I worked with Steve Jobs at Apple, where every OS update killed startups. AI founders are about to face the same thing

I saw it firsthand when I worked on the iPhone, iPod, and iPad under Steve Jobs. Every product launch and OS upgrade generated excitement for users and existential fear for founders. Founding teams spent years building capabilities that Apple could absorb into the operating system overnight. Life’s work became dead on arrival.

Sherlocked, but Not Forgotten

There are several companies that folded worth mentioning, but here are three that stand out to me:

Tile kept pace with AirTag for a while because even though Apple made a slightly nicer tracker, Tile had years of market leadership, retail distribution, meaningful hardware revenue, and a defensible head start. But the balance did eventually tip toward Apple when they launched AirTag with deep integration into the Find My network and the U1 chip. Suddenly, Tile no longer had access to the same system-level advantages. It lost access to the oxygen that mattered: defaults, permissions, hardware integration, and distribution. The company was eventually acquired by Life360 in 2021 for approximately $205 million — a fraction of its peak valuation. 

Pebble invented the modern smartwatch category years before the Apple Watch hit shelves. The company built a passionate developer ecosystem and sold millions of devices. But Apple reserved the deepest iPhone integrations — notifications, payments, health data, system hooks — for itself. Pebble wasn’t outcompeted feature-by-feature. It was boxed out structurally.

Even f.lux, which pioneered blue-light reduction software to help us sleep better at night, learned the same lesson. Apple initially rejected its iOS implementation for using private APIs. It wasn’t until Apple launched Night Shift directly inside iOS itself that f.lux experienced existential competition.

Other tech giants, like Google with search and Microsoft with Office, also shuttered numerous companies with authority and efficiency. But they weren’t destroying startups simply because they built better products. They entered the market with viable alternatives, consistently improved those products, and then maintained control over the platforms.

The key thing for founders facing similarly harrowing dynamics to remember: when a platform decides to compete, it’s impossible to win with price alone. Survival requires understanding how platforms collapse distribution, bundle features into defaults, and remove the dependencies third parties rely on. 

Survived, Thrived, and Still Alive

Recent history also provides examples of companies that survived platform attacks by evolving beyond standalone consumer features.

Dropbox should have disappeared the moment Apple and Google bundled cloud storage directly into their operating systems. Instead, it became a multi-billion-dollar enterprise software company because it expanded beyond consumer sync into collaboration, team workflows, e-signatures, and cross-platform infrastructure.

Spotify survived Apple Music despite Apple owning the operating system, the App Store, the hardware ecosystem, and the distribution advantage. In addition to investing in brand and artist partnerships, Spotify built network effects around playlists, discovery, creators, podcasts, and social behavior that could not simply be copied into existence overnight. Its value came from the ecosystem surrounding the platform, not merely the app itself.

1Password faced extinction once Apple and Google bundled password management directly into their platforms for free. Instead of competing feature-for-feature at the consumer layer, it moved upmarket into enterprise identity management, developer tooling, secrets infrastructure, and organizational workflows. The consumer feature became the wedge. The enterprise system became the business.

As we saw with Dropbox and 1Password in the last cycle, offerings from smaller companies that deeply integrate with customer architecture and offer tailored features can become the wedge in enterprise AI.

AI Is Firmly in Its Sherlocking Era. Be Aware.

Every new Claude release, ChatGPT capability expansion, or workflow agent launch creates excitement among users and customers. It should also unsettle founders.

Products historically most vulnerable to Sherlocking shared a common trait: they were single-purpose features built on platforms they did not own — small enough to bundle, and lacking network effects, alternative distribution, or deep operational integration.

AI-native companies need to operate as more than model wrappers or generalized copilots. To compete with foundation models in any given vertical, startups must become operationally embedded inside enterprises, law firms, financial institutions, and medical facilities. 

The best enterprise AI companies will integrate deeply into internal operations spanning approvals, compliance systems, procurement flows, analytics pipelines, reporting structures, and institutional knowledge. Once that happens, ripping them out becomes painful.

This matters because the frontier labs are optimized for horizontal scale, not deep operational integration. OpenAI, Anthropic, and Google can build extraordinary foundation models. But they cannot realistically provide white-glove implementation and workflow redesign for every logistics provider, hospital system, insurer, law firm, or manufacturer in the world.

That asymmetry creates enormous opportunities for startups facing fight-or-flight moments. Margins and automation capture attention and investment. But customer integration and high-touch service make up the moat.

A Parting Word to My Fellow Founders

The next generation of great AI companies can’t beat hyperscalers and tech giants with endless budgets on price. They definitely can’t win by competing head-on with OpenAI or Anthropic on generalized intelligence. However, it’s possible to thrive — and grow — if you can accomplish what giant platforms historically struggle to do: become indispensable to the operation of a customer’s business.

We’re in the early days of AI-for-everything. As in previous cycles, scores of young companies will be Sherlocked. The model I’ve used to build and scale Nest and now Mill — going deep on vertical integration — works. Founders interested in longevity should build and forward-deploy teams vertically around specific offerings or products. Hardware, software, and product design should all work on something together. The hyperscalers are delivering world-class innovation on a nearly daily basis. But they’re also clunky and siloed. If you want to survive and grow in the face of fierce competition, make sure you never join that group.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.