Your buyer doesn’t need another deliverable

Your buyer has never opened their inbox hoping someone finished an outbound sequence.

They have a business to run. A problem they haven’t solved. A decision they need to feel good about before they put their name on it.

Meanwhile, we’re celebrating because the emails are written, the ads are built, and the category research is done.

That gap is what I’m talking about when I say we’re optimizing for the production of a deliverable instead of the execution of a high-quality playbook.

“I built AI agents to research my category.” Cool. Which unanswered question could change how you compete?

“I built AI to write outbound emails.” Cool. Which buyer problem are you communicating more clearly than everyone else?

“I built AI to create ads.” Awesome. What will the buyer see in those opening seconds that makes the benefit immediately recognizable?

Imagine a marketing leader considering a new system. The email promises efficiency. The ad shows a beautiful dashboard. The research behind both says her company is growing.

But she’s worried about adoption. The last system took months to get into the team’s routine. She needs to understand what would change, what the switch would take, and why this time would be different.

In that illustrative scenario, every deliverable can pass review while her actual concern goes unanswered.

Skipped the thinking. Solving the wrong problem. Grading the wrong thing.

The fix, in one minute

Stop prompting. Start interviewing. Find someone who has done the work well. Record how they would do it, what they would rule out, and why. Give that transcript to your AI to build the plan. Then refine it against real work.

Leave with a reusable playbook and a way to judge whether it helps the business.

Here’s exactly how I’d do it

Put the decisions back into the work

01 / ChooseOne real assignment
02 / InterviewThe expert’s decisions
03 / ExtractA usable playbook
04 / TestOne new example
05 / RefineWhat buyers show you

1. Pick one piece of work and name the business job.

Choose something you actually need to do this week. Keep the deliverable. Add the decision it needs to support.

Use in your next assignmentName the business job

01
Category research: Which buyer need could we credibly serve better than the alternatives? Leave with a positioning choice supported by evidence.
02
Outbound: Which relevant problem would give this buyer a reason to engage? Look for qualified conversations about that problem.
03
An ad: Which benefit should the buyer recognize immediately? Check recognition before spending more, then measure the relevant next action.

Write: “We need [deliverable] to help [person] understand or decide [specific thing], so [business result].”

“capture what we want to achieve with our design, not how.”

Nielsen Norman Group’s point: define the need before choosing the solution. My application here: an ad names the format; recognizing a useful benefit names its job. Source: User Need Statements ↗

2. Find the expert. Turn the microphone on. Ask why.

Find someone who has done comparable work well. That might be your researcher, your creative lead, your best salesperson, or you. Ask them to bring a real example and the evidence that it worked.

With their permission, record the conversation. Tell them your assignment. Then ask:

Use in your next assignmentInterview for the decisions behind the work

01
What would you need to understand before starting?
02
Walk me through your example. What did you notice first?
03
What did you do next, and why in that order?
04
What did you rule out? What would make you choose differently?
05
What could look great and still fail?
06
How did you know it worked, and what would you change for my situation?

Get into the greatest detail imaginable. If they say “lead with the pain,” ask how they knew which pain mattered. If they say “make it disruptive,” ask what the buyer needed to see and why that visual would communicate it.

Leave with a transcript of decisions, tradeoffs, and evidence.

Gary Klein’s Critical Decision Method examines “actual, non-routine events that require expert decision-making.”

The useful lesson: a real example gives you decisions to unpack. Ask the expert to reconstruct the work so you can hear the judgment that a generic list of best practices leaves out. Source: Critical Decision Method ↗

3. Give the transcript to your AI. Ask for the build plan.

Here’s the instruction I’d use:

Copy this with your transcript

Use this expert interview to build a playbook for my assignment.
Extract:
• The intended business result and the decision this work supports.
• The inputs and evidence needed before starting.
• The decisions, their sequence, and why they matter.
• What to avoid, when to change approach, and when to ask for help.
• Quality checks before use and outcome measures after use.
Preserve examples and source timestamps where available. Separate the expert’s advice from your suggestions. Flag missing information instead of inventing it. Return the playbook and a rewritten ask. Do not produce the deliverable yet.

