The deep dive, eleven of twelve

The record proves the vow was kept.

The machine has quietly destroyed the accidental audit trail that used to sit in every file. Careful practice and careless practice now leave the same mark, which is a new problem and not one solved by being conscientious.

On Declaration XI of the manifesto: The record proves the vow was kept. By Chris Dias.

The professed seal, pressed in green wax, with the numeral XI at its centre

Everything in this series up to now has been about what a professed human does. This declaration is about the awkward question that follows, which is how anybody else would ever know.

Picture the complaint arriving eighteen months from now. A client says the advice was wrong, or a tribunal asks how a citation reached a skeleton argument, or an insurer wants to understand what happened before deciding whether to indemnify. The question put to you will not be whether you supervised the work. Everybody says they supervised the work. The question will be what you actually did, and the honest answer for most firms today would be a shrug and a sincere assurance about their normal practice.

Paragraph 7.2 of the SRA Code has an unglamorous elegance to it in this light: you are able to justify your decisions and actions in order to demonstrate compliance. Able to justify. Demonstrate. The Code for Firms adds at 2.2 that firms keep records to demonstrate compliance. The rule already assumes an artefact exists. At the Bar, Core Duty 9 requires you to be open and co-operative with your regulators and Core Duty 10 to manage your practice so as to achieve compliance, which are both duties you cannot discharge with recollection alone. What the manifesto asks for is not a new obligation. It is the artefact the existing obligation quietly presumes.

And AI makes the presumption urgent for a specific reason that has nothing to do with regulators being difficult. In the old workflow, the supervision was visible in the file whether anybody intended it or not. There was a draft with a trainee's name on it, then a version with my handwriting across the margins, then a fair copy, then a letter. You could reconstruct the thinking from the paper. That accidental audit trail has now gone. The machine produces a draft that already looks like a fair copy, and if I read it carefully and change nothing because it was right, the file records exactly the same thing as it would if I had never read it at all. Careful practice and careless practice have become indistinguishable in the record. That is a new problem, and it is not solved by being a conscientious person.

So the record has to be built deliberately. What I want in it is modest and specific. What the machine produced, and which tool produced it. What was verified, particularly anything citable, because that is the failure mode the courts have already seen. What was changed, and the reason it was changed. What was rejected outright, which is the most valuable entry of all, because a rejection is the clearest possible evidence that a professional was present and disagreeing. And at the end, a named professional's signature attached to that specific state of the work, rather than a general assurance floating over the matter.

The rejections deserve a moment. If a supervision record contains nothing but approvals, it is not evidence of supervision; it is evidence of a process producing approvals, which is what a rubber stamp also produces. The trail of things a human refused is the trail that proves the human was capable of refusing.

This is what my colleagues and I have been building through Countersigned, and I should declare that plainly rather than let it sit implied. The conviction underneath it is the one from Declaration VIII: AI errors in law are supervision failures rather than technology failures, and if that is true, supervision is something you can build infrastructure for. So the review is designed into the workflow rather than bolted on afterwards, each AI-assisted matter passes through layers of professional scrutiny, and what comes out the other end is an immutable record ending in a solicitor's signature. The act of professing, engineered to hold at machine volume.

I am aware of how that can be misread, so let me be exact about what a record is not. It is not the supervision. A beautifully complete audit trail attached to work nobody genuinely read is Declaration VIII with better documentation, and I would rather have an honest firm with a shabby record than a counterfeit one with an immaculate one. The record does not create the diligence; it evidences it. Which means the only firms for which this is worth building are the ones already doing the work.

But for those firms, the record is not bureaucracy. It is the thing that makes their diligence visible to a regulator, defensible to a court, insurable by an underwriter, and saleable to a client who is entitled to ask how their matter was handled. It converts a professional's private care into something the outside world can rely on, which is precisely what professing was invented to do in the first place; the vow was always taken out loud, in front of witnesses, because a private commitment protects nobody.

Accountability that cannot be evidenced is sentiment. I have met a great many sincere practitioners and never once been able to file sincerity.

Accountability that cannot be evidenced is sentiment.

Declaration XI

Next in the deep dive.

Twelve declarations, taken one at a time, each set against the rules that already bind solicitors and barristers in England and Wales.

Declaration XII: Any firm can be AI-native; every firm must remain professed All twelve