Swimming with sharks
No business should organise itself around whatever technology happens to be in vogue. Applying this to AI, there’s far too much happening for any small practice to follow – and there’s no need to try. Sorcha Corcoran needs a bigger boat
When it comes to small legal practices adopting managed artificial intelligence (AI), the irony is that it promises great things in terms of saving time, but, at present, many principals simply don’t have the time or headspace to come up with an AI strategy, let alone actually try out any of the tools.
Getting beyond the potential risks involved, primarily around privacy and confidentiality, often seems too much of a mountain to climb, and it’s getting in the way of seeing the opportunities.
Simply knowing where to start is often the biggest obstacle.
“There’s far too much happening in AI for any small legal practice to follow, and there’s no need to try. A business shouldn’t organise itself around whatever technology happens to be in vogue,” says John O’Shanahan, founder of LeanBPI, a consultancy that supports Irish SMEs in improving productivity through practical digital adoption, process improvement, automation and, increasingly, AI.
He advises taking one problem within the firm and starting from there.
It could be that too much of the week goes on monitoring company emails, or maybe preparing a particular document means finding and copying the same information from multiple places.
“I would keep coming back to four questions:
- What problem are we trying to solve?
- What will this actually give us?
- What are the risks?
- What still needs human judgement?
“If you can’t answer those, don’t use AI for that task,” he says.
“A firm that gives a few hours a month to understanding one tool, testing one use-case, and talking to other practices about what they’re doing will be in a much better position after a year than a firm that tries to implement an AI strategy in one go.”
We are the champions
For solicitor and notary Dave Lane, someone within a small legal firm taking proper ownership of managed AI adoption is really important, before doing anything else.
Generally referred to as ‘AI champions’ across sectors, such individuals act as internal advocates, helping teams move from basic experimentation to safe, repeatable and practical use of the technology.
Last year, Lane was keen to take on that role at McGroddy Brennan Solicitors, where he was working at the time as a senior associate (see below: ‘Don’t stop me now’).
Concerns about unmanaged or ‘shadow’ use of AI were mounting in the profession – especially in terms of employees uploading documents to their phones and risking GDPR breaches.
“To me, effective managed adoption has always been a matter of selling AI to staff from within the firm. At McGroddy Brennan, that meant me showing everyone the practical use-cases of AI every day, starting with things as basic as drafting an email,” he says.
Influenced by his fund-accounting background, he developed a strict AI policy and devised a checklist approach – all the while ensuring adequate controls were firmly in place.
Before a document was uploaded to an AI model, any private information was redacted and the model was strictly told not to use personal information for data processing.
“We only used the pro versions of AI tools – ChatGPT for drafting emails and letters and Perplexity AI for research. I provided guidance on what model to use for what task.
"I kept it simple, always warning that the priority was confidentiality, and once a document had been created by AI, it had to be checked and reviewed by a solicitor before being issued,” Lane explains.
The take-up was good at McGroddy Brennan, as people started to see the benefits of using the tools, particularly in relation to document review and administration.
“They were getting reviews done within half an hour that would have taken half a day before. Putting in a simple direction or prompt into ChatGPT meant they could take the output as an initial draft and edit accordingly,” says Lane.
Now I’m here
In a professional services business that John O’Shanahan worked with, staff were given AI-generated first drafts, which they then reviewed and personalised.
The time spent on email administration came down by around 50%. “That process translates very well into a legal practice,” he says.
“One thing I’ve seen over and over, in every sector, is that confidence comes from doing rather than discussing,” he adds.
“The moment one process visibly improves, and staff can see they’ve got time back, they start bringing the next opportunity themselves. That’s the point where adoption stops being something you’re pushing. I expect AI in legal practices to follow that same pattern.”
Guy Fagan is a digital consultant with over 20 years of experience. He runs Guy Fagan Digital Consultancy, which focuses on digital strategy, web development, and AI integration for businesses in various sectors.
“I spend my day looking at workflows, communication, and operational efficiency with clients. Last year, I noticed that, increasingly, questions were coming from people working in the legal profession, specifically focused on what they do, day to day,” he says.
“Clients were expecting faster replies. They were dealing with huge workloads, and the capacity [to adopt a formal AI project] was a major constraint.”
Seven seas of Rhye
Drawing on all of this interaction, Fagan identified what the ‘AI pattern’ looks like in many small and medium-sized legal practices today, to help those in the profession to navigate the situation they face:
- AI as a first-draft and first-pass tool: AI can produce a draft letter, a skeleton argument structure, a summary of a document bundle, or a proposed checklist. Then the solicitor revises, verifies, and applies professional judgement. “Using AI in this way is good in terms of reducing friction, so solicitors can spend more time, day-to-day, on judgement, client care, strategy and outcomes,” Fagan says.
- AI as a ‘compression’ tool: summarising long documents and producing chronologies can reduce the time spent getting to the core issues. This is especially helpful for smaller teams managing multiple live matters. “AI summates well, but remember, you’re not asking it for a decision – the human element, judgement aspect, and safety are crucial here,” Fagan notes.
- AI inside practice workflows – not just in a browser: rather than copying client material into consumer tools, many firms are looking for AI features embedded in case- or document-management environments because ‘where data goes’ is the real question.
- AI in communication: AI can help legal professionals to explain complex issues in clean and clear language. That clarity can never come at the expense of accuracy. Used carefully, AI can help to produce clearer drafts, better structure and more consistent tone – as long as human review is the gatekeeper.
“We know that AI is good at producing language, but it’s the professional’s job to make sure that language is true and appropriate,” Fagan says.
“I do find that a lot of firms are using AI for client-friendly explanations in plain English. But you can’t build trust and relationships with raw outputs from ChatGPT.
"Clients need to feel that solicitors are giving a sense of themselves. You don’t want soulless responses; the humanity has to come across or client retention will suffer.”
Hammer to fall
Guy Fagan reckons any small legal practice that hasn’t formally embraced AI yet can’t go wrong if they start with these four steps:
- Ask what AI tools staff are currently using. Ban confidential client data from unapproved tools. Pick one low-risk workflow to pilot.
- Create a one-page AI policy, keeping human sign-off central.
- Train staff about the dangers of hallucinations and importance of confidentiality, verification, and human checking.
- Review the tools before ever using client data.
Don’t stop me now
Two years ago, Dave Lane took it upon himself to study AI and get proactive about it within the firm he was in at the time, McGroddy Brennan Solicitors.
“There was talk in the profession about jumping on the bandwagon. I wanted to look under the hood. I believed it wasn’t just a fad, but was going to be quite a significant technology,” he says.
A solicitor and notary public, with more than 20 years’ experience in the Irish legal profession, Lane has completed three AI-related diplomas; AI training by Google, the Compliance Institute, and the Law Society; as well as starting an MSc in artificial intelligence.
One of the biggest lessons Lane has learned is that ‘prompt engineering’ is key. This is about structuring, refining, and optimising the text instructions you give to AI models to maximise the output.
“An average prompt is seven to 11 words, but that’s not good enough. The more thought you put in to your prompt, the harder AI has to ‘think’ about it and the better output you’ll get.
"For example, the other day, I uploaded a commercial lease to an AI model to review – properly redacted and compliant with standards – using my own checklist, and got the response in 25 seconds.”
Not everyone is going to have the inclination to educate themselves to the extent that Lane has. But any firm deploying tools needs to bear in mind that basic AI literacy is now a legal requirement under the EU AI Act.
In other words, they must take practical steps to ensure their staff have an appropriate understanding of AI, its risks, limitations, and proper use.
Sorcha Corcoran is a freelance journalist.