Illustrative ad playbookFrom “make it compelling” to a decision you can test

01 → EvidenceWhat did buyers actually say?
Use interview evidence about adoption anxiety. Keep the source beside the brief so a plausible assumption does not become a supposed customer fact.
02 → DecisionWhich concern should the ad address?
If the evidence supports it, focus on how the team’s working day would change. A generic claim about efficiency will not explain the switch.
03 → ExecutionWhat should the buyer see?
Demonstrate the change with a product example you can substantiate. Give the production agent the approved example and the intended benefit.
04 → BoundaryWhat if the proof is missing?
Return the gap to the operator. Do not invent an implementation promise or a customer result to make the ad more convincing.
05 → ReviewWas the benefit understood?
Before increasing spend, ask target buyers what they understood without explaining the ad. Capture what still worries them.
06 → FeedbackWhat changes in the next run?
Locate the missed decision: concern, proof or next step. Update the playbook and check qualified visits or conversions against the result you intended.

The operator supplies and approves the judgment. The system applies it to the next ad and makes the result available for review.

4. Have the expert check it. Run one new assignment.

Ask the expert to correct missing steps, bad assumptions, and advice that only applied to the original example. Then give the approved playbook and a fresh assignment to the AI.

Review against criteria you set before seeing the output. Does the research support the positioning choice? Does the email give this buyer a credible reason to respond? Can someone understand the ad’s benefit without your explanation?

“The people closest to product requirements and users are best positioned to define success.”

Anthropic’s guidance supports involving domain experts in evaluation. My takeaway: the person who understands good work should help set the standard before the AI starts grading its own output. Source: Demystifying evals for AI agents ↗

5. Check the business result. Refine the playbook.

Passing review means it is ready to test. Put it in front of the people it was made for and check the result you named in step one.

For research, test whether buyers recognize and value the proposed difference. For outbound, track qualified conversations rather than counting every reply as a win. For ads, pair benefit recognition with qualified visits or conversions. Compare with your current approach where practical; a small test gives you direction, not proof of causation.

When something misses, locate the decision behind it. Was the buyer concern wrong? Was the proof weak? Did the next step ask too much? Update the playbook so the next assignment benefits from what you learned.

Make the playbook part of the system

The system at a glanceTurn the expert interview into repeatable execution

01 →Extract the decisions
Capture inputs, tradeoffs, sequence and reasons from an actual example. Flag missing information instead of filling it with invented expertise.
02 →Approve the playbook
Have the expert correct assumptions and context-specific advice. Define quality criteria before anyone sees the new output.
03 →Run a fresh assignment
Make the approved playbook and required evidence inputs to execution. Send exceptions to a named operator.
04 →Improve the recurring work
Check the business result, locate the decision behind a miss and update the playbook. Make the next run use the correction.

Check the result: research supports a positioning choice, outbound starts relevant conversations, or ads communicate a benefit buyers recognize.

The interview is how you get the judgment out of someone’s head. For recurring work, give that judgment a permanent job.

For category research, that could mean an agent gathers evidence against the questions the expert identified, flags conflicting findings, and returns a recommendation for the operator’s positioning decision. The resulting decision then informs the ad and outbound work. Each assignment contributes to the same business argument instead of starting from a fresh guess.

Start with one recurring assignment. Keep the approved playbook, test case, review criteria and owner together, and make the next run use them. You are building a way for expert judgment to keep shaping execution after the expert leaves the conversation.

Start with the ask sitting in front of you

I made Does my ask suck? to help you begin. It asks your AI to uncover the missing decisions, interview you, and help find relevant practitioner expertise when you need it. If you have a transcript, bring it.

Get the free skill → Copy the instructions or download the files. No email. No signup.

Your next move: choose one recurring assignment, find the person whose judgment you want in it, and record the walkthrough. Turn that into one playbook. Test it on one new piece of work before you scale it.

The AI didn’t need a better prompt.

It needed a better conversation.

Maddie

Adapted from my LinkedIn post on deliverables and decision making. Research excerpts and applications included in the essay. Buyer scenarios and playbook examples are illustrative.