Give AI a Job and Get More from the Same 24 Hours

This article began in a parking lot at a swimming pool. I was waiting for my son, sitting behind the wheel, talking to AI by voice. I wasn’t at my desk and I wasn’t typing. I had a little time and an idea worth exploring.

We talked about what AI could mean for people who run their own businesses. Then I asked it to turn the discussion into an article I could review when I got back to the office.

A conversation is a good place to begin. But the business value starts with a different question: what job could I give this thing?

We all get the same 24 hours. I’m interested in how much more useful those hours can become.

From asking questions to assigning work

Most people start with AI by asking it questions. Explain this. Rewrite that. Help me think through an idea. That’s useful, and a good conversation can save real effort.

But running a business involves work that comes back tomorrow, and the day after. The inbox fills again. Another customer wants an appointment. Another enquiry needs the same information you sent last week.

If you have to start a new conversation and explain the task every time, you’re still carrying the administrative load. You get a faster answer, but you’re still the one setting everything in motion.

Giving AI a job means being clear about the result you want, the information it can use, and when it should come back to you. That’s the practical meaning of an AI agent: a system with an assignment, access to the tools it needs, and limits on what it may do. I care less about the word “agentic” than whether the job gets done properly.

Start with the inbox

Email is a sensible place to start, because so much business work arrives there mixed in with things that don’t deserve your attention. A sales enquiry, a supplier update, a complaint and several newsletters land together, and you have to read enough of each to decide what matters.

An email triage job could sort those messages using rules you’ve agreed on. New enquiries get attention. Complaints are flagged for you personally. Routine updates are grouped. Anything uncertain stays visible rather than being quietly dismissed.

It could then prepare replies wherever the business has clear, approved information. If a prospect asks about services, the draft should draw from your actual service descriptions. If someone asks about availability, it should use a current, authorized source, or flag that you need to check.

It shouldn’t invent a price because a reply sounds better with one, or promise a delivery date because the customer seems impatient. A polished sentence can still be wrong.

I’d start with drafts that need your approval before anything is sent. You see the incoming message, the proposed reply, and any question the AI couldn’t resolve. Then you correct it, approve it, or handle it yourself. I’d much rather open a draft and make a decision than face another blank screen.

Connect the calendar to the job

Scheduling looks small until you add up the interruptions. Someone asks to meet. You check your diary and offer times. They suggest another. You check again.

With the right calendar connection, an AI workflow can suggest options based on your real availability. But an empty slot isn’t automatically a good time. You may need travel time between jobs, a buffer after a customer meeting, or a protected morning to finish proposals, and personal commitments matter too. The system needs those rules before its suggestions are worth anything.

Start with suggested times and a reply for you to approve. Allow it to book or change appointments later, once it has proved reliable. And make sure the whole job fits together. If you have to repair the handoffs between checking availability, offering times and confirming the booking, you’ve just moved the work around.

Work that happens without another prompt

This gets more interesting when a job runs on its own, at a set time or when something happens. Some work should happen because it’s Monday morning or a new enquiry has arrived, not because you remembered to ask.

You might want a short morning briefing with today’s appointments and the messages that need you. Or a weekly list of enquiries awaiting follow-up, with proposed next steps ready for review.

Not every AI product can do this out of the box. Ask whether it will actually run while you’re away, and how you’ll know if it fails. And make sure it gets your attention sensibly. If every routine action sends another notification, you’ve just traded an email problem for an alert problem. Ask for a short summary, and an immediate alert only when something warrants it.

Voice gives ideas somewhere to go

The parking-lot conversation showed me something else. Sometimes the obstacle isn’t how hard a task is, it’s the effort of getting started.

If writing an article means sitting at a desk, opening a document and typing a first draft, it competes with everything else waiting there. By the time you sit down, the thought may have lost its shape.

Speaking is a different starting point. You can explain what you’re thinking, circle back, and work through an example while the idea is still fresh. My contribution was the perspective, the experience, and the judgment about what mattered. The AI’s contribution was organizing that into something I could work with. The review still mattered, especially whether the finished piece sounded like me.

Owners already think about their business between appointments, on the way to a job, and while waiting for someone. Voice turns that time into a practical place to capture ideas, like notes after a customer visit, an explanation of a recurring service issue, or the outline of a proposal. You don’t have to sound polished; you just need enough substance for a useful draft. (My example was a parked car. Reviewing and approving work can wait until you’re safely stopped.)

Not every spare moment has to be productive, either. Waiting can simply be waiting. The value is having the option when an idea arrives.

Choose one job and make it earn its place

I wouldn’t start by handing AI the whole business. Pick one repetitive task you know well enough to judge the result. Email triage or routine enquiry replies are good choices, because you can see the work and catch mistakes before anything goes out.

Write the assignment as clearly as you would for a person helping you. For example: every weekday morning, review new customer enquiries and prepare replies using my approved service information. Send complaints and pricing questions straight to me. Give me everything else as a short review list, with drafts clearly marked, and send or delete nothing.

Then check it. Did it find the enquiries? Were the drafts accurate? Did it leave the decisions where you wanted them? Give it only the access the job needs, keep a visible record of its work, and test awkward enquiries as well as easy ones.

Finally, measure the effort that’s left. If checking and correcting its work takes as long as doing it yourself, something needs to change. Maybe the assignment is too broad, the source material is weak, or that job really does need more of your attention.

Put the recovered time to work

For me, the appeal is spending more of the day on work that needs an owner’s attention: a customer relationship, a difficult decision, a good idea that would otherwise never get written down. And sometimes simply finishing on time.

So look at one piece of work that keeps coming back to your desk. Decide what good work looks like, where your judgment is needed, and what AI could prepare for you. Then give it the job and see whether it earns its place.

You’ll still have the same 24 hours. The aim is to spend fewer of them repeating work that could be prepared for you, and more of them on the business and the life you’re trying to build.

Want help setting up an AI assistant that actually does a job in your business? See how I handle AI and automation or book a discovery call.