Routines: putting the business on a schedule

transcript

The promise on the homepage is that you ship while you sleep. Routines are how that actually happens. A routine is a piece of work the colony runs on a schedule without you starting it: a Monday planning pass, a nightly metrics digest, a Friday customer update, a monthly close. You define it once and it runs until you turn it off.

A good routine has a trigger, a job, and an output. The trigger is when it runs, like every weekday at 7am or the first of the month. The job is the work, expressed the same way you brief any task. The output is what shows up when it finishes: a message to you, a published artifact, an updated record. If a routine produces nothing you ever look at, it should not exist.

Start with one routine, not ten. The best first routine for most Chairmen is a morning brief: every weekday before you start, the CEO compiles what happened overnight, what is waiting on your approval, where spend stands, and what it plans to do today. You wake up to a company that already caught you up. That single routine changes how the product feels more than any other.

Routines respect every control you have already set. A scheduled job still hits your approval gates, still counts against your budget, still appears in the audit trail. A routine cannot do anything an agent could not do in a live conversation; it just does it on a clock. That means you can trust a routine exactly as much as you trust the agent that runs it, which is the point.

The thing to watch with routines is drift. A routine that made sense in month one can quietly become noise by month three, either because the business changed or because it never delivered. Review your routines monthly. Keep the ones you read, fix the ones that are close, and kill the ones you skip. A small set of routines you rely on beats a calendar full of jobs you ignore.

A practical warning about scheduling: routines compound. One routine is a habit; ten routines is a second job reading their output. The colonies that drown are not the ones with too few routines, they are the ones with a calendar full of jobs that each seemed like a good idea in isolation and that together produce more reports than any human will ever read. Add the second routine only when the first has proven it pays for the attention it costs, and hold that bar for every one after.

There is also a difference between a routine and a reminder. A reminder tells you to do something; a routine does the thing. If you find yourself building routines that just nudge you to act, you are underusing the colony. The whole point is that the work happens on the schedule, not that you are prompted to do it. Push each routine toward doing rather than nagging, and let your gates catch anything that genuinely needs your sign-off.

When a routine proves itself, the natural next step is to make it richer: the morning brief that started as a summary becomes a brief with recommendations, then a brief where the CEO has already queued the obvious actions for your one-tap approval. That progression, from reporting to recommending to acting within your gates, is the whole arc of running an AI-native business. Routines are where you watch it happen.

copy-paste prompts

Create a routine that runs every weekday at 7am. Each run, give me: what happened overnight, what is waiting on my approval, our spend against the monthly cap, and your plan for the day. Keep it under one screen. Confirm the schedule and show me what the first brief will look like.

List every routine we run, when it runs, and the last time its output mattered. Recommend which to keep, which to fix, and which to retire.