AI balancing act

The legal sector is considered one of the most obvious areas where artificial intelligence (AI) can be of significant use. With its ability to perform some tasks in a fraction of the time it would take humans to, AI offers substantial benefits for law firms. Unsurprisingly, many firms have embraced AI, while others are considering adopting it in some form. A poll conducted at the recent annual sLaw Society Private Client Conference reveals that 45% of attendees use AI regularly, while 28% plan to explore its use within the next three months. 

Yet many issues must be addressed when using AI. As private client work involves handling sensitive personal matters where discretion, empathy, trust and strong client knowledge are particularly important, firms in this field must be especially cautious about deploying AI. 

“The use of AI in practice can be transformational,” says Eduardo Ustaran, partner and co-head of the data privacy and cybersecurity practice of Hogan Lovells Cadwalader, which advises organisations on AI usage. “As well as aiding research and drafting, it can compare and summarise documents, aid due diligence and compliance, onboarding and conflict checks.” 

But its risks can be considerable too. “Confidentiality, privacy, privilege and data protection are all things that must be carefully considered when using AI,” he warns. “Lawyers – especially less experienced lawyers – may become overly reliant on it and, crucially, a major risk is relying on it without question as it can make mistakes.” 

So how are firms managing these various aspects of using AI? 

How firms are using AI

Firms that have embraced AI use a range of AI tools, from general products – like Microsoft Copilot, OpenAI, Anthropic and ChatGPT – to specialised legal products – like Thomson Reuters CoCounsel, Harvey, Luminance and Lexis+ AI. Some firms are also developing their own forms of AI, such as Hogan Lovells Cadwalader, whose global technology company, ELTEMATE, offers aids for a variety of legal tasks. 

Certain firms have been quick to adopt AI. Penningtons Manches Cooper uses a mix of general AI products, including generative AI products, like Microsoft Copilot, legal tools, such as Thomson Reuters’ products, and subscribes to PLC, says senior knowledge lawyer Sarah Robinson, who’s also vice chair of the Law Society’s Private Client and Estate Member Advisory Committee. “We also use Ask iManage for workflows and other tools for our specialised practice areas, like Avail for our real estate work.”  

AI is well embedded in the daily work of Natasha Holme, managing associate in the private wealth disputes team at Mishcon de Reya. Holme, who primarily uses Legora for various tasks, says: “I consider at the outset of a task how it could be done more efficiently with AI and whether it’s suitable to use it. For example, I may use AI for first drafts of bespoke documents for clients, for turning rough notes or call transcripts into attendance notes, generating correspondence, and converting client and colleagues’ comments into revised drafts – which it can do in minutes.”  

Holme also insists that AI provides a valuable second pair of eyes. “You can cross-check a draft will against client instructions and review estate accounts for arithmetic errors, and it’s excellent at consistency checking.” But AI really comes into its own when it’s used for reviews and summaries, she explains, like summarising the terms of a trust deed and reviewing large document bundles.

Regulation and guidance 

An ongoing consideration for firms is the obligation to comply with professional rules governing AI use. To date, the UK government hasn’t regulated AI but relies on industry regulators – such as the Financial Conduct Authority and the Competition and Markets Authority – to ensure AI is used responsibly in their sectors. The Solicitors Regulation Authority (SRA) points out, however, that the UK government guidance on the use of AI in the public sector – the AI Playbook for the UK Government – is also a helpful aid for private organisations. 

Meanwhile, the EU’s AI Act (Regulation (EU) 2024/1689) is the world’s first comprehensive legal framework for the development and use of AI, and it’s also launched related initiatives, such as the AI Continent Action Plan. While these don’t bind firms in the UK – unless they operate in the EU – they are becoming a well-recognised standard.

The Law Society has produced an AI strategy and a guide to generative AI, and although the SRA has not yet issued any specific guidance on AI for law firms, it is working on one. In the meantime, on 23 November 2025, it issued de facto guidance, publishing Risk Outlook report: The use of artificial intelligence in the legal market. It considers AI to be another form of technology and, accordingly, solicitors remain fully responsible for the accuracy of their work. Its guidance on effective supervision was updated on 12 June, with the addition of a supervision and AI section, which states: “When using AI tools in the delivery of legal services, regulated firms and authorised individuals should consider what effective supervision of AI-assisted or AI-generated work looks like in practice. This includes ensuring that outputs produced with the assistance of AI are subject to appropriate human review, scrutiny and professional judgement. 

“Regulated firms and individuals should refer to the information available on our SRA Innovate webpage when considering the use of AI in legal practice and must ensure that an authorised individual retains ultimate responsibility for any legal services delivered with AI assistance.”

