Haz Hubble is the Co-Founder and CEO at Pally (YC S25).
When I started building Pally, an AI assistant that keeps people on top of their messages, I assumed it needed its own app. We built one, with onboarding, settings and a home feed. Fewer than 20% of new users were still active in their second month.
We removed the app and moved the product into iMessage. Users now sign up by texting a number. With the same AI underneath, month-two retention rose to over 70%. That result made me question how much of consumer software exists only to support the app around it.
The Cost Of A New App
Every app asks people to do work before they get anything back. They have to find it, install it, create an account, learn where things are and remember that it exists. Some share of users gives up at each step.
The phone they’re installing it on is already crowded. The average user has more than 80 apps, and about 62% of them go unused in a given month. In the U.S., people spend around 80% of their app time in their three most-used apps. A new app is asking for a place that most people have stopped giving out.
Three Ways Past The Screen
There are three broad approaches.
The first is new hardware. OpenAI has said it will show its first consumer device, a screenless product designed with Jony Ive, in the second half of 2026. The last well-funded attempt, Humane’s AI Pin, failed. A common explanation is that it asked people to change their habits without giving them enough reason to.
The second is the super-app. WeChat hosts more than 4.3 million mini-programs, and roughly 764 million people used them each day in 2025. People in China order food and pay bills through services they never downloaded, because those services sit inside an app they already open all day. The model depends on one company controlling the platform, which is part of why nothing comparable exists in the U.S.
The third is putting AI inside the messaging apps people already use, such as iMessage, SMS, WhatsApp and email. Language models make this practical because a conversation can now do the job that buttons and menus used to do. The assistant can also message first, so users don’t have to remember to open anything. Of the three approaches, this is the only one that doesn’t ask people to form a new habit.
Why Now
Satya Nadella made a version of this argument about business software in December 2024. He said, “The notion that business applications exist will probably collapse” as AI agents take on more work. He reasoned that most apps are databases with rules on top, and agents can apply those rules on a user’s behalf.
The same reasoning applies to consumer products. When an assistant can read your messages, check your calendar and act on a request written in plain English, many of an app’s screens become optional. Messaging is where this shows up first, because it’s where people already spend their attention.
The Platform Risk
Building inside another company’s messaging app means that company can shut you out. Meta did exactly that in January 2026, when WhatsApp stopped allowing general-purpose AI assistants such as ChatGPT and Perplexity on its business platform. After EU regulators intervened, Meta restored access in the European Economic Area in July. As of late July, the ban had not been lifted in the U.S.
Apple’s position is different. In June 2025, a federal judge refused to dismiss the Justice Department’s antitrust case against Apple, and the allegations the court let stand include restrictions on iMessage and on super-apps. Apple has reportedly made several settlement offers to the DOJ this year. While that case is open, blocking third-party assistants from iMessage would hand regulators a fresh example of the conduct they’re already suing over.
That doesn’t resolve the risk. Apple could still change its rules, and no messaging platform has made a lasting commitment to AI developers. Anyone building this way should plan for losing a channel.
What Happens Next
1. Within five years, most consumer apps will work mainly as back-ends. Their data and logic will still matter, but people will reach them through an assistant instead of opening them.
2. A phone number will replace the App Store listing as the way many people first try a new product.
3. Screens will be generated when a task needs one, such as signing a contract or reading a chart, and will close when the task is done.
4. Consumer products will be judged by completed tasks. Time spent in an app will start to look like a cost to the user.
5. Regulators will treat AI access to messaging platforms the way they treated browser choice and app store rules, and at least one major platform will be required to open up to third-party assistants in the U.S.
What Builders Can Do Now
If you’re building a consumer product today, make three changes.
• Track how long a new user takes to get value, counted in messages rather than screens. If the first exchange doesn’t help them, fix that before anything else.
• Build the version with no interface first. Add screens only where text doesn’t work, such as payments or reviewing long documents.
• Keep your data and logic on your own infrastructure, so you can move between iMessage, SMS, WhatsApp and email if a platform changes its rules.
For Pally, this means treating iMessage as our first channel, not our only one. We keep the product’s logic on our own infrastructure so it can run over SMS, WhatsApp or email if Apple’s rules change. We add a screen only when a user can’t finish a task in text, and every new feature has to work without asking anyone to open something new.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?