If an AI tool went wrong this afternoon, could you switch it off?
Most firms couldn't say how, or who's allowed to. Get a free written plan for your own AI tools and connections, or have a tested one-button stop installed.
Why this exists
AI is connected to more than you think
AI tools now read inboxes, update records and talk to other software on your behalf. Staff connect them to company systems without anyone else knowing. When something behaves badly, the first question is what's connected to what, and in most firms nobody has written that down.
of UK accountants and bookkeepers say their practice uses AI in some way.
Xero, Cebr and Censuswide, 2025of UK private practices created an AI policy last year.
Tolley, 2024The free plan
A written shutdown plan, free
Answer a few questions about the AI tools and connections your firm uses, and you'll get a written plan back: what to switch off, in what order, who's allowed to do it, and where each setting lives.
What it can switch off, and what it can't
Exactly what one button covers
One button stops
The AI systems and automations I've built for you
Your connections to AI models and their API keys
MCP servers and the tools your AI agents are allowed to use
One button doesn't stop
AI features inside other companies' software, such as your accounting or case management system
Staff members' personal AI accounts
Anything that needs your own admin login to change
Those are written into the plan as steps for a named person, in the right order. I won't pretend a button can do them.
How the installed version works
One button, set up and tested
It's wired into your systems when it's installed, and lists every connection it controls.
You decide who is allowed to press it.
Jobs already running stop cleanly, and you can see where each one got to.
Every use is logged with who pressed it and when.
Switching things back on is a single step too, once you're happy.
The drill
Tested, with a record
Nobody knows a button works until someone has pressed it. The installed version includes a shutdown drill: we press it, check everything stopped, switch it back on, and you get a dated record. Repeat it on whatever cycle suits you. It's the kind of evidence an insurer or regulator will actually want to see.
Part of doing AI properly
Knowing how to stop a system is part of using it responsibly, whichever regulator you answer to. The Kill Switch sits alongside the AI policy and usage log from the governance kit.
About the team AI rolloutPrice
Free, or installed from £950
Questions