Your AI’s memory is not your knowledge management system

When you move between ChatGPT and Claude, what happens to the context you’ve built over months of conversations?

A recent post by Jo Clubb about moving between AI platforms prompted me to share how we approach this at SportSync. After replying, I kept thinking about a distinction behind that conversation: an AI that knows you is not the same as a knowledge system you can manage.

They’re not the same thing.

An AI platform’s memory can make conversations more relevant. But I wouldn’t want that to be the only place where my working context lives.

There are two separate decisions:

Where do I keep my knowledge? And how do I make it useful?

The first is about access, portability and dependence on a provider.

The second is about how I work and what I create: methodologies, reports, presentations, reflections and the reasoning behind decisions.

Where does that work live? How is it organised and connected? And how do I build on it rather than start again?

Keeping Word documents and Excel files in folders on a Windows computer can be part of that foundation. But storage alone is not a knowledge management system, even when an AI tool can access those files. Access doesn’t, by itself, explain how they relate, which version is current or how they should inform the work.

In high-performance football, that means more than retrieving a session plan. It means reconnecting it with the evidence, experience and reasoning that shaped it, then using that knowledge to develop the next piece of work.

At SportSync, we use Notion as the foundation for this work. I’ve used it for six or seven years, initially in my own football practice. What I value is being able to adapt the structure as our work evolves.

But putting everything in Notion doesn’t automatically create a knowledge system either.

At SportSync, we use an approach we’ve been developing through our experience in high-performance football, our work with clients and what we’re learning as AI evolves. It connects four layers:

• Mapping processes and workflows.
• Building a structured knowledge base around the work.
• Defining governance through policies, protocols and working rules.
• Developing automations and agents with clear roles, context and boundaries.

The point isn’t simply to store more. It’s to connect what we know with what we create and how we work, so people and AI can use it in context.

Notion is the foundation we’ve chosen, not a substitute for that work or for thinking about access, exports and recovery.

That’s the distinction I keep coming back to: an AI remembering things about us is useful. Building knowledge we can review, reuse and bring into different tools is a different responsibility.

I explored those four layers in my previous article on AI agents in high-performance football. This conversation brought me back to the same foundation, from a different angle: changing the AI tool shouldn’t mean rebuilding our working context from scratch.

Where do the knowledge and work you build each week live, and how easily can you use them again?