Apple's Ternus Pushes Federighi Beyond a Working Siri

John Ternus spoke on Apple's Nov. 2 earnings call, his first as chief executive. Ternus succeeded Tim Cook on Sept. 1. He praised Siri AI as “the kind of Siri that we always wanted.” Siri, he said, handles “more than 2.5 billion requests every day.”

In a closed-door meeting in recent weeks, Ternus pressed software chief Craig Federighi to move faster and pursue AI ambitions beyond improving Siri, Bloomberg's Mark Gurman reported in his newsletter.

Apple's goal, as described in the newsletter, is an assistant that pursues goals between conversations, as ChatGPT Dots, Meta's Muse and Grok can. Apple's assistant currently carries out requests initiated by its users, with more detailed control over apps still being added.

The meeting account comes from one newsletter citing anonymous sources, and Apple has not publicly confirmed it. The newsletter author's forecast of an agent-like Siri as early as next year carries no Apple commitment.

What Siri does now

The Siri released this fall can assemble trip details from a user's messages and emails, then put those details into a message to someone else. It can draw on information displayed on the screen and private material stored on the device. Federighi's team rebuilt the assistant's engine, using AI models based on technology licensed from Google.

More precise app actions arrived in the iOS 27.3 beta last week. Users can organize Safari tabs or sort folders in Files. In Photos, a request can have Siri rotate and improve a picture before sharing it. Support for Apple's own apps could reach the public with the December release.

Third-party app actions remain a “preview,” with wider availability likely tied to iOS 27.5 or 27.6 before next June's developer conference. An app maker could enable restaurant bookings or let Siri pass information between apps to complete a request.

An agent, as used here, can take on an ongoing responsibility. ChatGPT Dots, Muse and Grok can remember projects and preferences, monitor connected sources for changes and decide when to pause or resume work. They can bring information to the user between conversations without waiting for each step to be requested.

Apple also lacks a computer-use tool. ChatGPT and Claude offer software that reads a screen, moves a cursor, clicks buttons and types into applications. Apple controls the platforms and devices on which hundreds of millions of people manage their lives, yet has no AI tool that can operate those systems on a user's behalf. Without one, that job falls to third-party companies controlling a user's machine, which carries security and privacy implications.

The software chief

Federighi has led Apple's software organization for more than a decade. He became its AI czar after John Giannandrea's removal last year, taking over a Siri effort that had failed to deliver. He helped get a working assistant to market in a relatively short time, despite technical and organizational problems.

His reputation as an AI skeptic includes concerns about security and privacy, alongside the technology's effects on society and employment. In June, he described his objections to chatbot design in concrete terms.

“If you use many of the existing chatbots, they're really focused on engagement to a large degree, and sycophancy, right? They kind of want to pull you in,” Federighi said. “They might encourage you to reveal things about yourself and then use that as a basis to establish a connection.”

He described Siri as a tool there to help. On treating it as a romantic partner, he said, “Siri's not into that.”

Inside Apple, Federighi takes a hands-on role in meetings and major decisions. He argues against proposals he disagrees with in technical detail. Some engineering and design employees feel that leaves them limited influence over product direction.

The planning calendars differ, too. His software planning typically runs in windows of 12 to 18 months. Hardware teams use multiyear road maps. That mismatch has contributed to delays and less ambitious initial versions of devices.

Questions about agents

Ternus described Siri as “primarily request-based today” on the earnings call. Its “orchestrator-driven architecture,” he said, leaves room to expand what it can do. He called agentic capabilities “promising tools for users,” then added: “We just want to make sure that users are aware of what they're giving access to.”

Gene Munster, managing partner at Deepwater Asset Management, was disappointed with those answers. “It's clear they don't want to talk about personalized agents,” he wrote on X after the call, calling them “the biggest risk to the Apple ecosystem.” His estimate, published after the call, put current personalized-agent use at around 20 million people. He wanted an explanation of Apple's position as agents change how consumers use technology.

Analyst Amit Daryanani had asked whether the primary customer relationship might move from the device to the agent. Ternus answered by describing the iPhone as an “intelligent personal hub,” where Siri can use personal data securely and privately.

Apple has already shown an agent-like task involving sensitive information. At its June developer conference, it demonstrated software that would find weak or compromised passwords, visit the relevant websites to change them, and save replacements in Passwords. It appeared in early summer betas, then disappeared. Apple's marketing website still lists it as coming in a future update.

The bench behind him

Mike Rockwell, who built the Vision Pro headset and led Siri AI development, is still seriously considering stepping down from his role or leaving Apple next year. His current job is vice president of Siri engineering.

Additional responsibility could fall to Sebastien Marineau, a Federighi deputy who leads teams integrating AI into iOS. Some inside Apple see him as a possible eventual successor to Federighi and more committed to AI. Amar Subramanya, a Google veteran hired last year, could also take on more work.

Federighi has given no indication that he plans to leave. Succession is a distant prospect, with little done to establish a clear path for a replacement.

There are spending limits as well. Earlier this month, many engineers began seeing daily AI token caps fall from $300 to $200, with more cutbacks planned. For customers, Ternus said usage limits are coming to features that rely on server-side models, including Siri AI. People who need greater capacity will be able to buy expanded access.

Work on the next operating systems is continuing. The M2 milestone of iOS 28 development begins Monday, Nov. 9, when Apple starts finalizing what it plans to include for users.

Related stories

Apple Cuts Some Engineers' AI Budgets to $200 a Day

Some Apple engineers lost a third of their daily AI allowance in October, with further cuts planned for some teams. Some employees say few workers used the old limit. Others read it as cost cutting and worry about layoffs. Apple declined to comment.

Implicator.ai

Apple Launches $299 HomeView Smart Display With Siri AI

Apple unveiled HomeView, its first smart display, at $299, shipping November 13. Messages and calendars appear only when a paired, recently unlocked iPhone is home, and no outside tester has tried its face recognition yet.

Implicator.ai

Apple Puts Gemini Behind Siri AI Relaunch

Apple's Siri AI relaunch uses Gemini-derived model work while keeping the interface under Apple control. After two years of delay and a $250 million settlement, the beta has to prove the privacy boundary, routing rules and iPhone-upgrade story behind the WWDC reveal.

Implicator.ai

Marcus Schuler

San Francisco

Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai