Writing · Practical AI
Stop giving AI tasks. Give it a job.
Most owners delegate tasks to AI and save forty minutes a week. Delegating an outcome instead is the same technology and a completely different result.
Ricky Bell · 25 August 2026 · 5 min read

Here is a question worth sitting with. If AI is saving you time, why does your Friday still look the same?
I think the answer is usually in what you asked it to do. Most owners I speak to are delegating tasks. Summarise this email. Draft that quote. Tidy up this description. All of it works, and all of it saves a few minutes — but you are still the one who noticed the email, decided it needed a reply, opened the tool, pasted it in, and decided what happened next. You have not delegated the job. You have typed faster.
The distinction I want to give you comes from Allie K Miller, who writes about this better than most: the mistake is delegating tasks when you could be delegating outcomes. Same technology, entirely different relationship.
What an outcome looks like
Take a plumbing firm. The task version is: “write a reply to this enquiry.” You do that eleven times a week.
The outcome version is a job. Its goal is that no enquiry sits unanswered past one working day. It checks what came in overnight, first thing. It drafts replies to the routine ones in the firm’s own tone. Anything that needs a real decision — a price, a promise about a date — it puts in front of you with a recommendation attached. Anything angry or urgent skips the queue entirely and reaches you straight away. It answers from your actual price list and your actual availability, not from its imagination.
Read those two paragraphs again. The second one is not a bigger instruction. It is a different kind of thing.
The anatomy of a job worth handing over
- The goal — one sentence. What is this job for? Everything else follows from it.
- The cadence — when it runs. Every morning, every hour, whenever something arrives.
- The boundary — what it may do on its own, and what it may only draft.
- The escalation rule — what must interrupt you immediately rather than wait for the batch.
- The sources — which price list, which calendar, which FAQ. Never “whatever you know”.
- The standing objective — the commercial thing you always want pursued, quietly, in the background.
Firstly, write those six down for the job you most want off your desk. Secondly, notice how much of that you have never actually written down for the humans either. That is not a coincidence — it is the same clarity problem, and this is a good excuse to fix it.
Where I would keep a hand on the wheel
Ownership is not the same as unsupervised. Anything irreversible, anything with a number in it, and anything that leaves the building with your name on it should still stop at a human. The system does the noticing, the gathering and the drafting — which is where the hours actually live. You do the deciding, which takes seconds.
That split — the machine’s persistence, your judgement — is the whole design philosophy behind what we build. I have written about why we deliberately give the AI as little to do as possible.
Overall: if AI is saving you forty minutes, you have given it a task. If it has given you back an afternoon, you have given it a job. Our free audit names which jobs in your business are the best candidates, with rough hours against each one. Because the hard part was never the tool. It was deciding what to hand over.
Frequently asked
What is the difference between delegating a task and delegating an outcome to AI?
A task is a single instruction you issue and then act on — “draft this reply”. An outcome is a standing responsibility with a goal attached — “no customer enquiry goes unanswered past one working day”. The first saves you typing. The second removes a job from your desk.
What should an AI job description contain?
Five things: the goal in one sentence, a cadence (when it runs), an autonomy boundary (what it may do alone), an escalation rule (what must reach a human immediately), and the sources it is allowed to draw on. Add a standing commercial objective if the job has one.
Is it safe to let AI own a job in a small business?
It is safe when the boundary is explicit. Give it ownership of the work and the drafting, but keep a human sign-off on anything irreversible or anything that leaves the building. Owning a job does not mean working unsupervised.