The Four I’s: How to Read
Any Company’s AI Bet

Dorsey's Framework

In March 2026, Jack Dorsey co-authored an essay with Roelof Botha at Sequoia titled "From Hierarchy to Intelligence." Dorsey, who co-founded Twitter and now runs Block (formerly Square), laid out a four-layer architecture for how companies should build with AI. The essay uses developer-heavy vocabulary for each layer: Capabilities, World Model, Intelligence Layer, and Interfaces.

Dorsey argued that AI allows companies to run with fewer managers and more individual contributors. The essay reads as his explanation for why he cut 4,000 Block jobs (nearly half the company) a month earlier.

The framework itself is what's useful for any reader trying to evaluate a company's AI bet. I applied it to Apple. The Four I's are my plain-English translation of Dorsey's four layers: Instrument, Information, Intelligence, and Implementation.

Instrument

The Instrument layer is the first building block. It's the tools a company uses to get work done. Dorsey clarified that these are not products themselves. They are building blocks that get the product to perform. Without the Instrument layer, there's nothing to build with.

Apple earns a decisive A+ on Instrument because it owns every layer from chip to app. No competitor has come close to Apple Silicon since it launched in 2020.

Apple designs dozens of these tools, plus the silicon chip that runs them all, in-house. Think Maps, Messages, Music, Photos, Siri, and Wallet. They all come built into iOS and run on more than a billion active iPhones. Underneath sits Apple Silicon, the chip that powers everything.

Apple's Instrument layer also includes Writing Tools, Image Playground, and the Private Cloud Compute system. This is how Apple personalizes voice recognition, image tools, and language features across every app on your iPhone. Even the apps they didn't build.

Every company's Instrument layer will look different on the surface, but identical underneath. It includes the AI models your company chooses, the agents your team scopes, and all the automations that run in the background. If you can name the AI your company runs on, you have your Instrument layer. If you can't, that's the first thing to build.

Information

Information is everything a company knows about itself, written in a format AI can easily read. This spans a company's operations, data, and decisions. Information lives in files instead of in managers' heads. Information splits two ways: how a company understands its own operations, and how it understands its customers.

Apple earns a B on Information because its internal knowledge is siloed team by team, and what it knows about its customers can't connect across devices.

Mark Gurman, the Apple beat reporter at Bloomberg, has documented a pattern: Apple teams can't share Information with each other. For example, its Face ID and cancelled self-driving car project both used computer vision. Yet they weren't allowed to talk to each other. This is a cultural pattern that predates AI. Apple has long prioritized secrecy and security over Information exchange. Despite some of its product delays and cancelled projects, the long-term trade-off has proved to be a wise choice.

That same instinct shapes how Apple treats customer data. Apple knows that in order to reach customers at scale, it must keep personal data encrypted. On every iPhone, Apple Watch, iPad, and Mac sits a personal-context graph: Calendar, Contacts, Health, Maps, Messages, and Photos, all behind a privacy wall no competitor has matched. AI processing happens on each device using Apple Silicon, not in the cloud. But when cloud services are needed, Apple's Private Cloud Compute system encrypts the data so that Apple can't read personal information. Everyone values privacy, so when excellent products meet those expectations, the brand earns trust.

But Apple made a deliberate trade-off: the same decision that earns consumer trust is what blocks cross-device Information flow. It's the same bet Apple makes on Instrument: own the building blocks on every device. The Information layer pays the cost. And Apple is fine with a B here because it's already earning an A+ on Instrument.

The Information layer is everything your company knows about its own operations (docs, blueprints, decisions) and its customers (data). What prevents AI from being effective is rarely the Instrument. It's usually the gap between what everyone at the company knows and what AI is privy to.

Dorsey traces this back to org structure in the same essay: "hierarchy and middle management impede information flow." A company whose knowledge lives in files needs fewer people relaying it between heads. In practice, that looks like fewer middle layers, less waiting, and faster decisions.

Intelligence

Intelligence is the layer where a company's tools and its knowledge combine into finished work: a report, decision, or action.

Apple earns a B here because its Intelligence doesn't actually run on its own technology: Apple Foundation Models runs on Google's AI model, Gemini.

Apple originally ran Intelligence on its own technology. Clean Up in Photos, Genmoji, and Visual Intelligence all shipped on Apple Foundation Models. Then Apple made a rare public admission: "After careful evaluation, we determined that Google's technology provides the most capable foundation for Apple Foundation Models." In plain English: Apple pivoted to start renting its Intelligence layer from Google. The rental pivot is one signal. The last thirteen months at Apple have been another.

In March 2025, Apple delayed Siri's personalization features and appointed a new Siri lead. By December, Apple had poached its new VP of AI from Google Gemini. In March 2026, Apple delayed Siri a second time, pushing personalization to iOS 27. Apple's longtime head of AI left the company a month later. A company confident in its Intelligence layer doesn't spend thirteen months reshuffling around it.

B is an honest grade for a $4T tech company that now rents its Intelligence from Google.

Intelligence is the layer where your tools and knowledge turn into finished work: a report, decision, action, or autonomous agent. If your AI produces output that's specific to your company, you have an Intelligence layer. If every prompt produces a generic answer, there's probably an Information gap: AI can't act on knowledge it doesn't have.

Implementation

Implementation is how your customers use a product. Think hardware, software, interfaces, retail.

Apple earns a resounding A+ on Implementation because no other tech company owns every layer of its product, from the chip to the storefront.

Apple ships hundreds of millions of devices a year across five product lines: iPhone, Apple Watch, iPad, Mac, and Vision Pro. Each device runs Apple's own operating system, on Apple's own silicon. Over 500 Apple Stores worldwide handle in-person sales, setup, and Genius Bar tech support. No one else comes close to silicon, software, devices, and retail all in one stack.

Once AI models all start to look alike, the company that controls the devices will win the decade. OpenAI doesn't sell iPhones. Anthropic doesn't sell Apple Watches. Local open-source AI models still run on MacBooks. So Apple owns the devices, the operating systems, and the surface where everyone meets AI.

The A+ here is Apple's priority bet: own the surface where customers meet the product. Apple is not prioritizing the middle layers right now. It will catch up on its own clock.

Implementation is whatever product your customers use. If AI shows up in the middle of your operations but never in your customers' hands, you're missing an opportunity.

Synthesis

Apple's bet, read across the Four I's: internal tools and external products are the play. Apple believes its middle layers can be rented, delayed, or left to catch up later. A+ on the bookends (Instrument and Implementation), B in the middle (Information and Intelligence). This shape is deliberate. Apple's not worried about renting Gemini to power Siri or delaying personalization features. Its silicon chips and handheld devices are the moat.

The Four I's framework reads any company. Apple made its bets. Every company places its own bets, whether they know it or not.