More recently, firms had sobering notice of the pitfalls of over-reliance on AI when Pinsent Masons was admonished by a High Court judge in Cork & Another v Smith [2026] EWHC 1199 (Ch) for using AI material that proved inaccurate. While the firm has referred itself to the SRA, others’ use of AI is under review to determine whether professional standards have been breached. Ustaran, whose team has produced the review How to prevail when technology fails, stresses that this highlights the need for firms to understand exactly how their lawyers work. “You must have clear and justifiable rules about how and when AI can be used,” he explains, “but it’s not enough to have rules in place – they must be implemented and adhered to.”

How to select

With these issues in mind, firms must select AI tools carefully. In February 2026, the SRA issued its recommendations on what questions firms should address when they do. They should first determine if they get the most out of the technology they already use. Then, they need to understand their needs, what they want the technology to do, and whether any considered technology is compatible with their existing systems. 

Ustaran advises firms to focus on understanding how AI can aid individual practices. “Your work should be the driver which dictates what AI would be useful for your practice,” he says. Small- to medium-sized firms, in particular, may need to start small, focusing on the highest impact tasks first. Client expectations and habits are also driving firms to embrace AI, he reports. “More sophisticated clients now ask how the firm is using AI and how it will be used in their cases, as they did during the transition from faxes to email.” 

Ustaran notes that it can be useful to source something off the shelf and tailor it to your own needs, but – as for any new system – testing before buying is essential: “The best thing is to have a pilot scheme for the AI tool – something that allows you to use it in a safe, controlled environment.” Buyers should conduct due diligence on the vendor’s security credentials and look for recognised certifications. 

David Wood, head of legal tech and innovation at Penningtons Manches Cooper, says his firm has a robust procurement process and, as Ustaran suggests, legal technology selection is driven by business needs. “We don’t buy tech for tech’s sake; it must be needs-driven. A request may come from a practice team that thinks it could work more efficiently with a form of AI. We work with them to understand the problem they are trying to solve. This process provides clear requirements and helps ensure the solution we put in place is a good fit and will be used. The AI tools we choose may be offered by trusted suppliers – like Microsoft – which makes onboarding easier. But there are many new AI tools in the market too, and we also consider what our competitors may be using. The choice can feel overwhelming – we keep an absolute focus on business needs and organisational strategy,” he says.  

On the cost front, he reports that his firm is preparing for a mix of options, from per-seat subscriptions to consumption-based and hybrid models. “It almost feels like the early days of business mobile phone plans and subscriptions – and with some frightening usage charges to hit the news”.

Benefits and risks 

For most lawyers, the most immediate benefit of using AI is time, as it removes the arduous work many tasks involve. Holme reports that many lawyers value the confidence AI can deliver when properly used: “It’s knowing you haven’t missed something and that you can achieve consistency which can greatly lessen anxiety.” Holme notes that while there’s a concern that AI will reduce the volume of legal work available, the opposite may be true. “AI expands what an individual lawyer can handle, and overall demand for legal services appears to be growing.”  

Robinson also finds that AI helps save enormous amounts of time in tasks like generating letters and documents and in analysing and summarising files. She and colleague Wood believe AI may also provide new revenue streams. “It can help you enter new fields of practice,” Wood notes, “and it can help you do things at a scale you weren’t able to do before.”

Yet the risks of using AI are real. A key consideration in private client work is heightened awareness of privacy and the need to reassure clients that their data will be safe. The Pinsent Masons case has reinforced the need for lawyers to have a clear understanding of AI’s limitations and the risks associated with its uncritical use. Wood says that the case has generated much reflection. “It’s like cyber-attacks; there’s an awareness of ‘it could happen to us’. Businesses are stopping to check their AI governance and supervision policies.” 

Good oversight is critical for ensuring that the risks of AI are well managed and that professional obligations are met. The SRA says that to ensure that AI use has senior leadership oversight, at a minimum compliance officers for legal practice (COLPs) should be responsible for its regulatory compliance. It stresses too that boards should oversee its purchase and future use. Ustaran reports that the people responsible for data management are now also overseeing AI in some firms. “The use of AI is just making our shoulders broader,” says Wood. “There are more responsibilities and tasks for our governance, IT, security and data teams to consider at present, but I think new roles such as AI officers will emerge in the near future.” 

