# CS-42: Human approval for AI actions

Category: AI & emerging interfaces
Level: Advanced
Suggested duration: 4–6 weeks

Original UXCharan practice scenario. Validate assumptions; do not present it as a commissioned client project.

## Problem
A fictional assistant can draft a customer message and schedule a task. Explore how a person reviews the exact action, understands its scope and approves or cancels before anything happens.

## People to learn from
Professionals using an assistant to draft external actions

## Core scope
- Inspect a proposed action
- edit it
- approve or stop

## Research questions
- Which actions need explicit approval?
- What information must appear at confirmation?

## Edge cases
- Action fails
- scope changes
- duplicate request

## Specific artefact
An approval state machine and action log

## Evaluation question
Can users predict what will happen before authorising it?

## Work to share
Problem and role; anonymised evidence; alternatives and rationale; task flow; prototype; accessibility considerations; testing findings; revisions and honest limitations.

## Request feedback
Share a viewable Figma, document or portfolio link with Saicharan on WhatsApp: https://wa.me/917093809350
Include CS-42, your name, project stage and your main feedback question. Confirm review scope, timing and any fee directly.

## Methods
- Research planning: https://www.gov.uk/service-manual/user-research/plan-user-research-for-your-service
- Research consent: https://www.gov.uk/service-manual/user-research/getting-users-consent-for-research
- Design exploration: https://www.designcouncil.org.uk/resources/the-double-diamond/
- Usability testing: https://www.nngroup.com/articles/usability-testing-101/
- Accessible forms: https://www.w3.org/WAI/tutorials/forms/
- Portfolio storytelling: https://www.nngroup.com/articles/ux-design-portfolios/