---
title: "Act, Confirm, or Stop? Smarter behavior for AI assistants, wearables & robots"
category: "talks"
date: "2026-06-29"
time: "3:45pm-4:05pm"
track: "Voice & Realtime AI"
room: "Track 6"
speakers: ["Amit Desai"]
sourceLabels: ["Official conference schedule", "Public YouTube metadata"]
scheduleTrack: "Voice & Realtime AI"
scheduleRoom: "Track 6"
scheduleLabels: ["Voice & Realtime AI", "Track 6", "session", "confirmed"]
---
# Act, Confirm, or Stop? Smarter behavior for AI assistants, wearables & robots

## Conference Context
- Date/time: 2026-06-29 · 3:45pm-4:05pm
- Track/room: Voice & Realtime AI · Track 6
- Speaker(s): Amit Desai
- Session type/status: session · confirmed

- Track: Voice & Realtime AI
- Room: Track 6
- Session type: session
- Status: confirmed

## Session Description
Voice is our favorite way to command AI assistants and robots — and it is error-prone. The industry's reflex is to chase accuracy, but accuracy is only one knob: we can control system behavior in other ways to increase user satisfaction. This talk shifts the lens from accuracy to user outcomes. Give the AI agent more than one move: besides acting, let it stop, reject, confirm, clarify, or disambiguate. The question stops being "how often are we right?" and becomes "what does each outcome cost the user?" Bad outcomes are not equally bad to users — so price them relatively, then have the AI system minimize that user cost. Call it OUCH: Outcome User Cost Heuristic; we optimize system behavior to minimize the OUCH. Same accuracy, lower user cost, greater user adoption. We will walk through practical AI assistant examples illustrating this approach, then show how the same framework extends across AI environments — smart speakers, TVs, glasses, embodied AI, robots, wearables, and vehicles — by repricing outcomes and swapping the confirmation UI. Why this matters now: the cost of voice-command errors is escalating as we move into AI assistants and embodied AI, where wrong actions can be more expensive and dangerous. Mainstream voice adoption will not come from chasing accuracy alone; we need systems to price in the cost of being wrong.

## Media Evidence
No related AI Engineer channel video found yet.

## Evidence Graph
This evidence graph is generated from currently linked source material: official schedule text, related video pages, cached transcripts, visible slide text, dense/reconstructed slide pages, and AI slide-classification audits.

### Media Signals
No linked video, transcript, or slide source has been attached yet.

### Agent Reading Notes
Use these signals to refine the synopsis, topic links, people/company context, and method notes. If a source is a related external video rather than an exact official recording, keep it framed as supporting evidence.

## Transcript Status
No official session recording transcript was found by exact title match on the AI Engineer YouTube channel during this run.

## People
- [[amit-desai]]

## Notes
- Pending transcript synthesis when an official recording or confirmed matching video is available.

## Synthesis
### Synthesized Breakdown
# Act, Confirm, or Stop? Smarter behavior for AI assistants, wearables & robots ## Conference Context - Date/time: 2026-06-29 · 3:45pm-4:05pm - Track/room: Voice & Realtime AI · Track 6 - Speaker(s): Amit Desai - Session type/status: session · confirmed - Track: Voice & Realtime AI - Room: Track 6 - Session type: session - Status: confirmed ## Session Description Voice is our favorite way to command AI assistants and robots — and it is error-prone. The industry's reflex is to chase accuracy, but accuracy is only one knob: we can control system behavior in other ways to increase user satisfaction. This talk shifts the lens from accuracy to user outcomes.

### Speaker And Company Context
- [[amit-desai|Amit Desai]] — Director, Voice & Assistant AI at [[roku|Roku]].

### Topics Covered
- [[agent-security]]

### Derived Links And Source Material

### Novel Concepts / Clever Methods
- No highlighted novel concept has been detected yet.

### Evidence Boundary
This synthesis is based on the official schedule and linked source pages. It should be revisited when exact session recordings or transcript-backed secondary sources are available.