Your First AI Workflow

By Aaron Sneed, Founder and CEO of DOES | Published and updated November 11, 2026. A practical method for selecting a first AI workflow and keeping decision authority with people.

An AI pilot is useful when it improves a defined task and leaves the result reviewable. The AI Council approach brings multiple perspectives to research, drafting and challenge; people remain responsible for consequential judgments, approvals and action. This guide describes the method, not a certified system or a promise of performance.

1. Frame a bounded task and baseline

Identify the user, task owner, current process, approved inputs and expected output. Choose a recurring task whose result a person can inspect. Record current effort, elapsed time, errors or rework, and review burden where meaningful. Define success before testing; do not select only favorable examples afterward.

2. Define information and decision rights

Determine what information may enter the tool and who may authorize its use. Use fictional or approved non-sensitive material for initial tests. Do not enter controlled, confidential or personal sensitive information without an approved environment and handling process. Separate permission to draft from authority to approve, send, spend, release or operate.

3. Gather evidence and challenge the draft

Ask for sources, assumptions and uncertainty. Verify important claims against original records. Use different review perspectives to test feasibility, quality, risk, ethics and consequences. Multiple AI outputs are not independent evidence, and agreement between models does not establish truth. Retain contradictions and unresolved questions for human review.

4. Review, escalate and stop when needed

Assign a qualified human reviewer and define what must be checked before use. Escalate missing evidence, contradictory sources, unauthorized data, unsafe outputs or decisions outside the agreed scope. Stop or narrow the task when a reliable answer cannot be supported. AI does not replace technical authorization, professional judgment, customer acceptance or required signoff.

5. Keep a decision record and evaluate

Record the question, relevant sources and dates, tool/model where known, assumptions, alternatives, review findings, decision owner and limitations. Compare the pilot with the baseline using the same measures and representative tasks. Include error/rework and review effort, not just drafting speed. Expand only when the evidence and information handling support it; revisit after tool or workflow changes.

An illustrative workflow

A team could test drafting an action list from fictional meeting notes. The reviewer checks owners, dates, omissions, invented commitments and wording before any real use. The team compares drafting plus review effort with its baseline. This is an illustration, not a DOES customer result, validated time saving or an invitation to upload sensitive notes.

Aaron’s public discussion of the AI Council includes a Business Insider interview; its reported savings describe his own experience. For your organization, define and measure the result of a bounded assessment or pilot. Explore AI adoption consulting or browse all resources.

Discuss a scoped engagement | info@doesdoesit.com. Start with a non-sensitive overview; scope, schedule, pricing and acceptance are agreed before work begins.