Be AI-native where it saves time and money.

I want to be careful about where this series sits, because a manifesto that defends the profession can be read as a manifesto that defends the profession's habits, and those are not remotely the same document.

Think about what my first years in practice actually consisted of. Reading a bundle to find out what was in it. Retyping a chronology somebody else had already typed. Checking that the name on page four matched the name on page forty. Drafting the same paragraph of law for the ninth time that month with two details changed. Assembling. Indexing. Chasing. Every one of those tasks was necessary, and not one of them required a vow, and every one of them was billed at a rate that pretended otherwise. I was not exercising judgement for most of my week. I was performing administration in a suit.

If a machine will now do that first pass in minutes, then refusing it is not professionalism. It is nostalgia billed by the hour, and the client is paying for the sentiment.

The codes are not neutral on this either, which surprises people who assume the rules exist to slow everything down. SRA Principle 7 requires you to act in the best interests of each client. Paragraph 3.2 requires you to ensure the service you provide is competent and delivered in a timely manner. There is no provision anywhere in the Standards and Regulations that protects slow, and none that protects expensive. On the Bar side the point is made even more directly. Guidance gC65 says that nothing in rC20, the rule that makes a barrister personally responsible for their own work, prevents delegating or outsourcing to another person discrete tasks which that person is well equipped to provide. Delegation has always been lawful. The machine is a delegate, and a fast one.

So the AI-native half of the position is not a concession I make reluctantly before getting to the part I really care about. I want it. The intake questionnaire that arrives already structured. The first draft of the representations sitting in the file before I open it. The consistency check that finds the date discrepancy between the witness statement and the payslips before an entry clearance officer finds it. Research triage that tells me which of forty results are actually about my point. Bundle preparation that takes an hour instead of a day. Every one of those things gives me back time, and the time comes back at the expensive end of my working life rather than the cheap end.

And the client has to feel it. This is where I get impatient with parts of the market. If the marginal cost of the first pass has collapsed, and the fee has not moved, then the technology has changed nothing except the size of somebody's margin. The honest version is fixed fees you can quote at the outset without flinching, because you know what the work will actually take now; scope that is wider for the same money because the machine absorbed the grunt work; and matters taken on that used to be uneconomic, which for anyone working in immigration means people who previously got no help at all. That last one matters more to me than the efficiency argument. A great deal of unmet legal need is not unmet because nobody cared. It is unmet because the arithmetic never worked.

There is a version of this transition that goes badly, and it goes badly through drift rather than through decision. The economics pull in one direction the whole time, because supervision is the expensive part and the machine's output looks finished. Nobody in a firm ever announces that scrutiny is being reduced. It just gets thinner, matter by matter, as volume climbs and the drafts keep coming back plausible. A firm that lets that happen has not become AI-native. It has become cheaper, and it will find out what the difference cost somewhere around the second Ayinde-style headline or the first insurance renewal.

Which is the argument of this whole series compressed into a sentence: the two halves depend on each other. Being genuinely AI-native is what buys the hours that make real supervision possible. Being genuinely professed is what makes the speed safe to sell. Take either half on its own and you get a familiar failure; a firm too proud to use the tools, or a firm too busy to read what the tools produced.

I would rather spend my week on the parts of the job I trained a decade for. The machine has offered to take the rest. I intend to let it, and to be honest with clients about what that should cost them.