Use It Yourself First: An Executive's First Month with AI

In many companies the AI agenda starts with the boss forwarding an article to the leadership group chat, captioned "we need to catch up." The irony: the person forwarding it has often never seriously used AI. Everything they know comes from articles, dinner conversations and second-hand stories.

Knowledge can come from other people; judgment cannot. To judge whether a vendor's proposal holds up, to answer an employee asking "will this replace me", to decide how much to spend — all of that needs first-hand feel, and feel only comes from use. Here is a week-by-week path: one month, twenty to thirty minutes a day.

Why this cannot be delegated

Three things depend on the boss's own feel. Judging vendors: someone who has never used AI cannot tell the gap between demo performance and everyday performance, and is easily impressed or easily scared. Answering the team: employees will ask whether the thing is reliable and what happens to their role, and second-hand answers convince nobody. Sizing the investment: how much to spend and what to expect should rest on your own experience, not on someone else's slide deck.

A staff research report helps, but it is no substitute for feel — just as reading someone else's notes on driving is no substitute for taking the wheel.

Week one: hand over your information work

Start with the lowest-risk, highest-frequency work: reading and writing. Have AI summarise long emails and reports before you decide what deserves a full read. Give it a few points and ask for a first-draft outline of a talk you need to give. When you hit an unfamiliar industry term or policy clause, have it explained in plain words.

The goal this week is not saving time. It is calibration: how well does it handle material you hand it — do the summaries catch what matters, do the outlines hold together? Jot down one or two things it did well and did badly each day; you will need those notes in week four.

Week two: brief it the way you brief a subordinate

Most people hit the same frustration in week one: generic answers, all correct and all useless. The problem is usually the asking, not the tool. This week, practise a single habit — every request carries three things: context (who you are, what the situation is), standards (what a good result looks like, what is unacceptable) and format (how long, what structure, for which reader).

You will notice this is the same skill as delegating to people: the clearer the brief, the more usable the output. Later, the habit can harden into a shared team asset — see Turning Prompts into Team Assets.

Week three: let it prepare a real decision

Pick something you are genuinely weighing — entering a new channel, adjusting a price — and have AI do the preparation: lay out the arguments for and against; list the assumptions your reasoning depends on and flag which ones need verifying; then have it take the opposing side and attack your plan.

Hold two boundaries. First, it organises and challenges; it does not decide — the conclusion must be yours. Second, do not paste sensitive material in casually: for customer, financial and contract information, think first about what it may see. This deserves a company-wide rule — see Four Data Boundaries to Set Before Using AI on Company Material.

Week four: write your own judgment list

Use three weeks of notes to answer two questions: what kind of work does it do well, and where can it not be trusted? Most people's lists converge. Strong: condensing, first drafts, reformatting, attacking a position from another angle. Untrustworthy: specific facts and figures (it fabricates with confidence), anything internal to your company (it has no idea), and judgment calls someone must answer for.

The list is yours; it does not need to match any expert's. Its value shows up later: at the next vendor demo you know what to press on, when the team raises doubts you have answers, and when budget comes up you know which gap the money should close.

The three most common ways to go wrong

Treating it as a search engine — one factual question, one wrong answer, and the whole thing gets written off, though its strength was never lookup but processing what you feed it. Expecting mind-reading — giving no context, then blaming the tool for mediocrity; back to week two. Pasting in sensitive material — ten minutes saved now, a long-tail liability planted; back to the boundaries of week three.

Feel is where organisational judgment starts

A month in, you will not be an expert, and you do not need to be. You will have something more valuable: judgment you tested yourself — where this class of tools is strong, where it breaks, when not to trust it. Every AI decision the company makes afterwards — vendors, budgets, training — stands on that base. The boss's first-hand feel is the starting point of the organisation's judgment.