Aim Higher with AI

Most people use AI to do the same work faster. A reply that took ten minutes now takes two. That is a real gain, and it is where nearly everyone starts. It is also the smallest win on offer, because the work itself has not changed. You still picked what to ask for, and you asked for what you were already going to do.

The bigger gains come from changing the ask. There are three directions to push, and none of them needs a new tool or a paid plan.

1. Aim at the outcome, not the task

Every task exists to make something else happen. The overdue-invoice email exists so you get paid. The weekly status report exists so your manager spots trouble early. The shift schedule exists so the floor is covered without burning anyone out.

Hand AI the task and you get the task. Hand it the outcome and it can help with the task, then question whether the task is even the best route.

You can climb this in small steps. Picture a property manager who writes a monthly report for each owner:

1. The task. "Write this month's owner report from these numbers." Faster, same report.
2. The meaning. "Here are this month's numbers and last month's. Tell me what changed, the likely reasons, and what the owner should do about it." Now the report says something.
3. The decision. "Which two issues here will cost this owner the most if nobody acts this quarter? For each, give me the case for acting now and the case for waiting." Now you walk into the call with a recommendation.
4. The format. "This report exists so owners trust me and approve repairs before they become emergencies. Suggest a better way to get there than a monthly PDF, then check it against last year's reports: which problems would it have caught sooner?"

The fourth rung is a conversation with your manager before it is a prompt. Asking the question still costs nothing.

A sign you should aim higher: you spend more time in meetings about a piece of work than doing it. The meeting is where the real outcome lives.

The same move works on chores that repeat. Instead of asking AI to sort out this week's scheduling clash, ask it to help you design a schedule that stops the clashes. A fix you make once beats a fix you make every Friday.

2. Ask for the expert version

AI will gladly hand you a tidy summary. It can also work the way the best person in your field would, if you ask. Four habits get you there.

Start from the raw material. Summaries hide the interesting parts. Give it the actual export, the whole email thread, the full spreadsheet, not your notes about them. (Strip names and private details first. More on that below.)

Ask what the pros use. An experienced analyst does not eyeball two columns of numbers. They pick a method that fits the question. Ask "What method would a seasoned [your job] use to analyze this, and why that one?" and then ask it to use that method.

Test what you already believe. Every team runs on a few theories nobody has checked. "Most complaints come from new customers." "Mondays are always slow." Give the AI the data and ask it to try to prove the theory wrong, not to confirm it.

Don't let it grade its own homework. An AI saying its answer is right tells you nothing. Check against something outside the conversation: the official source, last year's actual figures, a colleague who knows the area, a calculator. For anything you will act on, decide up front what would prove it wrong.

Not sure what "deeper" would look like for your task? Ask before you start:

Here's a task I'm about to do: [describe the task].

Don't do it yet. First, tell me how the best person in my field would approach it: what they would check that I haven't mentioned, what raw information they would want instead of a summary, and how they would prove the result is right. Then rewrite my request so it asks for that version.

Run the rewritten request. Compare the result with what you would have asked for.

3. Borrow from other trades

Plenty of good ideas come from someone in a different field looking at your problem. AI has read about most fields, so you can bring them in on demand.

Describe what the work has to do. "A newsletter" points to newsletter advice. "Getting busy parents to read something on their phone and do one thing before Friday" points to marketers, app designers, and kindergarten teachers. Rewrite the task as the job it has to get done, and the useful fields show up on their own.

Pull in a panel. A board presentation can use a journalist's nose for the headline, an editor's red pen, and a skeptical investor's questions. A safety briefing can borrow from airline crews. Ask which professions handle problems like yours best, then have the AI apply each one's methods in turn.

Ask for their methods, not their voice. "Pretend you're a detective" gets you a costume. "Check this the way an auditor would: list each claim, find what backs it up, and flag anything with nothing behind it" gets you an audit. Name the work, not the character.

Here's what I'm trying to make happen: [describe the goal as what it needs to achieve, not as a document type].

Name three professions outside my field that solve a version of this problem well. For each one, explain the specific method they would use and apply it to my situation. Finish with the one idea most worth trying first.

Why most of us under-ask

The ceiling is usually what we believe the tool can do, not what it can do. That pattern is much older than AI. Decades of goal-setting research found that people handed a specific, hard target outperform people told to "do your best." And once Roger Bannister ran a mile in under four minutes in 1954, after years of people calling it impossible, another runner did it within two months.

Three habits push the ceiling up:

• Set the bar above what you need. Ask for the version you would be proud to show your boss's boss, and say plainly that the target is realistic. AI tends to stop at the bar you set, so set a high one.
• Watch someone further along. Seeing a colleague pull something off is the fastest way to believe you can. Ask around what people have tried, or browse the guides for your role on this site.
• Make failure cheap. Try the ambitious version on something low-stakes: an internal draft, last month's data, a slow afternoon. If it flops, you lost twenty minutes and learned where the edge is.

The tools also improve every few months. Keep a short list of things AI could not do for you, and try them again each season. Some of them will have quietly become possible.

Bigger asks, same judgment

Aiming higher usually means sharing more of your real data. Before you paste customer, patient, or employee details, remove names and identifying details, and check what your employer or regulator allows. It also does not change who is accountable. AI can build the case. You still make the call.

Try it this week

Pick one task you will do anyway. Before starting, run the "best person in my field" prompt above. Do the version it suggests, and notice what you would have missed.

This article was prompted by Felix Rieseberg's essay How to be more ambitious, which is worth reading in full.