The holiday handover
Holiday season is underway. Someone on your team is about to disappear for three weeks, and the plan is a handover call on their last afternoon.
Hand a colleague a documented playbook of the role instead. All the processes. The key contacts. The decisions you make and how you make them. Three weeks away gets easier for both of you. Same for sick leave, which nobody plans for at all.
I built the handover agent for people leaving jobs. When I left Aware, I used it to document the entire People function: 13 chapters, roughly 15 hours of my time over two weeks. The external HR agency taking the function over told me it was the best handover they'd ever received from a client.
Four months of client work later, I realised it works also for anyone who wants to document their work and their responsibilities in a structured way. It's a way to analyse your own work and find the things you could do better. Good preparation for planning, goal setting, or a performance conversation. When I ran it on myself, it turned into a way of taking stock of the whole role. It sets a baseline you can come back to in a few months, refresh, and keep current.
Why businesses should care
For a business, high fidelity knowledge about how the work actually gets done is gold. It secures continuity. When someone leaves, temporarily or permanently, you get expert knowledge at a level no exit interview will produce.
But think ahead and other use cases open up. Use the documentation to identify automation opportunities and strategic initiatives.
It also leaves you with a clean set of documentation for audits, due diligence, and governance work.
There's a people development case too. Completing my own handover playbook, I was struck by how much work I had done, the wealth of operational knowledge I had. It was a valuable opportunity to zoom out and holistically reflect on the value I had been creating and the number of plates I'd been spinning. These interviews were an opportunity for me to take stock. They gave me a basis to talk about my achievements and responsibilities as I prepared for job interviews and talked with my network. Run this across a team and you get people who can describe their own work accurately, which is more than most managers get in a performance review.
Not Bossware
The agent is human-in-the-loop first. The employee doing the interview controls what goes in and what comes out. Every draft gets signed off by the user. You connect the systems you want to connect and nothing else.
The agent asks questions one at a time. It follows up when an answer is thin. It asks for a contrasting example when one example isn't enough. Then it gives you a draft. You edit it, approve it, and enrich it with related links from your own files.
Any stack works
The process works regardless of the tech stack. As a personal tool in ChatGPT or Claude, as I saw in January. But also as a shared tool for a team or an organisation, i.e. as a project or a gem. For a client working in a Microsoft environment, I've built it in Copilot Studio, with Power Automate handling the file read and write. I've also built a version on Mistral AI Studio, ideal if you need an EU-headquartered model.
You implement this in the stack you already use. Your data is controlled and routed exactly the way it is today. There's nothing to change, because the method comes to your environment rather than the other way round. Access and privacy stay with the processes and tools you already run. Nothing new to install. You control this, and you keep controlling it.
See it working
I've put together examples of the agent running in four different setups. The videos are on the new site: handover.works.
I'd like to hear what you think. If you've got a view on exit management or knowledge management, I'm happy to talk.