AI for mid-size firms: 'It’s about not wanting to be left behind'

This Gazette roundtable brought together representatives from mid-size firms to discuss key findings and trends identified in Clio’s UK & Ireland Legal Insights Report 2026. The discussion focused on business-critical issues around AI adoption, with the group sharing their AI journey from experimentation to implementation. They looked at how they are deriving value from AI tools and applications; the challenges in terms of data governance, cybersecurity and client trust; and the impact on talent strategy in a changing profession. Finally, they explored the opportunities and challenges facing mid-market firms in the next two to three years. 

AI journey

Firms are at different stages in the AI journey. Key challenges include selecting which products to pilot in an evolving market. As Victoria Hodges, real estate partner at Teacher Stern, explains: ‘There’s nervousness about making a wrong move and choosing a product that is then superseded. And AI is an investment from a financial perspective, and also from a time perspective.’ 

Microsoft Copilot is a popular starting point, with Wedlake Bell and SA Law among firms that have completed a firm-wide rollout. David Hymers, Wedlake Bell’s IT director, explains: ‘We rolled out Copilot at the end of last year, and now it’s embedded into our organisation. While we’ve seen good adoption, we run monthly usage stats, and there are still people who are not using it.’ 

Hymers explains: ‘Copilot’s a pretty safe bet because a lot of firms run predominantly on Microsoft technology, and Microsoft has massive resources for development. Although Copilot has improved in the last 12 months, we’ll reach the point where we need to bring in another tool, although probably not across the whole firm. But we don’t want a situation where we’re using lots of different tools.’ 

'There’s nervousness about making a wrong move and choosing a product that is then superseded'

Victoria Hodges, Teacher Stern

He highlights a relatively new development in GenAI: AI native legal services. ‘Orbital, a real estate AI platform which we use at Wedlake Bell, has just launched its own conveyancing firm – they are starting to come for our lunch now,’ he says.

Firms are using AI embedded into core systems, as well as piloting agentic AI platforms Harvey and Legora, although cost can be a barrier for smaller firms. While Winckworth Sherwood’s AI journey ‘has gone from experimentation to embedded adoption via our document management system [DMS]… SME firms don’t have the same resources that big firms have… to play around with expensive AI tools,’ says partner and IT director Christel Aguila. As well as running parallel trials of Harvey and Legora, and rolling out a limited number of Copilot licences, ‘the advantage of having AI embedded in our DMS is that it benefits everyone, not just lawyers’. 

Firms are piloting generalist AI and legal-specific tools and platforms to work out the best combination for their requirements. Specialist partnership, employment and regulatory firm CM Murray has rolled out Copilot and is currently piloting legal AI platforms. As founder and managing partner Clare Murray explains: ‘We are regular users of Copilot. And in the legal tech space, we’ve trialled CoCounsel and Harvey, and now we’re trialling Legora and trying to figure out whether we should wait for Claude legal before making a decision – they all want a big chunk of cash up front.’ 

Return on AI investment

Murray highlights the need to leverage AI investments. ‘[The cost] really does focus us on making sure that we’re not just adding to our cost base, but leveraging specialist AI tools to be revenue-generating.’  

Adoption is the first step to achieving a return on AI. Peter De Maria, senior partner at Doyle Clayton, which specialises in workplace law, is facing similar dilemmas around choosing the right solution and managing firm-wide adoption. ‘Some [users] are over-eager, and use it for everything, whereas others haven’t even touched it. Prompts are key, so we’ve been sharing a lot of prompts in order to get the best out of Copilot.’

Hymers initially focused on quick wins: ‘The reason why Copilot is an easy way into AI is that it’s embedded in apps that lawyers use. So when we started rolling it out, we sold easy use cases, like how it can do a recap of your meetings, find emails in your mailbox and summarise documents.’ Users then ‘started thinking about how it could help their team or role. And when we checked in with them a few months later, some of the things they were doing were quite inventive’. 

Ursula Bosworth, marketing and operations director at Littler, also focuses on changing habits. ‘We’re constantly encouraging people to use it with very active questions – have you used it for this? When people feel like the old ways might be faster, I encourage them into that playful mode of curiosity… when you get a win, you’re more likely to use it for other, more sophisticated things.’

Simon Walsh is a business litigation partner at SA Law. He relates ‘strict rules about who can do what. We don’t use Copilot for client work because it gets things wrong… We use CoCounsel for legal work. The challenge is to encourage people to move beyond just using it for research, where it is brilliant.’

