Apple CEO Pushes for Agentic Siri as AI Rivals Gain Ground

Apple CEO Pushes for Agentic Siri as AI Rivals Gain Ground

Apple’s newly appointed CEO, John Ternus, is privately urging the company’s software leadership to pick up the pace on AI, with a more capable agentic Siri at the center of that push, according to Bloomberg’s Mark Gurman. The reported internal pressure comes even as the company presents a satisfied face about the recent Siri AI launch.

Gurman reports that Ternus has encouraged Craig Federighi, Apple’s senior vice president of software engineering, to accelerate the roadmap and move beyond incremental improvements to the voice assistant. The friction is said to stem partly from Federighi’s reputation as a skeptic of AI’s rapid adoption.

What the current Siri AI can and cannot do

Article image

The revamped Siri AI is a stronger voice assistant than its predecessor, with useful features that have earned user praise. Gurman has noted that its ability to automatically fix weak passwords already looks somewhat agentic, though AI-generated passwords bring their own trade-offs.

What Siri AI still lacks is the kind of autonomy shown by agents from AI-first companies that aim to organize tasks proactively, operate a computer, or write advanced code. Those rival agents are uneven: some, such as OpenClaw and Anthropic’s Claude Code, are highly functional, while others, like Google’s Gemini Spark, are not yet there.

The gap is less about hardware than software. Apple has built powerful devices well suited to running local AI models, yet has moved more slowly than competitors on the software side. Rivals have shown they can turn promised agent features into working products, albeit at different speeds.

Why Apple's slower pace has cut both ways

Apple’s caution has, in the eyes of many observers, actually helped its reputation during a period of growing public backlash against AI. Some users are frustrated by software bloat from mediocre or unnecessary AI features, particularly as hardware prices climb to accommodate them amid a memory shortage.

Users also take comfort in Apple’s privacy-first approach, which has helped distinguish its AI work from rivals. Meta’s Muse agent may be an extreme example of a data-hungry personal agent, but privacy trade-offs are not absent from other products. Gemini, for instance, relies on more of your data to perform at its best, setting privacy risk against usability.

At the same time, Apple spent two years promoting Siri AI only to deliver an improved voice assistant rather than a groundbreaking AI product. That gap may mean users are now ready for more advanced capabilities.

Competitors are also developing their own hardware. Gurman suggests that if they pair their software with AI-native devices built around assistants that can see, remember, and act, Apple’s longstanding advantages could begin to matter less.

What an agentic Siri could look like

An agentic Siri would likely build on features users already appreciate. The on-screen awareness capability, for example, currently works only with what is visible at that moment, such as the portion of a long email you have scrolled to. A more capable agentic Siri could scroll through an entire email on its own and surface the insights you need.

In the near term, however, Ternus’s push will probably focus on delivering promises Apple has already made. Gurman points out that Siri AI still lacks advanced in-app actions, which Apple first announced in 2024 and which would let third-party apps create plug-ins and pathways through Siri. That functionality would allow Siri to tap into both in-house and third-party applications with greater precision. In the first beta of iOS 27.3, Apple began discussing the feature again.

If Siri AI became a competitive agent, many users who trust Apple’s ecosystem might stop installing apps like ChatGPT or Claude on their devices. That remains unlikely for now given the wide gap between Apple’s AI and consumer products from OpenAI and Anthropic, but it is not off the table as public AI backlash and privacy concerns grow.

Even among newer agent products, the field is immature. OpenAI’s Dot and Anthropic’s Duet were found by ZDNET’s David Gewirtz to be a little odd and not quite ready for prime time.

OpenAI’s Computer History feature, which builds a timeline of app and website activity and is available only on the Mac, raised privacy concerns when it launched in August. The feature is opt-in and can be turned off, but OpenAI said it does not encrypt sensitive information the feature sees, leaving it vulnerable to hackers. If Apple moves forward with AI, it could design more secure versions of such features in-house, leaning on its trusted commitment to privacy and user data protection. For users who value the subtlety of Siri AI, though, anything resembling ChatGPT or Claude might not land unless each feature is opt-in.

Leave a comment

Your email address will not be published. Required fields are marked *