Three numbers about AI in law, three companies selling AI to law, and nobody in between

A number about how much time AI saves a law firm has to come from somewhere. Following three of them back to where they started produces the same journey three times.

The first is five plus hours saved per lawyer per week at Talbots Law, on title checks. It originates in a case study published by Orbital, the platform Talbots bought. The trade press republished it with a line at the foot recording that the article was submitted by Orbital. Underneath the figure is one person saying that if you do a lot of title checking it could easily save five plus hours.

The second is that 84 per cent of UK solicitors save up to six hours a week on drafting. It is published by LEAP, which sells practice management software to law firms.

The third is the collapse in entry-level legal hiring, which comes from GenieAI, which sells legal AI.

Three claims. Three suppliers. Each one is a company describing the effect of its own product on the market it sells into.

The road they travel

The pattern is identical and it has four stages. A supplier produces a figure about its product. The trade press republishes it, sometimes with a disclosure line and sometimes without. The figure detaches from its origin and starts appearing without attribution. It arrives in marketing material, in pitch decks and in the reasoning of firms deciding how many people to hire.

At no stage does an independent party measure anything. There is no point on that road where someone with no commercial interest counts how long a task took before and after.

The exception, and what it shows

GenieAI's work on hiring is the strongest of the three, and this paper described it wrongly on 1 September. The company analysed job listings across Adzuna, Reed and LinkedIn, three of the largest boards in Britain, and found paralegal vacancies down about 21 per cent over the year and more than 60 per cent since 2023, consistent across all three platforms. That is a real method on real data.

It is also, still, a company measuring the market it sells into, and measuring vacancies rather than time. It tells you fewer paralegal jobs are being advertised. It cannot tell you that AI is why, and the company's own framing attributes the fall to AI adoption.

Separately, GenieAI surveyed its own customers and found 78 per cent of in-house legal leaders saying AI now handles work that once needed junior hires, and 44 per cent saying they had paused or reduced paralegal hiring for 2026. That is a survey of people who bought the product, asked about the product.

What would settle it

The apparatus to answer this already exists. Conveyancing has had a published price series since the SRA transparency rules came into force in December 2018, because a regulator required firms to publish what they charge. Prices are collected. Time is not.

Somebody could take a sample of firms, measure the hours a matter takes before and after a tool arrives, and publish the result with a method attached. It would not be cheap and it would not be quick. It would also be the only figure in this field that did not belong to a company with something to sell.

Until then, every number in circulation about AI and the speed of legal work in Britain is a number published by somebody who benefits from it being large.