Product · active
Three worked illustrations examine different AI tasks: comparing supplier quotes, drafting a customer reply and organizing questions about a payment request. Each shows a useful output, the facts that still need checking and the authority that remains with a person.
Apply for business coaching Read the manual it comes from
What it is
This playbook helps you define an AI task within a business decision: what information it can use, what output would help and what it must not decide.
The examples distinguish comparison, drafting and payment authority. They do not establish a single rule for every business or use of AI.
The manual's last section names the line and points across it. This product is what lives on the far side: the detailed review on where the model adds signal and where it adds fluent noise that the business mistakes for signal.
The examples make the checks visible. They do not establish that one rule fits every business or prove a measured outcome.
These are illustrative situations, not reports of client engagements. The model’s role depends on the task, information and permission to use it.
Quote A is $8,000 plus delivery. Quote B is $8,600 delivered. AI can line up the stated scope and identify the missing delivery cost. The owner asks A for the delivered price before choosing. The useful output is the missing question; declaring A the winner would overstate the evidence.
A customer asks for delivery next Friday. The schedule shows Monday, and no one has approved overtime. AI may draft: “Our current schedule is Monday. I’m checking whether Friday is possible and will confirm by 3pm.” A person verifies capacity and sends the reply. The draft must not promise Friday or imply the check has happened.
An email asks to change a supplier’s bank details. A fluent summary does not verify the sender. Use the company’s established verification process and a previously trusted contact route. AI may organize non-sensitive questions, but it must not authenticate the request or authorize the payment. The outcome remains open until the authorized person completes the checks.
Who it is for
If you run the manual's process well and now reach for a model on bigger calls, this is the guide to where that instinct helps and where it quietly costs you. It assumes you can already get clean output. The question it answers is whether the output belongs in the business at all.
This playbook is for the operator who already uses AI and risks trusting it on the one call where the model should have stayed out.
Write what AI is allowed to produce, what source material it may use, who checks it and who can act. If verification is unavailable, keep the decision open rather than treating a polished draft as approval.
What this is not
Knowing that a decision sits in the keep-it-out zone tells you to stop reaching for the model. It does not make the decision. The hardest calls, the ones the playbook flags as model-out, are exactly the ones where business coaching earns its cost. The manual names that line; the coaching is where it gets crossed.
Back to the manual →When the work is live
Stan reviews each request for fit. All corporate work starts with an inquiry and is quoted separately. Stan confirms the people, stakes, scope, and fee in writing before work starts. The page identifies where AI belongs; an engagement reviews the live decision that remains.
Apply for business coachingWork with me · $1,500/month 1:1 business work · Request scope
