Practical framework
Choose the right workflow. Measure the result that matters.
Use these questions before approval, during a pilot, and after deployment. Start with a business problem, then examine who benefits.
Opportunity checklist
Is this a good first AI project?
A strong first workflow is repetitive, bounded, measurable, and reviewable by someone who understands the work.
- Happens frequently
- Consumes measurable time
- Uses information you can access
- Has a clear current owner
- Produces an output a person can review
- Allows errors to be corrected
- Has a meaningful before-and-after measure
- Can be tested without disrupting the whole business
Good starting question: Which recurring task would you gladly stop doing by hand—but still want a person to supervise?
Step 1 · Baseline
Describe life before the system changes it
Measure the existing workflow in human and business terms. Without a baseline, a productivity claim cannot show who actually benefited.
- Hours and schedule predictability
- Pay and staffing levels
- Customer prices and fees
- Wait time and service capacity
- Language and accessibility barriers
- Small-business implementation cost
- Infrastructure burden
- Worker and creator decision rights
Step 2 · Distribution
Name who receives each gain
For every claimed saving or improvement, identify the recipient and the mechanism that delivers it.
GainAsk
Time savedShorter hours, more capacity, fewer jobs, or simply more assigned work?
Cost reducedLower prices, higher pay, community investment, or increased profit?
Expertise scaledWider access, better service for current users, or a new barrier for people offline?
Knowledge usedCredit, compensation, bargaining power, or no meaningful say for contributors?
Step 3 · Control
Protect voice before and after deployment
- Participation. Include affected workers, customers, contributors, and communities before decisions are final.
- Transparency. Explain what changes, what information is used, and how outcomes will be measured.
- Choice. Preserve meaningful alternatives, review rights, and the ability to appeal or opt out where appropriate.
- Correction. Set a review date and change the system when the promised benefit does not arrive.
Step 4 · The test
More money, time, access, or control?
A credible plan should show at least one concrete improvement without quietly reducing another. Record the evidence, name the people affected, and revisit the result after deployment.
MoneyPay, prices, fees, and compensationTimeHours, workload, waiting, and capacityAccessAffordability, language, disability, and reachControlVoice, choice, privacy, portability, and appeal
Important: These are possible outcomes, not automatic consequences of better AI. They depend on deliberate choices and accountable institutions.
