ARTICLES
Artificial intelligence, law and professional judgment
I have written about artificial intelligence and legal practice since the first edition of Andrew and the Marvellous Analytical Engine. Some of those pieces have survived the intervening year rather better than others, which is perhaps appropriate for writing about a technology notable chiefly for refusing to remain still.
The articles collected here have been selected and, where necessary, substantially revised to accompany the second edition. They deal with the questions which seem to me most durable: how lawyers misunderstand AI, how to instruct it, where responsibility lies, what happens to evidence and trust, how legal practice changes, and what the technology may mean for the economics of costs and fixed-cost litigation.
They are not intended as a continuously updated news service. They represent the position at the publication of the second edition in 2026.
What Lawyers Get Wrong About Artificial Intelligence
The useful questions are not whether the machine thinks or whether lawyers should fear it, but what task it is doing, what can go wrong and how the result will be checked.
Programming in Plain English
Prompting is less a collection of magic phrases than the old professional discipline of giving clear instructions, controlling the material and checking the result.
Artificial Intelligence Regulation Is About People, Not Machines
The important regulatory questions remain familiar: who chose the system, who checked the work, who supervised it, and who answers when it goes wrong?
Artificial Intelligence, Evidence and the Problem of Trust
Synthetic evidence changes not only how false material is made, but how genuine digital material earns trust through provenance rather than appearance alone.
Remaking Legal Practice
As AI becomes ordinary and disappears inside legal software, its largest effects may be on fees, training, professional value and the meaning of judgment.
Artificial Intelligence, Costs and Proportionality
AI may reduce production time while making excess easier. Costs law will still have to distinguish efficiency from unnecessary work, and time from value.
Fixed Costs and Thin Margins
In low-value PI the durable AI opportunity is not an autonomous robot solicitor, but a disciplined production process with human intervention where judgment becomes necessary.