Your AI writes well. It knows nothing.AI with your company's context
An AI that knows what your company knows.
Your strategy, your offers, your target clients, your processes, the way you write. I gather them, keep them up to date, and give them to ChatGPT and Claude, with each person's permissions. Your team keeps its AI; it just stops re-explaining everything.
3 to 5 minutes · No credit card needed
The problem
Context is what your AI is missing.
It answers quickly and well. But for a company it doesn't know.
Re-explaining everything
Your offer, your target, your tone: every conversation starts with a paragraph of context, different depending on who writes it.
Gaps filled at random
Without your prices or your proof points, it invents brackets to fill in, or worse, plausible-looking figures.
Everyone their own AI, everyone their own version
A project here, a custom assistant there, last year's files. Nobody knows what the next person's AI has read.
What “context” means
What your AI reads at my place.
Your strategy
Vision, target clients, offers, pricing, competition: the 14 sections of the CED, written with you.
Your goals
What the company is aiming for this year, and where it stands, measure by measure.
Your organisation
Roles, responsibilities, job descriptions: who does what, and who decides.
Your methods
Processes, rituals, lessons learned, written by the people who practise them, and dated.
Your voice
Your editorial guidelines and each person's voice: it writes like your company, not like an AI.
“I already have ChatGPT.” Exactly.
I don't replace your AI: I give it what it lacks. Even if you drop your files into a project, the essentials are missing.
| What matters | Your AI alone | Your AI and your files (projects, custom assistants) | Your AI and Marcel |
|---|---|---|---|
| What it knows about your company | Whatever you re-explain, every time | The files you uploaded | The team's strategy, goals, organisation and methods |
| Who writes it | Nobody | You, alone, if you have the time | I interview you: 14 strategy sections, 66 document templates |
| Up to date | — | As long as someone re-uploads the right version | Every document is dated; the ones to re-check come up |
| For the whole team | Everyone on their own | One copy per project, to rebuild | A single version, written together |
| Permissions | — | The same access for everyone | Everyone sees what their role opens up, and so does their AI |
| From one AI to another | — | To redo in every tool | The same memory in Claude and in ChatGPT |
| Its sources | It fills in the gaps | Sometimes | It tells you what it read |
An example
“Prepare the proposal for the Saint-Paul clinic.”
The same request, in the same ChatGPT. Except it has read your target clients, your offer and your price list before answering, and it says so.
See the before / after →The safeguards
Your AI sees what you see. Never more.
Everyone's permissions
A team member's AI reads what their role opens up to them, nothing else.
Hosted in France
Your data is hosted in France. Marcel's AI is provided by Anthropic, with no training on your data.
Nothing without your say-so
Nothing goes into your strategy unless you add it. What sets the rule is for leadership to decide.
Your questions
Do I need to switch AI?
No. Everyone keeps their Claude or ChatGPT: you paste in a link, log in with your Marcel account, and you're done. The three steps →
Why not just drop my documents into a ChatGPT project?
It works for one person, for as long as the files stay accurate. What's missing is what makes company context: a single version written together, kept up to date, with permissions by role, and the same one in Claude and in ChatGPT.
What if I have nothing written down?
That's the case for most companies. I start from your website and your documents, suggest the answers, you confirm: a 20 to 30-minute session is enough to get going.
How do I know what the AI has read?
It tells you: every answer built with Marcel cites what it consulted. You check the source instead of taking its word for it.