How should a mid-sized law firm evaluate legal AI tools?
A mid-sized law firm should evaluate legal AI tools by mapping the transactional work it wants to change, then scoring each tool against that work, not against a vendor's demo. Check fit to your matters, whether it works in Word, whether it uses the firm's own precedents, data handling, knowledge ownership, published pricing, and whether a realistic pilot will be adopted. If the firm is not sure where to start, an independent specialist can help you pick the right one.
How should a mid-sized law firm evaluate legal AI tools? By starting with the work, then the products. The usual route is a demo: a vendor shows something impressive with somebody else's documents, and the conversation moves quickly to packages and seats. Harvey, Legora, Spellbook, Luminance and Microsoft Copilot can all look strong in a controlled walkthrough. That is useful later. It is not, on its own, an evaluation.
How should a mid-sized law firm evaluate legal AI tools?
Start with a clear picture of the transactional work you want to change, then judge every tool against that picture, using your own matters. The question is whether the tool will change the work your fee earners already do, in the system they already use. This is written for transactional work at mid-sized firms with between 5 and 500 fee earners in the UK, Ireland, the USA and Canada, and for in-house teams doing the same kind of contract work.
Product tests for a drafting tool are in how to evaluate a legal AI drafting tool, and the questions to put before you sign are in ten questions to ask before buying legal AI.
What should you clarify before you speak to vendors?
You should be able to say what clients actually pay the firm for, where the know-how lives, where time leaks out each week, and which decisions must stay with a lawyer. If you cannot, a demo will fill the gaps with the vendor's story rather than yours. Write this down with the people who do the work: the document types that take up the most hours, the precedents lawyers actually reach for, and where automation would erode the advice clients value.
What criteria should you use to compare legal AI tools?
Score each shortlisted tool against the same criteria, on your own matters.
| Criterion | What to check | Why it matters |
|---|---|---|
| Fit to transactional work | Does it help with the drafting, review or negotiation you actually do? | A mid-sized team rarely needs an all-purpose enterprise suite first. |
| Works in Microsoft Word | Is it a Word add-in that respects tracked changes and your document system? | Lawyers under deadline will not switch to a separate portal. |
| Uses the firm's own precedents | Is the starting point your approved wording, and can you see the source? | Generic language takes longer to check and is harder to stand behind. |
| Data handling | Where is client data processed, and is it used to train models for others? | Confidentiality does not pause for a pilot. See our security questions for legal AI vendors. |
| Knowledge ownership | Does the firm own the clause library and negotiation record, including on exit? | Otherwise you rent capability that never accumulates. See who owns the knowledge your AI creates. |
| Pricing transparency | Is the all-in per-user price published, including seats, implementation and support? | Opaque pricing is a poor start to a long contract. Scriptum's figures are on our pricing page. |
| Pilot design | Can you try it on live matters, with a small group, against agreed measures? | A demo is not a pilot. |
| Adoption | Will the people who draft every day use it without a large change programme? | Unused seats are the most expensive kind. |
A starting checklist, not a ranking of any product.
How should you run the pilot?
Run it on live matters, with the lawyers who will use the tool, for long enough to get past the novelty. Agree in advance what "good" looks like: quality of the first draft, time to a partner-ready version, whether preferred positions are respected, and whether people still open it in week four. Our post on how to implement legal AI in a mid-sized law firm sets out a practical sequence: start with one practice group, set the rules, then widen.
When should you bring in an independent specialist?
Bring in an independent specialist when the firm is not sure what it is trying to change, when partners cannot agree on the problem, or when a previous purchase has added a subscription without changing the work. Outside help is also useful when the shortlist is being driven by demos rather than a map of the practice. Many mid-sized firms can work through the criteria above themselves. Reach for a specialist when those conversations are stuck, or when you want someone who is not selling the software to run the process.
One example is Chris McMenemy's EmbedRight Canvas for mapping where AI belongs in the work. Its author describes himself on his own site as a founder, engineer and venture builder who has delivered more than $100m of technology into businesses over twenty years. EmbedRight is a two-part canvas for businesses that need AI to change something real rather than add another subscription. It maps what clients are paying for, where the expertise lives, where time and money leak out, and what has to stay human, then ranks those findings by impact and AI suitability so you finish with a short list of what to do first. The canvas is free to use with your own team, and McMenemy also offers to run the session. That is one independent approach, not the only one. A trusted peer, a consultant who is not selling a tool, or an internal lawyer with enough distance can play a similar role.
Start with the work, not the demo, or you will end up buying a subscription rather than a change in how the firm practises.
Where does Scriptum AI fit?
Scriptum should be judged on the same checklist as anyone else. It is a Microsoft Word add-in that drafts from the firm's own precedents, captures agreed clauses into a clause library the firm owns, and publishes its pricing. It is built for transactional work at mid-sized firms with between 5 and 500 fee earners in the UK, Ireland, the USA and Canada. The Draft product page shows how it works in Word.
If you already know the work you want to change, book a demo and put Scriptum through this evaluation on one of your own templates. If you are not sure yet, do that mapping first, with or without an independent specialist, and only then sit through a demo.
Common questions
How should a mid-sized law firm evaluate legal AI tools?
Map the transactional work you want to change, then score each tool against that work on your own matters. Look at fit, whether it runs in Microsoft Word, whether it uses your precedents, data handling, knowledge ownership, published pricing, the design of the pilot, and whether lawyers will actually adopt it.
Should a law firm start its evaluation with a vendor demo?
No. A demo is useful later, once you know which part of the work you are trying to change. Starting with a demo makes it easy to buy a product that looks impressive but does not fit the way your firm actually practises.
When should a law firm bring in an independent specialist?
When the firm is not sure what it is trying to change, when partners cannot agree on the problem, or when a previous tool has added a subscription without changing the work. A specialist is also helpful when the shortlist is being driven by demos rather than a map of the practice.
Do mid-sized firms need a different evaluation from large firms?
Yes, in emphasis. A mid-sized firm should favour tools that fit transactional work, work in Word, and can be piloted without a large implementation team. An all-purpose enterprise platform is often more than a 5 to 500 fee-earner practice needs first.
Can a firm evaluate legal AI tools without outside help?
Yes. Many firms can work through a structured checklist with a partner, a practice manager and the lawyers who draft every day. Bring in an independent specialist if that group is stuck, or if you want someone who is not selling the software to facilitate the discussion.