Car-sharing platform
Getaround : EN + FR one knowledge base
designed with the VP Product, as an advisor
Fintech scale-up
Fintech scale-up: a governed knowledge base and 80% more reliable AI answers
My client is a B2B fintech scale-up that builds a financial management platform for businesses. Its teams brought me in to rebuild their knowledge base and make it a source of truth again, for their 300 employees as well as for their AI agent.
1 · Context
This scale-up builds a financial management platform for businesses. After a period of hypergrowth, the company entered a cycle of strategic consolidation, led by new leadership.
In an organization of 300 people, access to reliable information directly shapes the quality of onboarding, the transmission of culture and the operational alignment of teams. Notion has always played that role there, as both the knowledge base and the cross-functional workspace.
Over the years, the workspace lost its coherence, and the reliability of its information went with it. AI raised the stakes: Dust, the company’s conversational AI agent platform, draws on the same content. At this point, governing knowledge is less a documentation issue than a matter of decision quality.
2 · Challenge
The brief at a glance
- Restore a single source of truth
- Make the content that feeds AI reliable
- Give every page an owner
- Secure onboarding
- Establish lasting governance
The challenge was not to tidy up a wiki, but to restore a source of truth that teams and AI could rely on without second-guessing. The central wiki was largely out of date: more than 70% of its content was either no longer reliable or duplicated.
Years of growth had buried pages in endless sub-pages, shared private spaces and siloed team areas. With no clear landmarks, people fell back on their personal bookmarks or turned to other departments, and the Dust agents returned answers as uncertain as their sources.
The business stakes were twofold: secure the onboarding of new hires, who struggled to find the HR, operational and cultural resources they needed to become autonomous, and give the internal AI content it could trust again.
3 · Roadmap
- 1
Field discovery
Twelve interviews with the People, Product, Operations, IT and Strategy teams, some of them held on site at the company’s offices, to understand real usage and points of friction.
- 2
Information architecture
A new architecture built on clearer navigation, explicit content ownership and governance distributed across teams.
- 3
Central wiki rebuild
The central wiki rebuilt from scratch, from verified information, under a new name.
- 4
Content governance
Verification badges and review dates: every page has an owner, who is prompted to re-verify it when the date comes due.
- 5
Design system and templates
An on-brand design system and standardized templates, one for overview pages, the other for reference resources.
- 6
Change management
Training for Notion champions, video tutorials and an admin dashboard, to keep the system maintainable over time.
- 7
Adoption and refinement
Feedback from the teams to refine the system and anchor its adoption.
4 · The transformation
Before
- More than 70% of the central wiki out of date or duplicated.
- No recognized source of truth, for the teams or for the AI agent.
- Information found through personal bookmarks or by asking the People team.
- Onboarding slowed by a lack of reliable resources.
- No defined ownership for keeping content up to date.
After
- AI answers 80% more reliable: Dust draws on verified, structured, up-to-date sources.
- Half as many requests to the People team: information is one click away.
- A single, on-brand hub for every employee.
- Explicit governance: every page has an owner and a review date, with an automatic reminder when it comes due.
- Secure onboarding, and an ongoing collaboration.
Super structured, really crystal clear, so thank you, it meets the brief.
Let’s talk about your project.
Tell me a little about what you’re working on. I’ll get back to you personally to discuss the next steps.
More case studies
Luxury event cleaning
Cascades Propreté : 130 agents feeding the ESG reports
built by AI for every event
Industrial group
DC Groupe : 61 steps launched in one click for each product
100% of exchanges recorded on the project