Apple AI is back at the centre of the conversation — and its new chief is refusing to let the hype set the agenda. In a recent interview with WIRED’s Steven Levy, Apple CEO John Ternus said the company is “absolutely going to leverage AI” in its products, while drawing a firm line: “We don’t want to ship AI for AI’s sake.”
For a company that has spent years being described as behind, that is a remarkably relaxed stance — and it reframes the entire Apple AI race: away from who can demo the biggest model, and toward who can quietly make software people love. Here is what Ternus actually said, why the timing matters, and what his answer means for Siri, for AI agents, and for everyone else building in this market.
- Inside the Wired Interview Behind the New Apple AI Message
- Siri and the Apple AI Comeback
- What It Actually Means to Ship AI for AI’s Sake
- Apple AI Agents and the Question of the End of Apps
- Apple AI Strategy Against Google, Meta, and OpenAI
- The Business Case for Apple AI
- What Apple AI Signals for the Rest of the Industry
- References
Inside the Wired Interview Behind the New Apple AI Message
A Familiar Criticism, a New Spokesperson
For the past couple of years, Apple has often been described as behind on AI — the classic narrative being that it watched rivals ship breakthroughs while it tinkered. The most visible symbol was the long-delayed, more powerful version of Siri. Speaking with Levy, Ternus — a longtime Apple executive who officially became CEO at the beginning of September — acknowledged that “everyone has an opinion” about where Apple AI stands. His response was disarmingly simple: “I really like where we’re at.”
Thirty Minutes That Set the Tone
The interview itself — WIRED’s Steven Levy spending 30 minutes with Apple’s CEO — became required reading precisely because it landed at such a strange moment for Apple AI. Momentum is genuinely improving, yet the reputation lags. Rather than promising a crash-through-the-tape sprint, Ternus framed restraint as deliberate strategy: the company is absolutely going to leverage AI, but on its own terms, in its own products, at its own pace.
Why the Timing Matters
Context is everything. In August, Apple finally shipped its rebuilt Siri. In early September, Ternus’ first memo as CEO hyped a “huge launch next week”. And across the industry, analysts openly debate whether the AI spending boom is showing cracks. Against that backdrop, a CEO publicly saying Apple AI won’t chase features for the sake of features is a statement aimed at three audiences at once: investors who fund the spend, developers who build on the platform, and the rival labs still shouting about scale.
What Ternus Is Not Saying
It would be easy to read the Apple AI comments as pure confidence. They are not. “I really like where we’re at” is also the most conservative possible framing — approval of the current position, with no promise about next year. Ternus spent thirty minutes praising products, not roadmaps. Investors heard stability. Engineers heard permission to keep shipping carefully. And the AI hype machine heard, politely, no.
Siri and the Apple AI Comeback
Years of Slipped Promises
Apple announced Apple Intelligence in June 2024, with a slate of features that included a much smarter Siri with deeper app awareness. Several of those features slipped; the revamped Siri arrived in the summer of 2026, roughly two years after the unveiling. That gap — announced versus delivered — is the single number most often cited by people who call Apple AI late. Ternus did not dispute the history. He simply declined to treat it as destiny.
A Summer Launch That Felt Anticlimactic
When the rebuilt Siri finally launched this summer, TechCrunch asked why it felt anticlimactic. The honest answer: two years of headlines raised the stakes, and the shipped improvements — however useful — landed as competence rather than magic. The lesson for Apple AI is subtle. Waiting long means the eventual release is judged against everything competitors have meanwhile made look normal.
Hardware as the Quiet Advantage
While the assistant conversation dominated headlines, Apple kept making the silicon argument. The refreshed iMac with M6 arrived promising a “huge leap” in AI performance, extending the company’s core claim: unified memory and on-device compute mean Apple AI can run genuinely useful models locally, where rivals pay per inference call. That is the part of the Apple AI story that needs no keynote to explain — and no subscription to sustain.
Why the comeback mattered goes beyond one assistant. With the relaunched Siri, Apple AI bought itself time. Competitors had spent two years putting generative features on every screen, while Apple spent its energy rebuilding foundations: a new foundation-model architecture, an in-house model tuned for the company’s own silicon, and a privacy story it could still tell with a straight face. Insiders described the summer 2026 release as deliberately un-flashy – fewer headline demos, more work on latency, reliability and developer APIs. Apple AI has always been a slow-burn product, and the relaunch let the company re-enter the race on its own timetable rather than under pressure from rivals or shareholders.
