TL;DR
- My view: voice is the next way we talk to our devices. For many tasks you will not open an app. You will tell your AI assistant, and it will use the app for you.
- AI agents reached the web first. Mobile is next: the assistant calls your app, AI runs on the phone, and agents operate apps.
- I expect an app that an assistant cannot use to be skipped. This is part 1. Part 2 shows how to get ready.
I write Code Meet AI, the newsletter where I document how I actually use AI to ship real products, receipts included. New guides land there first.
I have shipped mobile apps for 9 years. In June I wrote about what WWDC 2026 meant for apps: Apple made AI visibility mandatory and AI integration provider-agnostic. Now, in November, the first signs of it are showing.
Voice is the next way we talk to our devices
The touchscreen changed how we use almost everything. The next change is voice: you tell your AI assistant what you want, and it does the job.
Say you need a flight to Lisbon on Friday. Today you open three apps, compare prices, type your passport details again and pay. I expect you to say: "Book me the cheapest flight to Lisbon on Friday morning." The assistant picks the app, fills the form and asks you to confirm. You never open the app.
AI agents learned to use the web first
An AI agent is an AI that takes steps for you instead of only answering. On the web, it happened in steps.
Anthropic released a beta in October 2024 where its AI could look at a screen, move a cursor, click and type. That is an agent using a website the way you do. Anthropic's MCP, the Model Context Protocol, is what it calls "an open standard" for connecting AI assistants to tools and data. Add plain API calls, which programs already use to talk to services. I count three doors for agents on the web: tools they can call, services they can query, and pages they can operate.
An agent uses what it can reach. A website with none of those doors waits for a human.
Mobile is next: three signs point there
1. Assistants are starting to call apps
Siri AI is Apple's new Siri that does things inside apps for you. Apple opened a developer preview to every developer on 3 November 2026. Apple's preview page says customers get it "in a few months". Apple's rule: "Siri can only perform actions you make explicitly available".
Google's version for Android, AppFunctions is an experimental preview as of November 2026. It lets an app tell the assistant what it can do. If this pattern holds, the assistant will choose among apps, and it can only choose the ones that said what they do.
2. The phone is starting to run its own AI
Apple opened its on-device model to any app in June 2025. That is the AI that runs on the iPhone and not on a server, and Apple calls it "free of cost" and "available when users are offline". On Android, Gemini Nano, Google's small model for the phone, runs in Google's AICore service and "enables offline functionality". I read "free of cost" as no per-request bill for the app, and offline means smart features that work on a flight.
3. Agents are starting to tap through apps
Google began a beta in February 2026 where Gemini runs steps in apps for you, "in a secure, virtual window". It started with select apps in the food, grocery and rideshare categories, in the US and Korea.
Meta's September 2026 announcement of Muse describes a personal AI agent that "actually does the work". I have not checked whether it has reached your phone yet.
I predict an agent will run your booking or checkout flow. It will not care how nice your onboarding screens are. If the flow confuses it, you lose the sale.
Start with one question about your app
The direction is clear to me. The shape and the date are not.
I see this as the biggest shift since the touchscreen. So adapt early. Ask one question about your app: what does it let people do, and could an assistant find that? If you do not know, that is where to start.
Part 2 is where the homework starts: how to get your app ready, step by step.
This one came from Code Meet AI, the newsletter where I write up the setups I actually run on my own projects.



Top comments (0)