AI Workflows
For a solo consultant, automation has to save more time than it costs to monitor. That favors small steps around existing work—not a complicated autonomous system that becomes another thing to maintain.
1. Turn rough notes into an agenda
Provide sanitized bullets from the previous meeting and ask for an agenda grouped by decisions, questions, and updates. Confirm every item before sending.
2. Extract action items from approved notes
Ask for task, owner, due date, and source sentence. Missing information should be marked, never guessed.
3. Reformat a reusable checklist
Convert your own process notes into a checklist for a specific project type. Keep the master process outside the AI tool and review omissions.
4. Prepare interview questions
Give the objective, audience role, known constraints, and questions already answered. Ask for gaps and follow-ups rather than a generic list.
5. Classify low-risk feedback
Group anonymized feedback into themes and attach each theme to the original item numbers. Read the source before making a conclusion.
6. Create a first-pass document outline
Ask for structure only, then supply evidence and write conclusions yourself. This is useful for proposals, reports, workshop notes, and guides.
7. Check a draft for ambiguity
Have the tool list unclear pronouns, undefined terms, unsupported claims, vague deadlines, and sentences longer than a chosen limit. Accept edits selectively.
8. Produce alternate explanations
Request a plain-language version, a technical version, and a 50-word summary of material you wrote. Check that the meaning did not change.
9. Generate synthetic test data
Create fictional names, dates, orders, or support cases to test a template without using real client records. Label the data clearly so it cannot be mistaken for evidence.
10. Draft a weekly review
Provide a sanitized list of completed work, open tasks, and blocked items. Ask for a review grouped into results, risks, and next decisions. You decide what becomes a commitment.
The four-gate rule
Before keeping any automation, ask:
- Data: Is the input permitted and minimized?
- Decision: Can a person still approve consequential output?
- Detection: Will an error be visible before it causes harm?
- Exit: Can the work continue and the data be exported if the tool disappears?
Start with one week, not a transformation
Choose one task you perform at least twice a week. Record the old time, the new time including review, and every correction. Keep the automation only if it produces a clear net benefit without weakening confidentiality or accountability.
NIST describes its AI Risk Management Framework as voluntary and use-case agnostic. For a small consultancy, that reinforces a practical lesson: assess the actual task and context, not “AI” as one undifferentiated risk.
Published September 4, 2026. These are tool-agnostic patterns; implementation details and provider terms vary.