Most firms looking for a Harvey or Legora alternative in the UK are answering four questions at once. Where the work lives, and whether the tool has to reach into a document management system today. Which corpus the practice depends on. What the COLP must be able to show. What the firm can sign without a procurement process. Answer those four in order and the shortlist thins on its own.
What each of these products is built around
Every serious tool in this market is built around one thing, and that thing explains what it does well. Harvey presents itself as an enterprise product, US-led and document-management-first, on its own published material. Legora presents itself as built around table-style review and collaboration between fee-earners. CoCounsel, from Thomson Reuters, is described by its publisher as built on the Westlaw corpus, and Lexis+ AI on the Lexis corpus, so the licensed content is the product and the assistant is the way into it.
Copilot and ChatGPT are general assistants, useful for drafting but not sold as legal verification tools. LegalAI Space is built around the matter and around closed-world verification of UK and EU authorities. This page will not tell you what any competitor charges, or how any competitor verifies its output, because we do not know either with the precision that would make a statement safe to publish.
Where the work lives
Ask where a fee-earner would find the second draft of a share purchase agreement at four in the afternoon. If the answer is iManage or NetDocuments, and your records policy discourages copies of matter files leaving it, a live DMS connector is not a feature on a list. It is the implementation.
We do not have one. iManage, NetDocuments and SharePoint appear as tiles in our connector catalogue; they are not live integrations. In our product, documents are added to the matter directly. For a firm with a disciplined DMS and a records policy built on it, that is worse, and Harvey's document-management-first position is a real advantage.
The corpus the practice depends on
Research splits into two kinds. Some depends on licensed material: headnotes, a citator, editorially maintained commentary. Some runs on primary sources: statute, judgments, regulator handbooks, EU instruments. Be honest about which one pays your bills. If your litigators check whether a case is still good law with a citator every day, a tool built on the corpus you already licence may suit you better than we will.
We hold no licence to that content, so we refuse it: Westlaw, Practical Law, Lexis, Bloomberg Law, Justis and vLex sit on a deny-list of ten paywalled domains. We cannot fetch those pages, so we cannot quote them, so we cannot verify them, and an unverifiable authority is the kind that caused the trouble in Ayinde v London Borough of Haringey and Al-Haroun v Qatar National Bank. Our allow-list runs to 74 domains, including legislation.gov.uk, BAILII, Find Case Law, EUR-Lex and the FCA, ICO and SRA handbooks.
What you can show your COLP
Since Ayinde and Al-Haroun were decided by the Divisional Court in June 2025, the question in front of a risk partner is no longer whether the firm uses AI. It is what the firm can produce, after the event, about how one particular piece of output was checked, and by whom. The SRA has addressed this ground in its Risk Outlook report (November 2023), its thematic review of compliance officers (December 2025), and its guidance update on new technology and COLP responsibility (February 2026).
In our product, the record is the audit and compliance register: runs, verification results, sharing, documents, members and conflict checks, exported to CSV, plus a printable inspection bundle carrying a content digest. The digest lets you show that the bundle you handed over matches the bundle that was produced. It is not a tamper-evident hash chain, and we do not describe it as one.
What the firm can sign
Every firm has a threshold. Below it a managing partner signs; above it a committee meets, with a security questionnaire, a data processing agreement and a partner vote. That threshold decides more legal AI purchases than any feature comparison. Above it, six months later, the pilot either works or it does not, and the person who championed it has spent something they cannot easily earn back.
Our free plan gives a firm two seats and 500 welcome credits, which expire after thirty days, with the standard agent set. Credits are firm-level and one credit is roughly a penny of model usage. A twenty-lawyer firm can run a real matter through the product before anyone books a meeting about it. That does not make us the right choice: a firm that needs a negotiated information-security schedule should run the full process, whoever it buys from.
Where we are the wrong answer
A global firm with iManage and a US practice should probably not buy us: we have no live DMS connector, so matter files would be copied rather than reached, and we have no US law coverage. A firm whose research depends on headnotes and a citator should keep Westlaw or Lexis, and a tool built on that corpus may suit it well.
A firm that wants redlining inside Word should look elsewhere, because we produce a Word file and do not live inside one. A firm that needs ethical walls as a technical control should not accept ours as a substitute: we have restricted matters and conflict redaction, which a COLP will know is a different thing. And a firm whose procurement requires a compliance certificate from the vendor will not get one from us today.