Phil Patterson

There is something reassuring about a solicitor’s office. There are the books, the certificates and the comforting impression that somebody in the room has actually read the terms and conditions.

You are there because they know something you don’t. They have spent years studying it, practising it and developing the judgement to apply it. You pay for access to that knowledge. It has been a perfectly reasonable arrangement.

The awkward development is that an awful lot of the information is now accessible through the device your client uses to order a takeaway. They can ask it to explain a clause, summarise a document or produce a first draft. These are not hypothetical capabilities. They are established uses of generative AI.

That does not turn everyone with a mobile phone into a solicitor. Nonetheless, the secret is out. Access to information is no longer the advantage it once was. Knowing what to do with it, and accepting responsibility for the result, is becoming the more important distinction.

I also run UK Trade Jobs, which gives me reason to think about the value of different kinds of work. There is a curious irony here. We spent decades encouraging people to leave manual work behind and pursue knowledge-based careers. Now a chatbot can help draft a contract, but it remains decidedly unwilling to come round and fix your boiler.

The professions should probably pay attention.

Sacred Ground

There can be few industries dealing with information as sensitive as the legal profession. People disclose the things they cannot tell their families, their employers or, occasionally, themselves. Legal professional privilege is not an inconvenient technicality. It is fundamental to the confidence that allows those conversations to happen. Introducing AI must strengthen that confidence, not quietly undermine it.

The Law Society of Northern Ireland’s guidance is clear about protecting confidentiality and privilege, understanding what happens to information entered into AI systems, and maintaining professional responsibility. These are obligations to carry into a new way of working, not reasons to avoid one altogether.

Firms should be able to explain which systems they use, what information those systems receive, who can access it and who checks the work. Buying a paid subscription is not, by itself, a security policy. Both free and paid tools require scrutiny of their contractual and technical safeguards. Tell clients what you are doing. Explain the controls. Give somebody responsibility.

“We use AI, we protect your information and your solicitor remains accountable” strikes me as a more confidence-building proposition than either pretending the technology does not exist or using it behind the client’s back.

The Billable Elephant

Fundamentally, large language models are useful at processing text and producing text. Summarising documents, preparing drafts and extracting relevant information are precisely the kinds of tasks the Law Society identifies as opportunities for greater efficiency. For an industry that does rather a lot of reading and writing, this should not require an especially lengthy strategic review. Much of the leg work can be reduced. Not every judgement, negotiation or difficult conversation. But a substantial amount of the work surrounding them.

And that brings us to the bill.

A corporate customer may previously have accepted 30 billable hours for preparing and reviewing a set of contracts. Where technology materially reduces that work, I would not build a business plan around the customer continuing to pay for the old process. Charge for expertise, responsibility and the value of the outcome. Do not expect the client to fund the preservation of unnecessary effort.

The same principle should apply to conveyancing. Where automation genuinely reduces the cost of routine work, some of that saving should reach the person buying the house. This does not mean every transaction is simple, or that professional responsibility has become cheaper. It means efficiencies should benefit somebody other than the partners.

Making legal services more affordable is not an unfortunate side effect of AI. It should be one of the objectives.

A firm that delivers good advice faster, at a more accessible price, has a fairly compelling proposition. Its competitors can disagree with the philosophy. They will still have to compete with the price.

But It Hallucinates

Yes, it does. In my own use, the improvement has been substantial and mistakes have become much less frequent. Nonetheless, “it sometimes gets things wrong” cannot be the beginning and end of the discussion. Which human being doesn’t? Which junior member of staff has never misunderstood an instruction, overlooked a detail or confidently produced something that needed correcting?

Humans have been generating confident nonsense for centuries. We have managed perfectly well without a data centre. Human error and machine hallucination are not identical, of course. But the answer to fallibility is scrutiny, supervision and accountability. We do not respond to a trainee’s mistake by concluding that employing trainees was fundamentally misguided. Nor should we compare a fallible AI system with an imaginary solicitor who has never made an error. Compare the actual processes. Check the sources. Measure the results. Keep a qualified person responsible. And, just as importantly, keep reassessing the technology. An opinion formed after a disappointing experiment several years ago is not a strategy for the next several years.

A Different Business

Through my work at Blue Canvas, I help businesses put AI to practical use. My view is that the opportunity is often larger than the obvious task everyone initially focuses on.
For a law firm, that means looking beyond drafting. There is an entire business surrounding the legal advice. Enquiries need answering, appointments need arranging, information needs collecting and clients need updating. AI can support administrative work and business planning as well as the legal documents themselves.

I would also be looking at business development: researching prospective clients, preparing useful communications and making sure a promising enquiry does not disappear into somebody’s inbox for a fortnight. Use it to reduce costs, certainly. But also use it to win work, improve service and give good people more time to do things clients actually value.
Then consider what those clients will need next.

I would be paying close attention to AI procurement, ownership of content and training data, employment changes, automated decisions, cyber incidents and responsibility when a system gets something badly wrong. These look to me like areas in which businesses will need thoughtful legal advice. At the same time, I would be cautious about relying on routine drafting and document production to generate the same fees indefinitely.

The work will change. So should the business.

This is not an argument for replacing every solicitor with a chatbot and leaving the office keys under a plant pot. It is an argument for being honest about what has changed.
Adopt the technology. Learn its limits. Protect the information. Be transparent with clients. Pass on meaningful efficiencies and build services around the problems that are emerging.

My money would be on the firms that do those things early and properly. Not necessarily the biggest firms, or those with the most impressive AI announcement. The ones that actually change how they work. AI is not going away. And your clients do not need your permission to use it.

Phil Patterson is founder of Blue Canvas AI