Question 1

What does an AI supervision system of record actually do?

Accepted Answer

It answers one question on demand: how was AI actually used on this matter, and who took responsibility for it. In practice that means a register of every AI tool the firm uses, independent verification of the outputs those tools produce, policies that run as code rather than sitting in a document, and a tamper-evident record of the lot. The COLP gets one place to look instead of several vendors' partial accounts.

Question 2

We already have an AI assistant. Why would we need this as well?

Accepted Answer

Because your AI assistant governs itself, and it only sees itself. The SRA's Risk Outlook on AI puts the first problem plainly: do not trust an AI system to judge its own accuracy. The second is arithmetic. Most firms in our range run two or three AI tools plus whatever staff use privately, and no vendor sees the others. We sit above all of them, and we do not ask you to remove any of them.

Question 3

What does the SRA actually require here?

Accepted Answer

There is no standalone SRA AI rulebook, and that is the point people miss. The SRA regulates outcomes rather than tools, so the duties that already bind your firm apply unchanged when a model does the work: effective governance and systems and controls, records that demonstrate compliance, and identifying and managing material risks. Its February 2026 compliance guidance adds that it would expect the COLP to be responsible for regulatory compliance when new technology is introduced.

Question 4

Does this change how our fee-earners work?

Accepted Answer

Very little, deliberately. Nobody has a tool taken away, which is what usually pushes AI use underground. Verification comes to the document through a Word pane, to the inbox by forwarding an output, or through the API for firms that want their tools wired in. The habit we are asking for is closer to copying in a colleague than learning a new system.

Question 5

What can we hand to a PI insurer at renewal?

Accepted Answer

An evidence pack drawn from records rather than adjectives: which tools were used, what proportion of AI-assisted citations verified, what was caught, and who signed each output off. Proposal forms are increasingly asking about AI, and the honest answer is much easier to give when it is a query against a record instead of a reconstruction.

Question 6

Is this an EU AI Act product?

Accepted Answer

No, and we would rather say so. The Act's high-risk obligations were pushed to December 2027 and mostly do not reach law-firm research AI. It is a framework we support, with Article 4 literacy tracking and Article 13 and 14 policies in the pack, rather than the reason to act. The reason to act is your regulator and your renewal date.

Question 7

How long does this take to stand up?

Accepted Answer

About a fortnight for a firm of 20 to 200 fee-earners: declare the estate, take the policy pack, run an attestation round, then let the evidence accumulate. The free readiness check tells you before you start which of those steps your firm actually needs. Governance runs as 19 checks on every governed task, so once it is on, it stays on without anyone maintaining it.

Reference	Source	Status	Action
Limitation Act 1980, s.32 Quote matched verbatim	legislation.gov.uk	Verified	None
Civil Evidence Act 1968, s.11 Quote matched, elision handled	legislation.gov.uk	Verified	None
Authority C (illustrative) Real case, but cited for a proposition the judgment does not contain	Find Case Law	Needs a check	Read the paragraph. The output overstates the holding.
Authority D (illustrative) No such case found in any allowed source	No match	Not found	Treat as fabricated. Remove before use.

Tool	Status	Risk tier	Owner	Evidence depth	Permitted use
LegalAI Space agents	Approved	Governed	COLP	Native evidence	All matter work
Harvey	Approved	High	COLP	Usage API	Research and drafting. Output must pass a check.
Microsoft Copilot	Approved	Medium	IT director	Audit-log ingest	Internal comms and summaries only
ChatGPT, personal accounts	Not permitted	Shadow	Unassigned	Attestation only	Blocked by policy. Staff attested.
A practice-area tool under evaluation	Under review	High	Head of KM	Manual, quarterly	Pending data protection review

The AI supervision system of record for law firms.

Your firm already runs AI. It probably cannot prove how.

Three layers, one evidence chain.

Govern the estate

Verify any output

Governed agents

Check anything your firm's AI produced.

Every AI tool in the firm, on the record.

Evidence a third party can check.

And if you have no AI tool yet, ours come governed out of the box.

The question is already being asked.

Find out where your firm actually stands.

The thinking underneath it

Why an AI Policy Is Not AI Governance

What the SRA Can Actually Ask Your Firm About AI (and How to Answer)

1,600+ AI Hallucination Cases: What Every Law Firm Should Learn

The questions firms ask first.

Ten minutes now, or an afternoon reconstructing it later.