-
-
Notifications
You must be signed in to change notification settings - Fork 1
Description
Problem statement
Home Assistant already supports a range of AI integrations, but getting them working requires technical knowledge, external documentation, and a significant time investment. For most users, AI features are effectively invisible. There's no guided path to get started, no contextual nudges when AI could help, and no clear entry point that doesn't require prior knowledge. We've documented the current flow in all its complexity below.
Due to the amount of dead ends and prior required knowledge, the value AI can offer remains locked behind a setup barrier that many users may simply never clear.
Fixing this directly supports OHF's approachability goal, and might also support Nabu Casa's goal of growing HA Cloud adoption by positioning HA Cloud as the obvious, zero-friction way to get started with AI.
There are already meaningful AI capabilities in Home Assistant today that many users never reach. Not because the features aren't valuable, but because the on-ramp is too steep. Shipping more AI features without fixing this first only compounds the problem: users hear about them, attempt setup, fail, and walk away with a negative impression. By fixing the foundation first, we multiply the impact of everything already built and everything that follows. Every future AI opportunity benefits from this being solid. It's not a one-time investment, it's infrastructure that pays forward indefinitely.
Community signals
- Nothing concrete that we could find, but the flow as outlined above speaks for itself. We suspect users that are really passionate about enabling AI features will find their way to it through video guides, but it's just objectively a bad experience.
Scope & Boundaries
In scope
- Improved placement of AI-related settings in the UI-
- Improved setup flow for AI integrations
- Nabu Casa/HA Cloud promoted as a zero-friction AI entry point
- Progressive disclosure of AI features once an integration is configured
- Contextual prompts to set up AI when relevant
- Correct labeling of AI integrations
Not in scope
- Changes to the underlying AI integrations themselves
- New AI capabilities (automation generation, dashboard generation, etc.)
- Forcing AI on users who don't want it
Foreseen solution
A guided onboarding flow for AI that any user can complete in under two minutes, without external documentation. HA Cloud is surfaced as the recommended entry point for users who don't want to configure their own AI backend. Once an integration is set up, AI features are progressively surfaced in relevant parts of the UI rather than requiring users to seek them out. What this might look like is included in this Figma file.
Risks & open questions
- How do we surface AI contextually without it feeling intrusive to users who aren't interested, while also avoiding adding "✨ generate ✨ with ✨ AI ✨"-buttons everywhere?
- How do we keep the setup flow maintained as the AI integration landscape evolves?
Appetite
Small - 1 cycle.
Execution issues
No response
Decision log
| Date | Decision | Outcome |
|---|---|---|
Metadata
Metadata
Assignees
Labels
Type
Projects
Status