An applied AI research lab in London.
We build software that takes on real work, and we study what it takes for that work to hold up inside a real organisation. ARC is what that has produced so far.
The problem was never the model.
Any model can produce an answer. Getting the job done is everything around it: the data, the rules, the exceptions, the approvals, and what happened last time. Most of that lives in people's heads and inboxes. Software that wants to do the work has to hold it.
Tools that help with a step hand the work back. The value is in the whole job.
Where a call matters, the system stops and asks. Knowing when not to act is the hard part.
What a job needed, what happened, who signed it off. Kept, it makes the next job start smarter.
Three threads.
Runs whole jobs across media businesses, deployed by our engineers and proven on the customer's numbers before they commit. The product.
Arc →Underneath ARC: a record of what every job needed and what actually happened. The research question is how a system learns from work without being taught it, and how it knows what it doesn't know.
How to show a system works on a business's own numbers before anyone signs. Blind comparisons against the team as it is today, on the metrics they set. We'll write it up as we go.
Writing.
Six months ago we planted a flag. This is what we found when we went looking for the work, and what we built instead.
Read →11 March 2026Announcing Viable SystemsWe're building the missing infrastructure that organisations need to actually put AI to work.
Read →London.
Viable Systems Ltd. 71–75 Shelton Street, Covent Garden, London WC2H 9JQ. Registered in England and Wales, company no. 17052121.
We work with a small number of clients at a time. If you have a job for ARC, the best first conversation is about that job.