Hymers argues that Copilot has improved sufficiently to be used for legal research too: ‘It is becoming more reliable. And there is a researcher agent in Copilot. We recommend that if people are doing legal research, they use that. In the last month, Microsoft brought real improvements to its reliability and cracked down on hallucinations. We have some lawyers who say, “It’s not reliable, I don’t want to use it and I don’t want my juniors using it, either”, but they’re going to be left behind because the rest of the firm is getting on with it.’

While there is a clear return on investment in terms of getting things done faster, return on experience is also an important value-add, for clients, fee-earners and staff in terms of streamlining people’s work and improving job satisfaction. 

Walsh says: ‘Part of the return on investment is return on experience. When you’re working away to answer a question at eight o’clock at night, and you can check something quickly, it means you can turn something around for a client right away… And if it makes life easier for fee-earners and staff because it streamlines their work, that’s a return on experience, which hopefully makes our firm a better place to work.’

Aguila offers an example: ‘When we were trialling Harvey and Legora, one of the heads of department told me they were in the middle of a court case and they had to research something. They used Harvey and it took them two hours when it would have taken a senior associate days to produce that specific content. At the time, we were still deciding whether or not to invest, and he said, “I am more willing to pay for this tool than to hire another person”. That is a big part of its value.’

As Walsh explains, an AI policy is the starting point for AI governance: ‘But getting that AI policy in the first place is a big stumbling block because it’s a lot of work, and involves a lot of thinking about what you allow people to do.’ 

Ensuring people adhere to an AI policy is even more challenging. ‘We monitor usage across all AI tools,’ says Aguila. ‘We recently implemented a tool that also allows us to see prompts… we can demonstrate to clients that we are applying a monitoring tool and when people use ChatGPT, Claude, Perplexity or Copilot a banner flashes up saying, “please do not upload any client or personal data”. We need to keep reminding people to use it responsibly.’

Undermining confidentiality

De Maria has experienced clients inappropriately using AI for legal issues: ‘For example, an individual using AI to raise a grievance, and their manager using AI to respond.’ CM Murray has a notice on the firm’s website and engagement letter warning clients not to upload their information into publicly available AI because it risks undermining confidentiality and legal privilege. Initial enquiries have become longer and more complicated as people increasingly turn to AI, notes Murray. 

Walsh adds that long, AI-generated instructions ‘look good, but the client doesn’t always understand what they are asking, and then they don’t want to pay for your time when they’ve asked 20 questions instead of three’.

Yulia Barnes, founder and managing partner of Barnes Law, has received instructions drafted with ChatGPT that clients have not read before sending: ‘And sometimes clients are not as appreciative of [lawyers’] expertise because they can use AI to produce a contract that looks good, but nobody understands it.’ 

Client expectations regarding data-handling and transparency were a focal point of Clio’s Legal Insights report, and firms generally include AI use in their terms of business. 

Walsh recalls clients specifically instructing lawyers not to upload their documents into any AI tool. 

Reliability underpins trust, underlining the importance of checking AI output. ‘We were early adopters of a real estate tool and people became disillusioned when the output wasn’t 100% accurate,’ says Hodges. ‘If a junior associate was doing the work, you’d check it before sending it to the client. But with AI maybe you don’t need a junior associate producing the first draft – and AI produces in minutes what would take an associate a day. 

‘Clients are sending us documents they have drafted [with AI] and asking us to check them, without appreciating that this is harder than starting from scratch when you don’t know the source material.’ 

Training and talent

As AI takes over the routine legal tasks usually done by trainees, firms have to find new ways of providing on-the-job experience. ‘Traditionally, people have been trained by working through a lot of painful, low-quality work, but although it’s good training, it’s really inefficient,’ observes Clio’s chief operating officer Ronnie Gurion. ‘AI can take away a lot of that drudgery, and firms are going to have to rethink how they train lawyers,’ he says, adding that much of the mundane work done by juniors is the type of work that is well-suited to AI. 

‘I see a change in the quality of lawyers who are newly qualified compared to the traditional training contract,’ says Barnes. ‘It wasn’t easy and it involved long hours, but people knew a fair bit when they qualified. Now, they get a bit of experience here and there, put it all together, send in a form and pass exams, but there is no real understanding of how to take a transaction from the beginning to the end because they only get to see bits of each transaction.’

Walsh believes it is possible to learn on the job while using AI for research. ‘The good juniors will take the time to click on the link and read the details,’ he explains. ‘And by taking the time to do that, they’ll expand their knowledge generally.’

