Practical editorial guidance, not a promise of income or a hands-on product benchmark. Links are standard, non-affiliate links on this site. Read our editorial approach.

A useful AI workflow starts with a job, source material, and a clear review point. The goal is not to outsource judgment. It is to make your thinking easier to examine and your drafts easier to improve.

The prompts below are starting templates. They are not a test of any particular model and they do not guarantee factual accuracy. Use an authorized tool and suitable data. Keep confidential, regulated, or identifying material out of public examples.

1. Find the assumptions in the offer

Start with something you actually wrote. Ask for criticism rather than a more impressive-sounding version.

A prompt to adapt
Read this offer. Identify what a buyer may misunderstand, what evidence is missing, and where the scope is too broad. Separate observations from assumptions. Do not invent clients, results, testimonials, or proof. Offer three questions I should investigate next.

Review whether the criticism is relevant to the customer you are trying to help. You do not have to accept every suggestion.

2. Prepare for a customer conversation

Give the assistant the customer situation and what you do not yet know. Ask for questions about past behavior, not leading questions designed to earn approval.

A prompt to adapt
Help me prepare questions about the most recent time a person experienced [problem]. Focus on the situation, current workaround, consequences, and previous attempts to solve it. Flag questions that lead the interviewee toward my proposed solution.

The resulting questions are preparation. They are not customer evidence until a real conversation takes place.

3. Compare smaller versions of the idea

Instead of generating a huge feature list, ask for the smallest test of the riskiest assumption.

A prompt to adapt
Here is my idea and the uncertainty I need to resolve. Suggest three smaller tests: a manual version, a simple prototype, and a conversation-based test. For each, state what it can teach me, what it cannot establish, and what I would need to observe.

Choose the test you can run responsibly with the time and access you have.

4. Turn a process into a checklist

Describe a process you understand, including where it can fail.

A prompt to adapt
Turn these process notes into a checklist. Include the trigger, required inputs, responsible person, human approvals, failure cases, and a manual fallback. Put unknown details in a questions section instead of guessing.

Have the person who does the work check the steps. A clean checklist can still be wrong.

5. Review the week without inventing a success story

Bring specific actions and observations. Ask the assistant to keep interpretation separate.

A prompt to adapt
Using only these notes, summarize what I did, what I observed, and what remains uncertain. Suggest one next experiment. Do not convert interest into revenue, estimates into actuals, or an unfinished task into a completed result.

Then decide the next move yourself. The assistant can help organize the decision; you still own it.

A simple quality check

Before using an output, ask: Is it faithful to the source? Which claims need checking? What did it assume? Who could be affected by a mistake? Which person must approve the next action?

Start with drafts and internal planning. Add automation only after you understand the failure cases and have a suitable review process.

Turn the read into a next step.

Use the free worksheet to put this into practice. No account or email required.

Get the worksheet

Need a little help applying this? Explore coaching and implementation.