AI agents could take on continuing work in law firms, but Matthew Stringer, founder and CEO of Stridon, says firms should first check whether each task is repeatable, checkable, supervised, bounded and worth the cost. His practical advice: start with one defined process and keep a person responsible for reviewing what the agent does.
Watch Desk analysis
What happened
In an article published by Legal Futures on 12 November, Stringer distinguishes asking an AI to complete a single task from giving it a continuing objective, such as keeping a client briefed when defined events occur. That second arrangement needs clear triggers, limits and approval points.
His five tests for a proposed use are whether the work is repeatable, whether someone can check the result, whether a named person has time to supervise it, whether access is bounded, and whether the value justifies the cost. He recommends starting with one process, setting a budget and measuring accuracy, corrections and time saved before expanding. Read Stringer’s guidance.
Why it matters
An agent that can act across connected systems raises a different management question from a chatbot that drafts a response for someone to review. Stringer’s tests give firms a concrete way to decide what they might delegate and where approval must remain with a person.
He gives an example of preparing a client briefing for around £2–5, while stressing that checking and correcting the result also count towards the real cost. Cheap generation is not much of a bargain if the review takes longer than doing the work yourself.
Our read
The strongest advice is refreshingly unglamorous: make the task specific, limit access, give someone responsibility and measure the whole workflow. Start small enough to discover what goes wrong before the agent is given a standing brief and a long leash.
What to watch
- Whether firms begin with a defined process and publish how they assess its results.
- How much staff time review and correction add to the apparent cost of agent work.
- Whether firms set clear access limits, approval points and spending caps before scaling.
Discussion spark: Should law firms let AI agents act on continuing instructions once boundaries are set, or should every consequential step require fresh human approval?
Sources and evidence
- From chat to Cowork: preparing your law firm for agentic AI – Legal Futures (12 November 2026, 00:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.