Policies and practical management

Given the importance and risks of AI, firms need clear policies that govern its use. These can vary depending on a practice’s size and focus. “It can be a simple list of dos and don’ts,” says Ustaran. “Or, it can be more complex, depending on how you’re going to use AI. Firms have long had policies on how technology can be used, so AI can be managed in a similar way.” 

An AI policy should include rules on when AI tools can and can’t be used, what jobs it can be used for, what information can be put into the AI system, how fee earners are to be trained and when they can use it. As technology advances and firms expand their use of AI, policies need to be regularly updated. 

AI training to ensure that fee earners and others are familiar with the available products and how they can and should be used is essential. A combination of face-to-face and online training is usual. Constant communication is considered vital too. At Hogan Lovells Cadwalader, Ustaran’s team meets weekly to discuss the latest uses of AI in the firm; at Penningtons Manches Cooper, teams constantly review and discuss AI usage.

A vital issue, however, is how AI affects the development of new lawyers who must acquire the knowledge and skills required for practice. Ustaran finds this topic interesting. “On the one hand, new lawyers are more digitally able. On the other, AI may be able to complete many tasks they traditionally undertook and learned from, and it may reduce the need for them.” He is sanguine about this. “AI has a greater impact on the way junior lawyers undertake some tasks – like research – but I doubt it will take over from them completely. Once, they used an office template for drafting; now, they can use AI. But they still must go through that process and understand what the contract should deliver. You still need the ability to judge how suitable the draft is, although AI may play a greater role.” 

This involves some foresight and good supervision. Holme says one challenge is that the lawyers best placed to train junior lawyers are often those least familiar with AI: “This creates a skills gap in both directions, which firms must address.” 

Some firms are finding ways to address this gap, for example by having junior lawyers first complete some tasks – like will preparation – manually, then have them complete the same task using AI so they can assess the AI-generated output and develop an understanding of the underlying process. They also have the junior lawyers produce a brief commentary explaining how they’ve used AI and how they’ve changed what it produced. Regular file reviews and feedback sessions need to be part of the supervisory process.  

To manage these issues, firms must create the right infrastructure and processes around it. Some measures are key to making AI work well in practice, insists Holme. “Buy-in and good communication are critical, from the firm leadership to senior lawyers, who must be seen to use the tools themselves, to tech-savvy junior lawyers who should share their knowledge with more senior colleagues.” 

Penningtons has an AI Steering Committee, Working Group and many AI Champions who help drive and monitor AI usage and lend support to those who use it. Mishcon de Reya also has AI Champions with AI expertise within its practice groups and a dedicated group to manage issues and provide guidance for users. A well-maintained prompt library also allows firms to identify what works and to replicate success.

Effect on clients

The use of AI is also affecting firms’ relationships with their clients. “Some may expect you to use AI to manage their cases, and to be able to work more effectively by doing so,” notes Ustaran. “They may consult it themselves and query your advice.” But a client’s use of AI can sometimes be advantageous, says Holme. “If they consult AI before talking to their lawyers, this can help them organise their thoughts, prepare questions and determine what they want to achieve.” However, it also means that they may arrive with AI-generated assumptions that need to be tested and often corrected. 

Robinson says firms need to talk with clients about how AI may be used on their matter. Like others, she says that AI is affecting what clients expect to pay, although they may not consider the cost of AI to law firms. “Traditionally, many clients may be billed on the hours worked,” says Ustaran, “but if we use a tool that can shorten those, they may expect to pay less. The advent of AI will make the concept of chargeable hours less relevant, and we have to learn how to deal with it by focusing on the value of what we deliver.” 

Firms may take different approaches to billing when they’ve used AI. One is to use more fixed fee arrangements when there have been significant time savings. But these, says Holme, lend themselves to non-contentious or discrete work and are less suited to litigation. She believes there’ll be a gradual shift towards a blended model: more fixed fees, more value-based pricing – and more transparent conversations with clients about the role AI has played in their work.

The future

AI continues to evolve. The most significant recent development Holme has seen is agentic AI, which can plan and execute multi-step tasks, deploy sub-agents to handle different parts of a workflow, self-correct and operate semi-autonomously. “You’re working with something closer to an active assistant, and this significantly broadens the scope of work AI can complete,” she explains. 

Managing AI will continue to be a balancing act. Robinson believes AI could generate new work, yet serious issues to confront may still lie ahead. “Following the Law Commission report Modernising Wills Law, for instance, there could be a move to the use of electronic wills, and firms will have to consider how AI can assist this process.” 

But these lawyers say that optimism is called for too. As AI relieves lawyers of time-intensive, repetitive work, they can focus more on providing good guidance and advice to clients – a satisfying prospect for the future.