Choose One Real Task Before You Choose Another AI Tool

You have access to AI. You have work to finish. The missing piece may be a small, repeatable way to connect the two.

My recommendation for a first workplace experiment: choose one task with clear inputs, a useful output and a result you can personally review. A campaign brief, client summary or onboarding draft gives you something concrete to inspect. “Make my whole business more efficient” is much harder to test.

This guide will help you choose the task, prepare the context and decide whether AI actually helped. You can use it on your own or bring the same preparation to a private AI Working Session.

Start with a task you understand well enough to review

For your first experiment, choose familiar work. Your experience gives you a way to notice omissions, weak assumptions and an answer that sounds plausible but does not fit the situation.

A useful candidate has four characteristics:

• It repeats. Improving it could help on another day, not just this once.
• It has a boundary. You can describe one input and one intended output.
• You can check the result. You know what accurate, appropriate and useful would look like.
• You can try it safely. The first attempt can stay a draft, using material you are permitted to share with the tool.

If the result would go straight to a customer, change a live system or make a decision you cannot evaluate, narrow the first experiment. Start with a draft or a fictional example and decide where a person must review it.

Turn a broad ambition into a usable brief

“Use AI for marketing” leaves almost every decision open. A bounded task gives the assistant context and gives you a way to judge the response.

• Campaign planning: Turn approved notes into a one-page brief for one audience and objective. Flag missing inputs instead of inventing them.
• Research: Summarize three permitted sources for one decision. Separate supported findings, interpretations and unanswered questions.
• Onboarding: Draft one message that helps a specific customer take one next step. Check its timing, promises and call to action.
• Weekly reporting: Turn an approved update into progress, blockers and decisions needed. Keep unconfirmed dates and owners visibly unconfirmed.

Write the review criteria before asking for a draft

Imagine handing the finished work to a colleague. What would they need to understand or do without asking you to explain it again?

For a campaign brief, your criteria might be: one defined audience; a stated objective; a message tied to the actual offer; evidence for factual claims; and a visible list of unresolved decisions. Add any real constraints, such as length, tone, approved terminology or a required format.

Now you have something more useful than “make it better.” You have a checklist against which to review the next attempt.

A fictional example: from campaign notes to a brief

Suppose a consultant has notes for a new advisory service. They know the intended audience and the kind of problem they want to address. The launch date and final price are not approved.

A polished draft that invents a launch date and a discount would create more work. A useful draft would organize the known information, propose options where requested, and leave the unapproved details in an “open decisions” section.

The important review questions are simple: Did the draft preserve the facts? Did it make the next decision easier? Can the consultant use the structure again?

A prompt you can adapt

Help me work through one task: [task].

The person using the result is [audience]. They need it to [decision or next action].

Use only the approved material below: [inputs]. The result should be [format], and it needs to meet these criteria: [review checklist].

First, identify missing information that would materially change the result. Ask up to three useful questions. Do not invent facts, dates, prices, quotes or customer evidence.

Then produce a first draft. Separate supported information from assumptions and open decisions. Finish with what I should personally verify before using it.

Review the response yourself. Asking an assistant to identify uncertainty can help organize your check; it does not establish that everything else is correct.

Measure the whole attempt

Record preparation, prompting, reviewing and rewriting time. Compare the final quality with what the task actually requires. A quick first draft is only part of the work.

After a few attempts, ask whether the workflow is worth keeping, changing or stopping. If it helps, save the inputs, context template, steps and review checklist. You now have a process you can repeat and improve, rather than a conversation you have to reconstruct.

Choose the support that fits your goal

For one immediate work task: the private AI Working Session gives you 90 minutes with me for $350 USD. We agree on one achievable outcome, work through it together and review the result. You keep a reviewed draft or example, an editable workflow, a context template and a quality checklist, with one follow-up question within seven days.

See the session and request a time →. We agree on scope, availability and booking terms before you pay. Your request continues on AI Teachers, my practical learning site; it does not reserve a calendar slot.

For developing a side-hustle idea with a group: Two Mondays to a Side Hustle is a separate workshop through The Wealth Experiment. We work on shaping an offer, building a small first version with AI and planning a customer test. See its page for current dates, pricing, preparation and what is included.

You do not need to take one before the other. A private session is focused on your task; the group workshop follows a shared learning plan.

If you are starting today, write one sentence: “The task I want help with is ___, and I’ll know the result is useful when ___.” That is enough to begin a more focused conversation.