'AI can take away a lot of that drudgery, and firms are going to have to rethink how they train lawyers'

Ronnie Gurion, Clio

‘Restructuring the whole training programme can still provide the same type of experience that you had pre-AI,’ says Aguila, reminding the group that AI is already embedded into working practices. 

Looking ahead, it is likely that AI will be included in legal training and CPD (continuing professional development) programmes. At Littler, Bosworth envisages ‘AI becoming part of CPD for everyone, which would cover effective prompting, as well as the firm’s AI policy, and AI risks.’

AI is a key factor in recruitment. ‘We hear more and more that firms are struggling to hire if they’re not providing AI tools,’ says Gurion, adding that AI is becoming an increasingly prominent requirement for lawyers who have already used it. ‘So if you’re not at least a mainstream adopter, you’re going to be disadvantaged.’ 

Road ahead

Looking forward, the group identified continuing challenges relating to governance, pricing and billing, with AI shifting firms towards fixed-rate models. AI is becoming the norm, so it is no longer seen as ‘innovation’. 

As Aguila observes: ‘For SME firms, it’s also about having the right talent to get value out of these tools and provide the transformation that your client is looking for. We’re all now using the same tools: Copilot, Harvey, Legora. It’s not innovation; it’s finding the right business case that will mean that your client will be willing to pay you the same or more because it’s adding value.’

Hymers adds: ‘As an IT director, one of my challenges is how do we pick the right tools? Some of the AI tools that are around now might not be around in a couple of years.’ He is currently looking at an AI tool for conflict checks.

'Copilot comes along and somehow it feels like a safe space for people – until a data subject access request comes through' 

Rachel Coburn, Hickman & Rose

‘Billing is a massive challenge,’ says De Maria. ‘And working out the value of what we are doing as human beings too. The optimist in me is saying our expertise, experience and judgement are still important. So, however much we’re using AI, it doesn’t replace experience and explaining things to people.’

On risk management, Rachel Coburn, compliance and HR manager at Hickman & Rose, remains focused on data as a perennial challenge: ‘We’ve spent so many years drilling into people not to put something in an email that they wouldn’t want to have read out in court.  And then Copilot comes along and somehow it feels like a safe space – until a DSAR (data subject access request) comes through. Another challenge is understanding how people are really using it. The opportunity for us is to modernise our case management system which will enable us to use AI.’

Murray is worried about the erosion of core legal skills: ‘I am concerned that if we outsource thinking, we will forget to think. I want to instruct a lawyer who’s read the papers.’ But, she adds: ‘In terms of opportunities, the workflow opportunity is incredible.’

'[The challenge is to] keep pace with what AI can deliver' 

Simon Walsh, SA Law

Bosworth highlights the speed of change as the biggest challenge for business operations, all of which are impacted by AI. ‘I think it will be a continuum of living in the paradox, speeding up and slowing down. And partnering with the right people within the firm and outside, to know what relationships you’re going to invest in, what technology you are going to invest in and how to cut through the noise.’ 

Barnes adds recruitment to the AI challenge, ‘because young people are used to using AI, and sometimes they stop analysing or challenging the information they see’. In terms of opportunities, she references a LinkedIn post that stated: ‘If you think AI will replace lawyers, you don’t actually know what I do.’ In fact, Barnes believes that ‘people using AI will create a lot of work for lawyers’.

Hodges highlights GenAI’s rapid trajectory: ‘It’s the speed of change, how quickly we’ve got to where we are now and how quickly things will develop in the next two or three years. And it’s about not wanting to be left behind.’ 

Like Aguila, she is concerned about keeping up with bigger firms while managing a smaller budget. Hodges highlights the human side of law as the differentiator: ‘A lot of law comes down to building and developing relationships with clients and I can’t see how AI can replace that.’

Walsh’s challenges are ‘keeping pace with what AI can deliver, and making sure that we’re using it to the firms’ and the clients’ best advantage’. 

Gurion sees the upcoming challenge for legal as ‘a bifurcation in the industry between firms that really embrace AI and transform their operations, and the rest. Everyone’s going to use it, but some will use it in a more measured way.’ In terms of the opportunities, firms need to decide what to do about it. ‘This means thinking about your tech stack. Do you have the right infrastructure to be able to leverage AI? Do you have the right training and people and mindset?’ A winning strategy is to get the right case management or system of record solution, and then think about bolting on additional AI tools. 

• This roundtable was kindly sponsored by Clio

Photographs by Noah Da Costa

Joanna Goodman is a freelance journalist and author of Robots in Law – How Artificial Intelligence is Transforming Legal Services