AI how-to guides Knowledge and documents
How to keep expert knowledge when people retire
Find the roles where critical know-how sits with one or two people, capture it in AI-assisted sessions while they are still there, and turn it into verified checklists, guides and answers others can search. The experts and their peers validate what is captured. Start with the most exposed roles and keep the knowledge alive with owners and review dates.
What AI does
- Spots knowledge concentration risk: domains where critical expertise sits with one or two people or is barely documented.
- Guides interview sessions with the expert, asking follow-up questions until the reasoning behind each step is clear.
- Turns transcripts, emails and documents into structured checklists, procedures and question-and-answer pairs.
- Tags and links the knowledge so people find the right answer when they need it, with the source.
- Flags gaps from the questions people ask that the knowledge base cannot answer.
What stays with people
- Deciding which roles and domains to prioritise.
- Explaining the judgement behind the work: the expert's time and goodwill are the core input.
- Validating that what was captured is correct, with the expert and peers.
- Owning each domain after the expert leaves and keeping it up to date.
How to put it in place
-
Map where the knowledge risk is
List critical roles and processes, who knows them and when those people plan to retire, rotate or leave. Prioritise the domains where knowledge sits with one or two people and is hard to find anywhere else.
-
Agree time and format with the experts
Explain why it matters and agree how much time each expert can give and in what format: interviews, walkthroughs of real cases, recordings while they work. Without their buy-in there is nothing to capture.
-
Capture with guided, AI-assisted sessions
Run sessions on real cases and decisions, not abstract descriptions. Use AI to transcribe, ask follow-up questions and draft the first structure, and pull in the emails and documents the expert points to.
-
Structure and verify before publishing
Turn the material into checklists, guides, short videos and answers to common questions, not raw transcripts. Have the expert and a peer validate each piece before it goes live.
-
Test with successors and keep it current
Let new joiners and backfills use the knowledge on real work and fill the gaps their questions reveal. Assign an owner and a review cadence per domain, and trigger reviews when processes change.
What you need
- Sponsorship from the business so experts get time to share what they know.
- A list of critical roles with retirement, rotation or departure risk.
- Access to the documents, emails and systems where the work happens, with consent and permissions.
- A searchable repository with owners, versions and review dates.
- People who will take over each domain and validate the content.
When it is not worth it
- When the work is already well documented and several people know it.
- When the process is about to be redesigned or retired: capture the new way, not the old one.
- When experts are not given time: AI cannot extract knowledge from someone who is not available.
Risks and how to control them
- Capturing what was said, not what is right: verify with the expert and a peer before publishing.
- Knowledge that goes stale as processes change: give each domain an owner, review dates and change triggers.
- Personal data and confidential information in recordings and emails: get consent, limit access and remove what is not needed.
- Experts who feel replaced: present it as passing on their legacy, involve them in the design and recognise their time.
How Thinkia does it
We prioritise critical roles and retirement or rotation risks, and agree time and format with the subject-matter experts. Tacit steps become checklists, videos and questions and answers with peer validation, not raw transcripts, and new joiners and backfills use the content so gaps are filled from real questions.
It runs on Thinkia Synapse, with knowledge gap mapping and alerts when critical expertise sits with one or two people. Ownership, review cadence and triggers when processes change keep the knowledge current, and capture continues in daily work, not only at offboarding.
The Thinkia solution behind this guide
- AI knowledge transferInstitutional knowledge that stays when people leave.
- Enterprise Knowledge AIYour organisation knows more than it can find.
Thinkia products involved
Frequently asked questions
How do you capture tacit knowledge with AI?
By working through real cases with the expert in guided sessions, where AI transcribes, asks follow-up questions and drafts a structure. The tacit part comes out when people explain why they made each decision, not when they describe the process in general.
Can AI replace the expert interviews?
No. AI speeds up transcription, structuring and search, and it can mine emails and documents, but the judgement behind the work is in people's heads. Their time is the key input.
When should we start?
Well before the person leaves. Capturing knowledge while the expert still works alongside their successor lets you test and correct it on real work.
Isn't this just a knowledge base?
A knowledge base is where it ends up. The difference is the capture: identifying risk, guided sessions, verification by experts and peers, and keeping it current once they have gone.
Is there a standard for knowledge management?
Yes. ISO 30401 sets requirements for knowledge management systems, and it is being revised. It is a useful reference for roles, processes and review, whatever tools you use.
Keep exploring
Related guides
- How to build a company knowledge base with AI
- How to automate employee onboarding with AI
- How to manage company documents with RAG
Related decisions
- RAG vs enterprise search: do your people need answers or documents?
- GraphRAG vs vector RAG: when does your knowledge assistant need a graph?