WATCH AND LISTEN ON
The companion
AI gives you back hours, and your brain, raised by grafters, immediately tries to spend them on more work. That is the trap in listener Rebecca's question, and the whole episode: the skill now is not capacity, it is restraint. Let the machine multitask, because you can't, and a good life is two parts contentment, one part desire, and no part refreshing the chat window.
1. Build the small language model of you
Kirsty's summer trick: a Claude project loaded with her book, every session synopsis she has ever written, plus the Gmail and Calendar connectors. When a time-sensitive opportunity landed mid-cooking, she voice-noted her phone from the hob and Claude drafted the reply in her voice, from her own material. She stayed at the stove.
“Look at the latest email from [name]. Draft me a reply. Use everything you know about me: my sessions, my synopses, my calendar, my tone. It's a draft, I'll review it before anything goes anywhere.”
Leader's note: Kirsty's edge was never glamour, it was being easy to work with and quick to reply. This keeps the responsiveness without surrendering your evening. The basics are what separate people; now the basics can be delegated without being dropped.
2. One step at a time
Henry's fix for the wall of bullet points every AI dumps on you after a task. Force it to report back one clear next action at a time, numbered, then stop. Your working memory gets one thing, which is all it can hold anyway.
“From now on, when you report back, give me one next step at a time as a short numbered list. One clear action, then stop and wait for me to say done before you give me the next. No paragraphs.”
Leader's note: this is also just good delegation, in both directions. If your updates to your team read like an AI dump, same fix applies.
3. Wince and send
Kirsty's own invention, now circulating under other people's names, which she has decided is a compliment. If something is scary to put out there, the pitch, the price, the post, you wince and you send it. The wince is the signal it matters, not the signal to stop.
Leader's note: teams copy what leaders visibly do. Wince and send in front of them once and you have given everyone permission.
1. The kettle rule
Next time you set an AI agent on a task, do not open a second chat window. Put the kettle on, do a plank, stare out of the window until it is done. Then go further: block your week into AI chunks and deep-work chunks, and stop flicking between them, because the flicking is just multitasking in a new coat. When the agent starts working, your job is to stop.
2. Discipline or gentleness
Henry prescribes cognitive discipline for the dopamine fairground. Kirsty would walk out of any room that tried it on her; she wants gentler, kinder, softer. Both are right, for different people. Work out which register actually moves you, then ask each person you lead which one moves them, because the one you naturally reach for is probably just yours. We often give the advice we need to hear the most.
3. Success leaves clues
Pick one person a few steps ahead of you and properly study them, not to copy but to model, the way Kirsty studied Lisa Nichols and Kindra Hall. Then accept the uncomfortable half: somebody on your team is doing this to you right now. Decide what you want them to catch. “I would genuinely rather you remember my message and forget my name.”
Transcript coming soon. It will live right here, in full, for every episode.