A practical field guide

Build a real one-person business with AI.

Go from a problem worth solving to a repeatable operation — without pretending AI can replace customer judgment, hard conversations, or responsibility.

Start with Step 1 See how to divide the work

For nontechnical and lightly technical professionals ready to turn useful experience into a focused business.

The route

  1. 01FindWho has a painful problem worth solving?
  2. 02ValidateWill someone act, not just agree?
  3. 03BuildWhat is the smallest useful offer?
  4. 04SellWhy should this buyer choose this offer now?
  5. 05DeliverHow will the customer get the promised result?
  6. 06OperateWhat proven work should become a supervised system?
Find your next move

Six outcomes. One business.

Start at the earliest stage that is still uncertain. Each step ends with evidence — not a vague feeling that you are ready.

01

Find

Choose a specific person and a costly, frequent problem before choosing a tool.

Move on when: You can name the customer, the painful moment, the current workaround, and why change matters now.
02

Validate

Test demand with conversations, commitments, and small experiments before building the full offer.

Move on when: At least one qualified person commits time, access, a referral, a pilot, or payment to continue.
03

Build

Turn the validated problem into the simplest deliverable that creates a meaningful result.

Move on when: A real prospect can understand the offer and receive a useful result from the current version.
04

Sell

Position, price, reach prospects, and ask for the first sale without hiding behind more building.

Move on when: A qualified buyer pays, signs a pilot, or gives a specific objection you can test next.
05

Deliver

Onboard, serve, support, and improve the outcome before trying to make the work autonomous.

Move on when: The customer receives the promised outcome and you can explain what worked, failed, and changed.
06

Operate

Make repeatable work easier with AI while keeping permissions, evidence, and stop conditions visible.

Move on when: The workflow repeats with visible proof, bounded cost, clear escalation, and a tested recovery path.
See the journey in practice

A product is where the lessons get real.

Shepherds Inbox is one example of moving from a specific problem to a delivered product, then improving the work with supervised AI.

The useful question is not “How much can AI do?” It is “What helps create a better customer outcome without giving up judgment?”

See the method applied
Start with
A specific person, a recurring problem, and a result worth delivering.
Use AI for
Research support, drafts, organization, routine checks, and repeatable low-risk work.
Keep human
Customer conversations, promises, pricing, quality decisions, and consequential approvals.
Automate when
The workflow is understood, the quality bar is visible, and exceptions have somewhere to go.
One guide for each step

Start with the six practical guides

Choose the guide for the decision in front of you. If you are not sure where to begin, start with Find.

Browse all field notes
Find

Find a Painful Problem Before You Choose What to Build

A practical way to turn domain knowledge into one specific customer-and-problem hypothesis that is ready for real conversations.

Read guide Validate

Validate Demand With a Behavioral Commitment

Replace compliments and AI opinions with an evidence ledger built from last-instance interviews and a concrete commitment test.

Read guide Build

Build the Smallest Useful Offer—Software Optional

Turn behavioral evidence into a clear, deliverable offer that may be manual, presold, or service-led before any software exists.

Read guide Sell

Sell the First Offer With a Direct, Human Ask

Turn a deliverable offer into a short prospect list, relevant outreach, a clear price, and an explicit next-step request.

Read guide Deliver

Deliver the First Customer Outcome Without Losing the Promise

Map intake, checkpoints, handoffs, quality, support, and recovery so the first buyer receives the result that was actually sold.

Read guide Operate

Operate a Proven Workflow With Bounded AI Authority

Turn a repeated, successful business workflow into a supervised SOP with permissions, receipts, cost bounds, stop conditions, escalation, and recovery.

Read guide