Most lawyers now have access to AI tools, but many don’t know how best to put them to work. Allan Carton sets out how and where management teams in law firms can provide a better steer and support.
Law firms still have time to get AI initiatives on track or to re-launch them to produce better results. Nobody should feel they are diving in at the deep end, uninformed or exposed now, as a lot has been learned from early adopters.
Build the firm’s knowledge into tools
Competitive advantage will come from tapping into legal knowledge, judgement and ways of working that your practice can build into its use of AI. It’s increasingly clear that success here depends heavily on engaging your people and developing the firm’s ability both to use these tools and improve them over time.
How to make the most of that knowledge and experience with AI varies, depending on the type of legal work and the task. Some of the key tools you want your people to understand and become comfortable using include:
- Prompt guides.
Tested prompts for everyday drafting, summarising and first-pass research or evaluation, with the lawyer in control throughout.
- Precedents, templates and clause banks.
The firm’s own material should be organised, kept current and subject to appropriate access permissions, so AI tools can draw on it, with the output reflecting how the firm works. Nothing new there, but keeping these materials in order has always been a challenge and has now become more critical.
- Playbooks.
For repeatable work such as triage of new enquiries, a written definition of what good looks like: the steps, the checks, what to flag and where a lawyer needs to make a decision. It should be written by the people who do the work and reflect how they approach it.
- Legal AI “skills”.
A short set of plain-English instructions that allows an approved AI tool to apply one of the firm’s playbooks, such as triage of new instructions, or carry out a recurring task, such as checking citations. It includes any templates, checklists or reference notes the tool needs to approach the task consistently.
- Agents. AI working through several steps on its own, using skills along the way, with a lawyer signing off anything that reaches a client.
This works a lot better with a structure and roadmap in place from early on (where each item is kept, who owns it and which version is current), so people can find what they need and collaborate on improving these tools. Accept that they won’t be perfect at the outset, or you’ll never get people using them. There will be some risk, but manage it proactively.
As AI takes on more of the routine work trainees used to learn from, helping to develop, use and maintain these tools shows them how the firm approaches legal work, what gets checked and why.
The material captured here belongs to the firm. Skills follow an open standard, so the same skill can potentially be used across tools including Claude, Copilot Cowork and ChatGPT without being rebuilt from scratch. That helps keep the firm independent of any one supplier.
What’s available now
There are more than enough AI tools available, but they’re not as yet joined up as much as we would like, so are more difficult to manage and configure than firms would want. That is starting to change.
Suppliers to larger firms (Thomson Reuters and LexisNexis among them) add value by bringing more components together, at a cost. For many SME firms already using Microsoft 365, Copilot is the most accessible starting point and has developed dramatically in the last six months.
Most SME firms, which for present purposes I’m treating as firms of 20 to 500 people, won’t want to fund pre-built integrations across several suppliers. They will need to be more creative about bringing the components together. Much of the knowledge is already there to capitalise on … with the right approach. That means defining workflows, then capturing the issues and decisions lawyers deal with at each stage, so AI tools can work with them.
Start with why. Get IT and lawyers talking about how.
The first conversation should always have been about why the firm wants AI: where you want it to make an impact. What results are you looking for? Then think about how your people will see and use it. What will help them engage?
Lawyers know the legal work. The IT team has to get to grips with technical tools that change constantly, with security, compliance and governance all critical and often difficult to configure. I know, having worked with them hands-on to understand the challenge. Without someone able to translate between legal work and technology, the two sides talk around each other and little happens.
It’s an understandable gap, so get on top of this early. The lawyers and their teams (legal executives, paralegals, secretaries and support staff) know the legal decisions that need to be made, how they interpret the facts, how they set strategy and tactics and what a good outcome looks like for the client … though much of that is rarely written down. The tech team can’t be expected to know that and rightly have their own priorities. Equally, most lawyers don’t and shouldn’t want the technical detail, but they do need to recognise the pressure IT is under and understand more of the fundamentals of AI than most do now.
To avoid constantly backtracking, rediscovering and reinventing the wheel, I recommend forming an AI Steering Group to think consistently through the decisions that need to be made: where AI should be used, what gets checked, how client confidentiality is protected and how a successful trial becomes everyday practice. It also keeps up engagement and momentum across the practice.
Identify use cases
First, help people to understand what prompt guides, playbooks, skills and agents are and how they might be used in their work.
Choose the right people in each legal area to map out what the process looks like (not in too much detail to begin with) and, within that, where these tools could help. The AI Steering Group might appoint “Champions” to lead this.
Once the flow is broken down, a lawyer can answer six questions without IT:
- how often does this piece of work arise,
- how widely is it done across the firm,
- what’s the benefit if it’s done well every time,
- are there enough precedents and examples,
- does it follow the same pattern each time, and
- is the process defined or could it be mapped?
Scoring the answers lets you compare candidates and decide what to tackle first. Add the resources needed and predicted benefits and you can plot them on a matrix.
Candidates include:
- triage of new instructions, covering allocation and a first pass;
- an initial evaluation in financial provision on divorce, kept under review as new information comes in;
- challenging your own arguments in litigation before the other side does;
- handling enquiries before contract in conveyancing more intelligently; and
- gathering and evaluating information in probate or the administration of an estate.
These all need to be tested and trialed under effective supervision. Keep developing those that work and stop those that don’t. Be prepared to get some wrong, learn from that and share it across the practice.
Used well, AI evaluates the information available in line with the guidance and rules you have set, then identifies gaps, issues and where to go next. Your tech team can help make sure the tools draw on as much context as possible. Final judgement stays with the lawyer for now, but AI makes their life easier if they can trust what it provides.
As confidence grows, the firm can reduce human intervention and make more use of agents, while keeping legal judgement and client-facing work under lawyer supervision.
Getting and keeping control
Clear guidance needs to come early, otherwise people find their own ways of working and nobody knows what’s being checked. Whoever uses the output is responsible for it and checks it as they would a trainee’s draft. Anything that rests on the law is checked against approved, reliable sources.
Your roadmap, managed by the AI Steering Group, sets how closely each kind of output is reviewed before it reaches a client. Talk to your professional indemnity broker early about how the firm uses AI and the controls around it.
Before approving any AI tool, find out where prompts and documents are stored and where the AI model processes them (often different places) and make sure the provider doesn’t train on your use of the tools.
Take stock now
A sensible first step is an informed review of where the firm stands: its current structure, short and longer-term aims, how AI is being used and any gaps in governance. That review should also cover the knowledge tools being built, the strongest use cases, agreed next steps and what the firm is telling clients.
Allan Carton
advises law firms on choosing practice and case management systems, adopting AI and listening to clients. He qualified and practised as a solicitor and holds an MBA from Manchester Business School.
Read more on Carton & Co’s approach to AI adoption.
