# Most people ask AI for answers. Your power users ask it about the work.

Source: https://tpiros.dev/blog/ask-ai-about-the-work

Picture an account manager with a client renewal on Thursday. They paste the contract into the company's AI assistant and type "summarise this". A tidy summary comes back in 10 seconds, they skim it, and they walk into the meeting knowing roughly what they knew before.

A colleague with the same contract types something longer:

```
I'm meeting this client on Thursday to renew the contract below.
Before you answer, ask me what you need to know about the account,
one question at a time. Then tell me what they're likely to push
back on and what I should have ready.
```

Both of them used the same tool on the same document. Only the second one walked in on Thursday with a plan.

## The box looks like a search engine

Most of us learned what to do with a text box and a blinking cursor on Google. You type a few words, you get an answer, you move on. An AI assistant gives you the same box, so people bring the same habit to it.

I think that habit explains a lot of what EY found in its [2025 Work Reimagined Survey](https://www.ey.com/en_gl/newsroom/2025/11/ey-survey-reveals-companies-are-missing-out-on-up-to-40-percent-of-ai-productivity-gains-due-to-gaps-in-talent-strategy): 88% of employees use AI at work mostly for search and summarising. The 5% who use it in ways that change how they work are typing something different into the same box.

## What power users ask about

The difference is the subject of the question. A search-box question is about a fact or a document. A power user asks about the job the document is for: who'll read it, what has to be decided, and what could go wrong.

Here are 4 questions that do that, and they work on almost any piece of work:

```
Before you start, ask me what you need to know. One question at a time.

Read this as the finance director who has to approve it.
What would they push back on?

What are you least sure about in what you just wrote?

What should I be asking you that I haven't?
```

The first one turns a vague request into a brief. People skip writing briefs because it feels like the work they were trying to avoid. Letting the AI ask the questions gets you the brief without the blank page.

The second puts a reader in the room. Most work is written for somebody, and an AI playing that somebody finds the weak paragraph faster than you will rereading your own draft for the fourth time.

The third catches the confident mistakes. It tells you which part to check before anything goes out.

The fourth sounds like a lazy question. Asked at the end of a job, it can come back with the question your manager would have asked in the review, which is a better time to hear it.

## You still check what comes back

Better questions get you more useful answers. They don't make the answers correct. You still check the numbers, the names and anything going to a client, and the third question tells you where to start.

## Easy to teach

Questions are the easiest part of using AI well to teach, because they don't depend on the tool. They work the same way in ChatGPT, Claude and Gemini, and they'll probably work in whatever comes next.

Give someone those 4 questions, have them try each one on a piece of their own work this week, and the box stops looking like a search engine to them.

That's a big part of what I teach on [the Power User Gap](/power-user-gap/), on a team's own workflows.
