Augment

Designing AI Agents: Control, Transparency, and Handoffs

An agent that acts on a user's behalf needs more than a chat box. It needs a visible plan, clear permissions, and an easy way to take over.

A brain made of circuit lines over blue and orange blocks
Photo by Steve A Johnson on Unsplash, with an Augment halftone treatment
Executive Summary

Key Takeaways

  • Agents take actions, so the main design question is control: what may it do alone, and what must it ask about?
  • Show a plan before the work starts and a running log while it runs.
  • Design the handoff both ways: from agent to human when it is stuck, and from human to agent when the user delegates.
  • Permissions should grow with trust. Start narrow and let users widen them.

Why agents change the design problem

A chatbot answers. An agent acts. It sends the email, changes the record, books the slot. That moves the risk from "the answer was wrong" to "something happened that I did not want".

NN/g describes generative AI as a shift to intent-based outcome specification: the user states the result they want and the system decides how to get there. That reverses who is in control. Good agent design gives some of that control back, on purpose.

Show the plan before the work

Before an agent starts a multi-step task, show what it intends to do in plain steps. Let the user edit, remove, or approve them. This single pattern removes most surprises.

1

State the goal

Repeat the request in the agent's own words so misunderstandings surface early.

2

Propose the steps

List the actions in order, with the tools and data each one touches.

3

Ask where needed

Mark the steps that need approval, such as anything irreversible or external.

4

Run and report

Show progress as it happens and summarize what changed at the end.

Permissions that grow with trust

Do not ask users to grant broad access on day one. Start with read-only or draft-only behavior, and let people widen the scope as they see the agent behave.

LevelWhat the agent may doGood for
SuggestPropose actions, never run themFirst week of use
DraftPrepare the work, wait for approvalCustomer-facing changes
Act with reviewRun, then show a summary to undoRoutine, reversible tasks
ActRun without askingLow-risk, well-understood tasks

Design the handoff in both directions

Agents get stuck. When they do, they should say so, say why, and pass the work to the person with the context needed to continue. Do not make the user start over.

The reverse matters as much. Users should be able to step in mid-task, edit what the agent has done, and hand control back. Treat the work in progress as a shared document, not a black box.

Keep a log users can read

A readable activity log is the agent's receipt. It lists each action, when it ran, what it touched, and a way to undo it. It also helps your team debug and builds the trust that lets users delegate more.

Frequently Asked Questions

An agent takes actions, so control, permissions, plans, and undo matter more than conversation quality.

Sources and further reading

  1. AI: First New UI Paradigm in 60 YearsNielsen Norman Group · Published · Updated · Accessed
  2. Generative UI and Outcome-Oriented DesignNielsen Norman Group · Published · Updated · Accessed
  3. People + AI GuidebookGoogle PAIR · Published · Updated · Accessed

Credits

  • Written by Augment Editorial Team, Product design and engineering, Augment Lab.
  • Cover photo Photo by Steve A Johnson on Unsplash, with an Augment halftone treatment, used under the Unsplash License.
  • Research Statistics and standards are cited from the sources above. Always check the original page for the latest version.

Augment Product Design

Ready to make your product clearer?

From first concept to go-live, we turn complex product ideas into clear, useful experiences.

Designing AI Agents: Control, Transparency, and Handoffs | Augment. Lab