What It Actually Means to Ship AI for AI's Sake
The Feature Checklist Trap
Shipping AI for AI’s sake means bolting a chatbot onto everything because a competitor did, announcing model specs no customer asked for, and re-labelling old features until they sparkle in a keynote. It borrows urgency from the market instead of earning demand from users. Plenty of vendors have done it; plenty of those features now sit unopened inside otherwise excellent products. The trap is real, and Apple AI — to its credit — has mostly avoided it.
The Invisible AI Test
Ternus offered a crisp definition of doing it right: “When we’re doing it right, people won’t have to think about whether it was AI that made this happen or more traditional software.” Call it the invisible AI test. The value arrives as a calendar suggestion, a photo that stays still, a message that reads well — never as a badge. It is a stricter standard than the industry’s default, because invisible AI cannot take credit for itself.
Products People Love Beat Demos That Impress
“What matters is, did we build something that they really love?” The question sounds soft; in practice it is brutal. Love is measured in retention, not applause. A comparison makes the difference concrete:
| Factor | AI-first launch | Product-first launch |
|---|---|---|
| Headline goal | Show off the model | Ship the outcome |
| User trust | Fragile — novelty fades fast | Grows quietly, feature by feature |
| Support load | High — expectations unclear | Lower — value obvious at a glance |
| Churn risk | High when demos underdeliver | Lower when workflows actually fit |
| Shelf life | One press cycle | Years of daily use |
None of this explains why a company would choose the slower path — except that the slower path is the one that keeps customers. The next question is where the philosophy meets the hardest test of all: agents.
Apple AI Agents and the Question of the End of Apps
The Question Levy Pressed
Levy pushed on the scenario every platform company dreads: AI agents so capable that individual apps stop mattering, interfaces dissolve, and the phone becomes a microphone. Ternus’ reply was famously sidelong — “That could make life really easy, I suppose” — before pivoting to the defence of craft: “the creation of products is a deeply human and personal thing, something that we love to do.” Translation: Apple AI will absorb agent technology, but it will not outsource the product.
Creation Is a Deeply Human Thing
That line is easy to mock and hard to argue with. Every agent hype cycle assumes users want fewer choices, fewer screens, fewer points of contact with software. Millions of people, it turns out, want the opposite: they want to feel the quality of the tool. A workflow where an agent books everything is efficient; a life where you chose everything is yours. Apple AI’s position is that both can be true — and that the app store survives the transition.
Where Agents Actually Fit Today
Real deployment tells a practical story. Agents earn their keep inside bounded jobs: triaging support mail, watching pipelines, preparing drafts — exactly the shapes being productised as AI employees and autonomous AI agents. What they have not done is make apps vanish. On the Apple AI platform specifically, agents are showing up as features inside existing apps — precisely where the “invisible AI” doctrine says they belong.
Apple AI Strategy Against Google, Meta, and OpenAI
The Rival Playbook
Google threads AI through search, Android, and cloud in one motion, monetising attention and infrastructure at once. Meta spends audaciously on open models to own the next developer default. OpenAI sells intelligence directly — subscriptions first, enterprise second, hardware ambitions third. Each is an offensive bet that the market rewards being visibly first. Apple’s countergamble is that the market eventually rewards being visibly right. The comparison table below lays those playbooks out side by side.
| Dimension | Apple | OpenAI | |
|---|---|---|---|
| Distribution | Owned hardware + iOS | Android, Search, Workspace | Apps, API ecosystem, partners |
| Monetisation | Device premium + services | Ads + cloud compute | Subscriptions + API usage |
| Flagship assistant | Siri (rebuilt 2026) | Gemini | ChatGPT |
| Signature move | Hide AI inside outcomes | Integrate across own stack | Model releases + agent tools |
| Biggest risk | Perceived as perpetually late | Antitrust + fragmented surfaces | Inference costs outrunning revenue |
Apple’s Quiet Leverage
The company competing this way controls three things none of the others fully do: a hardware business running at software-like margins, a privacy brand that makes on-device AI credible by default, and an installed base of more than two billion devices that no rival assistant ships onto so naturally. Apple AI does not need to win the argument on the internet; it needs to win it inside products people already own.
