Every firm that asks what an AI review produces wants the same thing, which is a number. What does the review find inside a twelve-partner practice, what does it cost to act on, and what does the firm get back. Adliora does not name clients, so the honest answer is to build a firm on paper, state every assumption and show the working. What follows is a model. The figures are mine, the arithmetic is yours to check, and nothing here describes a real practice.
The firm on paper
Call it a twelve-partner practice in the north of England with forty-one fee earners and eleven support staff. Two thirds of the fee income comes from residential conveyancing and private client work, the balance from commercial property and employment. Turnover is 6.2 million pounds. The firm runs a mainstream case management system, Microsoft 365 across the business and a document bank going back eighteen years. There is no AI policy. Three fee earners already use a general assistant on their own accounts because it saves them an hour a week, and none of them has mentioned it to the COLP.
What the review turns up
The first finding is almost never the tools. It is the distance between what a firm believes about its data and what its supplier contracts say. On the model firm, two of the three personal accounts sit on consumer terms that let the provider train on whatever is typed into them, and one fee earner has been pasting in extracts from a matter file. That is not a technology problem to be fixed with a licence. It is a confidentiality exposure that has been running for four months with nobody in a position to report it, because nobody in the firm knows it is happening.
The second finding is the absence of a record. The firm cannot show, for any piece of work an AI tool has touched, that a named human read the output and took responsibility for it. Where the work was sound, and most of it is, that gap costs nothing at all. Where a citation turns out to be invented or a clause is quietly wrong, the firm has no way to demonstrate the supervision it will be asked about. The duty is not new and does not bend for a machine. The SRA Standards and Regulations require you to supervise the work you are responsible for, whoever or whatever produced the first draft.
The third finding is the opportunity, and it is duller than the marketing suggests. First drafts of routine correspondence, summaries of long documents, a first pass over a bundle before a human reads it properly. Not judgment, and not the parts a client pays a partner for.
Putting numbers on it
Here are the assumptions, and every one is an argument you are entitled to have with me. Twenty-five minutes saved per fee earner per working day on drafting and summarising. Forty-one fee earners. Two hundred and twenty working days in the year. That comes to 3,758 hours of released capacity, worth 544,910 pounds at a blended recoverable rate of 145 pounds an hour.
Capacity is not revenue, and that number is where most vendors stop. A firm turns released time into fees only if there is work waiting to fill it. Assume a quarter converts, which is optimistic for a busy conveyancing department and pessimistic for a growing employment team, and the model firm bills 136,000 pounds it would not otherwise have billed. Against that sit the costs. Licences at 30 pounds a head each month come to 14,760 pounds a year, and the review, a written policy and half a day of training take the first year to about 21,000 pounds.
So the model firm finishes year one roughly 115,000 pounds ahead, on a turnover of 6.2 million. That is under two per cent, which is the figure worth carrying away. Now break the assumption. If five per cent of the released time converts rather than twenty-five, the firm bills 27,000 pounds and remains ahead of its costs. The case survives being wrong by a factor of five, and that, rather than the headline, is what makes the work worth doing.
Where the model is weakest
The time saving is the soft part. Most published figures are reported by the people using the tool, and self-reporting flatters. Halve it if you are cautious. The conversion rate is the other soft part, and it turns on client demand rather than on any decision the partners make. Neither weakness touches the first two findings, which carry no upside, only loss avoided. A firm that closes its confidentiality exposure and starts recording human sign-off gains nothing on the profit and loss account and removes the two things most likely to end badly.
Put your own numbers where mine are. Your fee earner count, your blended rate, and your honest view of how much released time turns into billed work. If the answer still clears the cost by a comfortable margin the decision makes itself, and if it does not, you have saved yourself a project.
The two-minute readiness check asks seven questions and gives you a private view of where your firm sits. Nothing is sent or stored. Send me your own figures and I will tell you plainly whether the case holds, including when it does not.