Everyone uses the same AI. So how do you stand out?
That’s the core question explored in episode #185 of Le Café du Market’, with Myriam Gabrielle Prot-Poilvet, a content strategy consultant and fractional head of content. Her answer is both simple and profound: competitiveness is no longer about which AI model you choose, but about the quality of the context you give it.
The problem everyone runs into
You’ve probably experienced this: you ask your AI something, and it gives you a generic, flavourless result that anyone could have produced. The reason? You gave it too little context. Or sometimes… too much.
LLMs have a context window, a limited capacity to process information. Too little, and the output is hollow. Too much, and the AI gets confused, starts making mistakes, and forgets things mid-conversation.
The challenge is finding the right balance: giving the right information at the right time. This is what’s known as context engineering.
Beyond the individual user: the team problem
Most advice on the topic is designed for a solo user. You’re told to put your files in Claude, organise your folders, name your documents in a certain way. Useful, but it falls apart the moment you scale to a team.
As soon as there are several of you, problems emerge:
- Everyone works from documents that have drifted apart
- Information gets out of sync
- Routines can’t be replicated
- Switching AI tools becomes a migration nightmare
To solve these problems, Myriam Gabriel built a centralised knowledge base in Notion, which acts as a context layer for all her AI tools: Claude, Notion AI, and whatever comes next.
The founding principle: single source of truth
Myriam drew on a concept from the 1980s: the single source of truth. The idea is straightforward: every piece of information should exist in one place only. No duplication. You link to it, you reference it, but you never copy it into multiple locations that could fall out of sync.
This principle led her to choose Notion as the foundation of her knowledge base, for several reasons:
- Compatible with Markdown, easily readable by AI tools
- Multi-user and synchronised in real time
- She uses it herself every day (no redundant parallel system)
- Version history available to roll back to previous states
- Connectors available to link AI tools directly to Notion
A structure built on 4 page types
The key to her system isn’t just using Notion; it’s structuring it with intention. She distinguishes 4 types of pages, each identifiable by a naming indicator:
1. Personal pages
Dashboards, consolidated views, summaries. Useful for her, but of little interest to AI tools.
2. Reference pages
Brand guidelines, ICPs (ideal customer profiles), positioning, tone of voice. These pages are stable and serve as a compass for all her output. An AI can consult them to make sure content aligns with the strategy.
3. Table pages (living pages)
Editorial calendar, KPIs, monitoring files, CRM, meetings. This data changes constantly. AI tools query them like a live database.
4. Instruction pages
This is where the real innovation lies. Rather than storing prompts inside Claude or a specific AI project, Myriam Gabrielle externalises her instructions into Notion. She builds entire processes, structured with steps, references, and conclusions, and when she creates a skill in Claude, she simply tells it: “Go and read the instruction on this Notion page.”
The result:
- When she improves an instruction, she updates it once in Notion, and all her AI tools benefit automatically
- She’s no longer tied to any one AI tool, since her instructions work regardless of the model
- Her prompts become a strategic asset, stored outside the silos of individual platforms
A system that improves itself
What makes this system particularly powerful is the feedback loop built into it.
At the end of every instruction, a synced block (which updates everywhere simultaneously) asks the AI to:
- Evaluate the quality of the result it produced
- Identify steps that caused problems
- Suggest improvements
If Myriam Gabrielle approves, the AI directly updates the instruction in Notion. And if the change is significant, it feeds into an AI journal: a table that logs what changed, why, which AI made the change, and what side effects were noted.
Every month, a verification routine makes sure everything stays clean, consistent, and up to date.
What it looks like in practice
Myriam Gabrielle uses this system for a wide range of tasks:
- CRM routine: the AI cleans HubSpot, creates deals, and updates KPIs without any human intervention
- Day planning (chief of staff mode): the AI cross-references the calendar, KPIs, and urgent tasks, then proposes a schedule with time blocks in Notion Calendar
- Prospect meeting prep: generates a content pre-diagnosis and a discovery meeting guide tailored to the client
- LinkedIn post creation: a collaborative working session with the AI, which doesn’t write for her but guides her step by step through angle, sources, structure, drafting, and evaluation
Freelance,
but never alone
A marketing centre of excellence where freelancers grow together, share, learn and contribute to every mission.
The takeaway
In the face of constant tool acceleration, Myriam Gabrielle offers a counterintuitive piece of advice: beware of FOMO. Chasing every new AI, every new feature, is exhausting and counterproductive.
True resilience means building a durable reasoning layer: an organisation of your knowledge that will remain useful whatever tool emerges in six months. The knowledge management principles of the 1980s are still valid. What changes is the surface they’re applied to.
”"There's a layer of modernity that changes every three days. But underneath, how do I build my own reasoning layer, one that's much more lasting?"
Myriam GabrielleHead of content part-time
Varied contexts, one shared commitment: creating impact.
Start-ups, scale-ups, SMEs, large groups: discover how we helped them grow, structure and perform.
Want to go further?
This article is inspired by episode #185 of Le Café du Market’, the B2B marketing podcast hosted by Axel Kaletka. Myriam Gabrielle goes into practical detail about how she built her system, with concrete examples of her Notion instructions and AI routines.
Listen to episode #185: Building effective AI: the real challenge of context
Le Café du Market’ is also a French-speaking community for B2B marketers, a place to exchange with peers, grow, and accelerate your marketing career.