Partnerships Are Not Surrender
Note also that restraint never meant purity. Apple licensed Google’s Gemini to power the rebuilt Siri — a deal widely reported at roughly a billion dollars a year — while keeping the assistant, the interface, and the experience unmistakably Apple’s. Buying the model and keeping the product is exactly the invisible-AI doctrine applied to corporate strategy. That combination is the strategy, not a contradiction of it: Apple will buy whatever it must and build whatever matters, and the user will still only ever see Apple.
Step back and the pattern is coherent. Apple AI is competing where it holds structural advantages — distribution, integration, trust — and declining to compete where it does not, however loud the demos get. Licensing Gemini while designing its own silicon shows the pragmatism underneath the philosophy: buy the component, keep the soul.
The rivals will keep racing benchmarks and naming the next breakthrough. Apple’s countergamble is that the race that decides the decade is the quiet one happening inside the pocket of whoever picks the phone up. That is also why the money conversation — which comes next — is not really about models at all.
The Business Case for Apple AI
Services Revenue, Not Demo Buzz
Apple AI is not sold as a subscription line-item; it is woven into hardware desirability and services stickiness. Over 1.3 billion paid subscriptions already ride on the installed base, and every feature that makes an iPhone harder to leave raises the value of that annuity. Shipping AI as infrastructure — better cameras, smarter dictation, useful Siri actions — protects revenue that a rushed AI branding exercise might put at risk instead. The spending chart below is the sanity check that restraint is not stinginess: Apple now spends more on research and development than at any point in its history, even while trailing its two biggest AI rivals.
Sticky Beats Trendy
The distribution picture explains the patience. Apple has counted more than 2.35 billion active devices worldwide, roughly 1.5 billion of them iPhones. A competitor’s feature must win attention on the open internet; an Apple AI feature arrives pre-installed in front of a billion people who chose the device first. That asymmetry lets Apple ship when something is good, not when the news cycle demands it.
The Cost of Getting It Wrong
Being late is survivable; shipping wrong is not. A hallucinated assistant answer on a rival app is a screenshot; the same failure on a billion phones is a regulatory event. Apple’s privacy positioning makes on-device, conservative AI the only version it can credibly sell — caution is not a lag, it is brand coherence under constraint.
What Apple AI Signals for the Rest of the Industry
Permission to Be Practical
When the most imitated company in technology says it will not ship AI for AI’s sake, every product team hears permission to be boring in the best way. The correction cycle has already begun: buyers are asking what the feature does, not what it runs on. Venture conversations that started with model names now start with workflows. Apple AI did not cause that shift, but it justifies it — and a useful catalogue of what practical AI actually looks like keeps growing at our AI models and tools hub.
Lessons for Product Teams
Three carry over directly. First, name the outcome, never the technique: users adopt results and ignore architectures. Second, let capability hide inside familiar surfaces — a rewritten email beats a chat window for most jobs most days. Third, treat latency, privacy, and battery as product features, because that is how customers experience them. Teams wanting a structured path from pilot to production usually start with a written digital strategy that forces exactly these decisions early.
What to Watch Next
Three tells will matter more than any keynote. Whether Siri’s usage numbers rise without promotion — invisible AI lives or dies on organic use. Whether the M-series on-device push attracts developer adoption at scale. And whether Ternus keeps the line after the next competitive flashpoint, when pressure to shout will be maximal. If he holds, the most consequential AI strategy of this cycle may turn out to be the one that refused to announce itself — and Apple AI’s restraint becomes the industry’s permission slip.
For product teams outside Apple, the practical lesson is uncomfortable but clear. The companies winning with AI are not the ones with the most features labelled AI; they are the ones whose users stopped noticing the technology months ago.
That requires the same discipline Apple describes: knowing which problems your customers actually have, refusing to ship demos, and treating reliability as a feature rather than a nice-to-have. Whether you build for two billion devices or two hundred users, the bar John Ternus has publicly set for Apple AI is now the bar for everyone else too – and the teams that clear it will be the ones still shipping in 2027.
References
I Spent 30 Minutes with Apple’s CEO. Now I Know Why John Ternus Got the Job — WIRED
Apple CEO says the company doesn’t ‘want to ship AI for AI’s sake’ — TechCrunch
Apple finally fixed Siri. So why does it feel anticlimactic? — TechCrunch
John Ternus hypes ‘huge launch next week’ in first memo as Apple CEO — TechCrunch
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.
