
AI Is Not Your Assistant. It's Your Process.
Most people still talk about AI like it’s a better version of a human assistant:
- “It can summarize your notes.”
- “It can send follow-ups.”
- “It can book your meetings.”
Useful? Sure.
But the metaphor is narrow. And it can limit what we design.
AI doesn’t have to be your assistant.
It can be part of your process.
Assistants wait for instructions. Processes define the work.
An assistant needs you to say:
Book a meeting with Sarah next week.
An illustrative example; Sarah and the details are invented. A process designed around your work could already know:
- You’re trying to close a partner deal.
- You always meet on Tuesdays.
- Sarah hasn’t replied in 3 days.
- There’s a board report due Friday.
It knows those things only if you’ve given it access to the deal record, your calendar and the email thread, and only as well as those records are kept up to date. With that context, it can propose a meeting, or send an invitation within limits you’ve set, in a way that moves the deal forward without asking you for every micro-decision.
That’s not assistance.
That’s orchestration, and it depends on boundaries.
Who said it could send?
In the example, the invitation goes out because you, as the deal owner, authorised a defined kind of message in advance: an invitation to an existing contact, in one of your Tuesday slots, about a deal that’s already open. That’s the limit. Anything that changes the scope of the deal, proposes terms or commits you to something new needs your approval first.
If it sends the wrong thing, such as the wrong date or an invitation to the wrong Sarah, it stops sending anything further on that deal and tells you, and you put it right with a correction, an apology or a new invitation. Cancelling an invitation doesn’t undo the fact that Sarah has seen it. That’s why the boundary matters before the send, not after.
From reactive to proactive, within limits
We’ve been building tools that wait for prompts. More leverage comes when the system anticipates needs and quietly removes friction, inside the limits you’ve set:
- Not “write a summary” — but “Here’s what’s changed since last time.”
- Not “send an email” — but “This went out under the rule you approved. Here’s a copy.”
- Not “check in with the team” — but “Here’s what’s blocking progress.”
The real shift isn’t speed. It’s scope.
Assistants speed up tasks. A well-designed process can take some tasks off your plate entirely: the routine ones that are cheap to get wrong and sit inside a boundary you’ve drawn. Others are better left as assistance, where the system drafts and you decide. A good process doesn’t make your judgement obsolete; it lets you spend it where it counts.
That’s the useful difference between automation and autonomy. Autonomy is authority you grant, for specific actions, with someone accountable for what happens.
When not to hand it over
With Sarah, keep the invitation as a draft for you to send when:
- the deal is at a sensitive stage, or her silence might mean something
- the message would propose terms, change scope or mention money
- the calendar or deal record is out of date, so the context is wrong
- you haven’t yet decided what it’s allowed to send
Your next step
Choose one recurring task you’d like to stop doing by hand, and write down four things: what it may do without asking, who approves anything outside that, who checks the results and how often, and what happens when it gets one wrong. The autonomy test walks through those questions in more detail. If you can’t answer them yet, keep the task as assistance until you can.
Updated:
Related
Want help turning ideas like these into reality? Let's talk.