High agency with clear decision rights

High agency means you notice a problem, propose options, act within the authority you have, and follow through. It does not mean unpaid extra work, constant availability, or ignoring a client’s boundaries. The most practical tool is a short decision-rights note that you agree with each client: what you can do alone, what you do and report, and what you must ask first.

Why this matters now

AI tools make it easy to do more work faster. That raises a question: who decides what the AI does and what you do? Microsoft’s 2026 Work Trend Index (May 5, 2026) surveyed 20,000 AI users in ten markets. It describes a group called “Frontier Professionals,” defined by advanced agent use, workflow redesign, and team AI habits. Compared with others, they were more likely to choose on purpose which tasks go to AI (53% against 33%). They were also more likely to do some work without AI (43% against 30%).

The report also names a “Blocked Agency” group (10% of the readiness analysis) where workers show initiative, but their setting holds them back. These are self-reported findings from AI users in other countries. Microsoft is a vendor. Still, the pattern fits a useful point: agency is about choosing, not about doing more.

A theoretical paper, Delegation and Verification Under AI (Huang, Xiao, and Vishnoi, revised June 2026), gives a related reason. In its model, how reliably a worker verifies AI output strongly changes the quality of the result. Good judgment about what to delegate to AI, and what to check, matters. This is a model, not a field study.

The decision-rights note

Use three levels. Agree on them in writing with each client.

Start with a short draft. Ask the client to move items between levels. Review the note every month.

The four moves, with a hypothetical case

This case is made up. A virtual assistant manages a client’s webinar sign-ups. The assistant uses AI to draft reminder emails.

1. Notice. The sign-up list has 12 addresses that look mistyped. Reminders will bounce.
2. Propose options. The assistant writes: “Option A: I fix the 8 obvious typos and mark 4 unclear ones. Option B: I email all 12 to confirm. I suggest A, then B for the 4.”
3. Act within authority. Fixing typos in the list is “act and tell.” The assistant does it. Emailing the sign-ups is “ask first,” so the assistant waits for a yes.
4. Follow through. After the webinar, the assistant reports how many reminders were delivered and what to change next time.

The assistant did not work extra hours. It did not guess. The client saw a problem solved with a clear record.

Where AI fits in the note

Add a line for AI use. For example: “I may use AI to draft first versions of reminder emails. I review every draft. I do not enter customer personal data into AI tools.” Check the client’s rules first. If a client does not allow AI, follow that rule.

Boundaries are part of agency

• Say your working hours and response times. Put them in the note.
• Do not accept unclear extra tasks as a favor. Ask for a scope change or a separate price.
• If a request is outside your skill or your authority, say so and propose another path.

For clients in other time zones, state “I will reply within X hours in your time zone” and list the overlap hours. This applies to many Philippine-based professionals, but any remote worker can use it.

Tradeoffs

• Some clients want to decide everything. Respect that. Make your proposals short.
• Too many “ask first” items slow the work. Move low-risk items down a level after the client sees good results.
• The research above does not prove that high agency gives higher pay or job security. It shows how some AI users describe their habits.

Do this week

Write a decision-rights note for one client. Put five tasks in the three levels. Send it and ask: “Is anything in the wrong level?”

Sources

• 2026 Work Trend Index Annual Report, Microsoft WorkLab, May 5, 2026.
• Delegation and Verification Under AI, Huang, Xiao, and Vishnoi, arXiv:2603.02961, March 2026 (v2 June 25, 2026).