained, without promising any.
- Acquisition
- Ads
Advertising
The agent prepares ad variants and tracks their results every day. The budget is still decided by your team.
Every claim is backed by proof, including the environmental claims governed by EU Directive 2024/825, applicable since 27 September 2026.
French Ministry of the Economy, Directive 2024/825
- Reputation
- Reviews
Customer reviews
After each purchase, the agent asks for a review. It prepares a reply to each review received and groups the recurring irritants.
No review is written in a customer's place. A fake review is a misleading commercial practice.
French Ministry of the Economy, online reviews
Convert and measure
- Conversion
- Pages
Pages and lead magnets
The agent prepares the landing pages and the downloadable guides, then tests two versions of the same page.
The cookie banner makes refusing as simple as accepting.
- Conversion
Email sequences
Each prospect receives a series of messages suited to what they have read and asked for. The agent writes the variants, and your team approves them.
Across 1,788 participants, messages tailored to each person's profile and written by ChatGPT were more persuasive than generic messages.
Matz et al., Scientific Reports, 2024
A consumer receives marketing emails only with their consent, and every message offers a one-click unsubscribe.
- Conversion
- Agent
The agent on the website
The agent answers visitors' questions from your approved pages, then offers a meeting or a guide.
Across 5,172 customer service agents, the AI assistant raised the number of issues resolved per hour by 14%.
Brynjolfsson, Li and Raymond, QJE 2025
The agent says it is an AI in its first message, and a person takes over the conversation on request.
European Commission, July 2026
- Management
- Results
Measurement
A single dashboard brings together traffic, leads, cost per lead and sales. Every month, we review with you what works.
A result is shown against a starting point. We record it before we start.
The 4Ps 3.0, in this order.
The method of the book “The 4Ps 3.0 of Go-to-Market” follows four stages. The most common mistake is to amplify before having paying customers.
A landing page and a demo video within a few days, to show something concrete.
Proof is a payment, such as a pre-sale or a signed contract. AI prepares the ground with the simulated buyer panel and the pre-order pages.
The first customers are served for real. We measure each stage and make corrections every week.
Once proof is in hand, the content machine starts, with advertising and work on AI citations.
Promotion comes after the first paying customers. Amplifying an offer that has never sold means paying to have it turned down by more people.
- We record the starting point, from traffic to AI citations, and we set up the access rights.
- The editorial calendar and the review chain are set up, with a person approving each piece of content.
- The first publications go out, with the first pages and the first email sequence.
- We measure the gap with the starting point, then you decide what comes next.
The four stages and their pitfalls are detailed in the book on the method.
- The brand's positioning and promise.
- The approval of each piece of content before it is published.
- Claims and their proof, from a product's performance to its environmental impact.
- The advertising budget and how it is split.
- The response to a crisis or a serious review.
Deepfakes: since 2 August 2026, realistic images and videos created by AI are labelled.
European Commission, July 2026
Influencer marketing: the label “Images retouchées” or “Images virtuelles”, under the French law of 9 June 2023.
French Ministry of the Economy, influencer marketing
Fake reviews: a misleading commercial practice.
French Ministry of the Economy, online reviews
Environmental claims: Directive (EU) 2024/825, applicable since 27 September 2026.
French Ministry of the Economy, Directive 2024/825
Cookies: refusing must be as simple as accepting.
Marketing email to a consumer: prior consent.
Mass-produced content: Google penalises pages made to manipulate its rankings, whatever the tool.
Simulations published on our site, with invented names. Each shows a marketing project and what stays with the teams.
Each retailer gets its product sheet, and both the posts and the replies to reviews are based on the approved sheets.
Posts that explain the treatments and their limits, with no before-and-after photos and no promised results.
Each listing goes out with its mandatory notices and its energy label.
On resale, the listing goes out with the new energy performance certificate and photos that do not mislead about the property.
On the marketing side, two real engagements.
One case study published with the client's consent, and one project we run ourselves.
The culinary encyclopedia and its menu and buffet engine
Four residences for sale, as off-plan units and investment packs
The method on this page, in books.
The 4Ps 3.0 give the order of the stages, and the two volumes of “16 leviers de vente” (in French) give the levers for each message. “Échouer en public” (in French) deals with reputation, which every publication puts on the line.
Does Google penalise content written with AI?
Google judges the value of a page, whatever tool wrote it. Its rules target pages produced at scale to manipulate its rankings, by hand as well as by machine. We publish fewer pages, and each one gives the reader something.
Should content created with AI be labelled?
For realistic images and videos, yes, since 2 August 2026. For text, the obligation covers content that informs the public on matters of public interest, unless it has been reviewed under an editor's responsibility.
How do you get cited by ChatGPT or in AI Overviews?
Google says no special optimisation is needed. Clear pages and sourced facts give assistants something to cite, and we track the citations obtained without promising any.
Does a simulated buyer panel replace market research?
It prepares it. Simulated answers are often wrong, and a simulated buyer never pays. The panel serves to rule out the weak versions of an offer before presenting it to real customers.
Who approves what is published?
Your team, or the expert you appoint. Until an agent has proved itself, nothing goes out without a person's approval.
How much does an engagement cost?
We do not display prices. The fee follows the free analysis, and we compare it with what the same work costs you today.