Many firms approach AI the wrong way round. They see a strong demonstration of a legal AI product, ask about licences and timelines, and six months later wonder why few people use it. The tool is rarely the problem. AI amplifies what is already there: good data and clear processes, or the lack of them. This is the order of work we follow with firms that want to be ready.
Step 1: assess where the firm stands
Start with an honest picture of four things.
Data quality and access. AI is only as useful as the information it can reach. If matters are coded inconsistently, documents are spread across email and local drives, and clients appear several times under different names, the results will be poor.
Systems. Legal AI tools work best when the firm’s matters and documents are in cloud practice management and document management systems. If they are not, readiness starts there.
People. Who is curious and will try new tools first? Who will need more support? Knowing this before you begin saves time and frustration.
Policy and risk. In the 8am 2026 Legal Industry Report, a survey of more than 1,300 legal professionals, mostly in North America, 43% said their firm has no formal AI policy and no plans to create one.1 Without a policy the firm is exposed on confidentiality, quality and professional responsibility.
Our AI readiness assessment is a quick way to get a first picture.
Step 2: fix the data
This is the step firms most want to skip, and the one that decides whether the rest works. The problems we find most often are:
- Duplicate contacts and matters. The same client appears under several names and spellings.
- Inconsistent matter coding. Practice area and matter type codes are applied differently by different people, so reports cannot be trusted.
- Documents outside the document system. Important documents sit in email, on local drives or in shared folders.
- Incomplete records. Contacts with no email address, matters with no billing code, file notes that stop halfway.
If a change of system is part of the plan, that is the time to deal with data quality. Deduplication, analysis and clean-up are part of our data migration work for that reason.
Step 3: write the policy
An AI policy does not need to be long. It needs to answer four questions.
Which tools are approved? Keep a list of approved tools and what each may be used for, so staff are not left to choose consumer tools for client work.
What data may go into them? Say clearly which kinds of client information can be processed by which tools.
How is the output reviewed? Decide who checks AI-assisted work before it goes to a client or a court. Set the review step for each use: correspondence, research, summaries, drafts. AI Critic is a free prompt we publish for pressure-testing an AI-assisted draft before it leaves the firm.
How are professional obligations met? The lawyer who signs the work remains responsible for it. The policy should say how supervision and review work when AI has assisted.
Step 4: choose the tools
With the first three steps done, the firm can judge tools against real needs. They fall into three groups. Where a product is named, the description is its vendor’s own and not our assessment of it.
- Legal research assistants that draw on a publisher’s legal content. Thomson Reuters describes CoCounsel as powered by Westlaw and Practical Law.2
- General legal AI platforms for research, drafting and review. Harvey presents its platform as one for law firms and in-house legal teams.3 Legora describes itself as collaborative AI for reviewing, researching and drafting.4 Clio describes Clio Work as matter-aware legal AI for research, analysis, strategy and drafting.5 Our catalogue also has pages on Harvey and Legora.
- AI inside the systems the firm already runs, such as the AI features of a practice management or document system, or meeting and call notes filed to the matter.
Check each candidate against the policy from step 3, against the systems it has to connect to, and against what the firm already pays for. Many firms hold AI features they have not switched on.
Step 5: pilot, measure, then scale
Start with a small group of willing people in one practice area. Agree what success looks like before you begin: time saved on a defined task, the quality of the output, how many people keep using it. Measure it. Use the result to decide whether and how to extend the tool to the rest of the firm.
A pilot gives the firm its own evidence, on its own matters, and it produces the colleagues who will help others adopt the tool. In our experience, firms that roll a tool out to everyone at once struggle with adoption, and firms that prove it on a small scale first have a much easier second stage.
Where to start
Our AI advisory and enablement work follows these five steps with a qualified lawyer leading it. For the figures behind the case for acting, read the AI adoption gap in numbers.
Sources
- 8am. 8am 2026 Legal Industry Report: AI Adoption Surges Through Turbulence as Firms Push Forward. 5 March 2026. Accessed 16 November 2026.
- Thomson Reuters. CoCounsel - The Professional AI Legal Assistant | Australia | Thomson Reuters. Accessed 16 November 2026.
- Harvey. Harvey | AI software for legal and professional services. Accessed 16 November 2026.
- Legora. Legora. Accessed 16 November 2026.
- Clio. Clio Work - Legal AI that actually knows your case | Clio. Accessed 16 November 2026.