You Don't Need Better Tools. You Need Context Offloading.

Individual: The 20-Minute Context Download

That SOP you have been meaning to write? (Yes, I have been there.)

Open a voice conversation with AI and talk through the process as if you are onboarding a new hire. Include the edge cases and the judgment calls you make without thinking about them.

Then ask AI to structure it.

You have just moved reasoning out of your head and into AI context, in 20 minutes. This is Workflow Literacy in its simplest form: you cannot delegate a process you have never made explicit.

One-on-One: The Context Interview

Once a month, replace a regular one-on-one with deliberate sensemaking. Ask three questions:

• What decisions did you make that had the most impact?
• Where do you rely on intuition?
• What would break if you were gone for 30 days? (I love this one.)

Record it, transcribe it, and have AI distill it into a structured brief.

Over time you build a searchable map of how your key people actually think and work.

Group: The Strategic Context Session

One monthly leadership meeting framed around exploration, not updates. No slides, just free-flow discussion that enriches your AI's understanding of the business.

Start with a provocation. What are we doing that might already be outdated?

Capture the discussion, then use AI to extract themes, tacit knowledge, and strategic implications. The goal is to record not just what people do, but why they do it.

Data is largely in people's heads, and we need to get it out. That is the whole argument.

So build the pipeline: a sense of ambiguity, then a voice recording, then context that lives in a system your core AI can reach. (I mostly use ChatGPT plus the apps that connect to Slack and Google Drive, which is where most of our context sits.)

This is also what the research keeps pointing at. MIT's Project NANDA found in its 2025 report The GenAI Divide that 95% of the companies it studied saw no measurable impact on profit and loss from their generative AI work. The authors are explicit that the cause is not model quality. It is a learning gap: generic tools work beautifully for individuals and then stall inside companies because they do not learn from or adapt to how the work actually gets done. (Source: MIT Project NANDA, The GenAI Divide: State of AI in Business 2025. I wrote about that failure rate in more detail in why 95% of AI implementation fails.)

The companies that win will be the ones that systematically get knowledge out of people's heads and into their systems.

So stop guessing, and start offloading context.

1. Generic output is a context problem, not a model problem. Switching tools will not fix it.
2. Polanyi's point still holds. The most valuable knowledge in your organization is the hardest to write down, which is exactly why nobody has written it down.
3. Start individual. One 20-minute voice download of one undocumented process, this week.
4. Then make it a rhythm. Monthly context interviews with your key people, and one monthly leadership session built for exploration rather than updates.
5. Put it somewhere your AI can reach. Context that lives in a folder nobody connects is not context.

This is the work we do with leaders in AI Leader Advanced, where the whole point is making your own thinking usable by the tools you work with.

Daan