Case study: the solo practice that hired an associate for the price of a seat.

Details here are composited from the pattern rather than a single identifiable practitioner, but nothing about it is exaggerated. Call him Daniel. A solo practitioner running a full book on his own — the clinical work he trained for, plus the intake, the notes, the follow-ups, the rebooking, the invoicing, and the low background hum of clients he meant to check on and never did. He was not short of skill. He was short of himself. There is only one of him, and the practice needed about two and a half.

He had resisted AI for the honest reason: it felt like the thin end of a wedge that ended with clients getting form letters from a machine. He was right to worry about that ending. He was wrong to think it was the only one. The version that worked kept him more present with clients, not less — because it took the edges, not the centre.

the context: drowning at the edges

The specific failure was always the same. A new client’s intake would arrive the night before, and he would skim it at 9pm because there was no other time, walking into the session with half the picture. Follow-ups he fully intended to send would evaporate under the next day’s load. Clients he had not seen in two months sat in a mental list he never worked through. None of this was a competence problem. It was a memory-and-time problem, and it was quietly costing him both quality and clients.

the shift: set the gate, then induct the worker

The first thing he did was not a prompt. It was a rule, written at the top of every workspace: this drafts, I decide, nothing goes to a client without a human sign-off. Only then did he induct the worker. He wrote a one-page induction — how he practised, his scope, what a good client summary looked like to him, his tone, the things he never said — and made it the standing brief. From that point on he was not writing clever prompts. He was refusing to re-introduce himself and his standards for every single task.

the approach: a room per client, three delegated drafts

For each client he built a room that held the induction plus that client’s real file — the raw intake, not his paraphrase of it. Then he delegated three things and read each like a supervisor reading a junior’s work. First, a summary of the intake. Second, session prep in the form of an interview: what are the ten questions this history raises that I have not resolved, what contradicts what, what did I say I would follow up on and never did. Third, a first-draft follow-up after each session. He corrected freely, kept the good, binned the rest, and signed off every word himself.

The interview move was the one that changed his clinical experience, not just his admin. It did not tell him what to do. It did what a good colleague does over coffee — asked the questions that made him sharper before he walked into the room. His attentiveness had been degrading with volume, and it turned out that was a memory failure, not an effort one. A room that re-read the whole file before every session handed the memory back.

“The associate he could never afford was not the chat box. It was the induction, the room and the clock — assembled around the one thing that stayed a human: his sign-off.”

the observable outcome: the evening came back

The measurable shift was time, and where it went. The intake summaries and follow-up drafts that used to eat his evenings became fifteen-minute review-and-approve jobs. The weekly sweep he scheduled — a list of who was due a follow-up, each with a draft check-in waiting — closed the gap between ‘I should contact them’ and it happening, which quietly improved retention without a single new marketing move. He did not see more clients per hour. He saw the same clients with more of his attention, and got a couple of evenings back that used to belong to admin.

the line that keeps this safe

Two commitments held it together as he leaned on it more. First, he drafts and I decide — nothing reached a client without his sign-off, however good the draft looked. Second, he was honest with clients about where AI helped in his process and where his judgement did the work, and found that transparency built trust rather than eroding it. His value was never that he could remember everything. It was that he could be trusted to decide. The worker made the remembering cheap so he could spend more of himself on the deciding.

what to do this week

If you run a practice on your own, do not try to automate the centre. Take the edge that fails most often — usually the intake or the follow-up. Set the sign-off rule first, write a one-page induction, build one client room, and delegate one draft. Sign off every output for a fortnight like a supervisor, then keep the version that hands you back the most evening.