Apple Resets AI Leadership: What It Means for Your Strategy
Apple is entering a new leadership phase in artificial intelligence. John Giannandrea, senior vice president for Machine Learning and AI Strategy, is stepping down and will serve as an advisor until retiring in spring 2026. The transition is already in motion.
For executives, the signal is clear: Apple is prioritizing speed, clarity of ownership, and model optionality after years of mixed progress and delayed releases.
Apple Confirms Giannandrea's Departure
Apple announced that Giannandrea's organization-responsible for machine learning, Siri, on-device intelligence, and its early generative efforts-will be redistributed. He joined Apple in 2018 after leaving Google and became one of the most influential leaders in the company's AI push. The next chapter will be led by new operators with a mandate to ship.
Apple's update also reinforces who's in charge: Craig Federighi continues to guide the broader AI strategy, including work on a more personalized Siri expected next year. CEO Tim Cook thanked Giannandrea for his contributions and welcomed the new leadership to accelerate what comes next.
Amar Subramanya Steps In
Apple has hired Amar Subramanya-previously corporate VP of AI at Microsoft and a long-time Google veteran-as its new vice president of AI, reporting to Federighi. He will lead Apple Foundation Models, ML research, and AI Safety and Evaluation. Siri remains under Mike Rockwell, who also reports to Federighi.
The rest of Giannandrea's teams will shift under Sabih Khan and Eddy Cue to bring applied work closer to product groups. Translation: tighter feedback loops, fewer handoffs, faster shipping.
Why the Reset Had to Happen
Apple was slow to move on generative models. While competitors pushed large-scale systems, Apple initially downplayed them, then shifted gears with Apple Intelligence-rolling out pieces across its platforms. Progress, yes. But the biggest bet-Siri-hit a wall.
The Siri Overhaul: Delay and a Strategic Pivot
Apple revealed a rebuilt Siri in June 2024 for iOS 18, powered mainly by its in-house generative models. After testing, the release was delayed to 2026. The fix involves a major pivot: Apple is turning to Google's Gemini to power advanced Siri capabilities instead of relying solely on its own foundation models.
The executive who led the Siri overhaul, Robby Walker, departed earlier this year after a reorg. The new Siri experience is now expected in spring 2026, likely around iOS 26.4-roughly when Giannandrea fully exits.
Google Gemini (developer overview)
What This Signals to Operators and Boards
- Build + partner is the operating model. Treat foundation models like core suppliers. Keep internal R&D, but maintain external options for reliability, coverage, and speed.
- AI Safety and Evaluation is now a line function. Make evaluation an independent gate with clear metrics for quality, latency, safety, and cost. No greenlight without passing thresholds.
- Organize for shipping. Move model teams closer to product owners. Shorten cycles. Incentivize shipped features and measurable user impact, not just demos.
- Timeline realism beats headlines. Stage releases (labs, preview, GA). Tie scope to release trains. Avoid promises you can't hit; users remember misses.
- Talent that blends research with delivery. Hire leaders who've shipped at scale, not just published. Subramanya's track record is the template.
- Economics matter. Track unit costs for inference, on-device vs. cloud trade-offs, and regional privacy constraints. Model performance is meaningless if it can't pencil out.
- Vendor optionality is a risk control. Keep at least two model providers qualified for critical paths. Build portability into your tooling and prompts.
- Clear user communication. Set expectations, phase capabilities, and avoid silent delays. Trust compounds-or erodes.
What to Watch Next
- How Apple sequences Apple Intelligence updates over the next four quarters.
- Subramanya's org design: central platform vs. embedded teams across product lines.
- AI Safety and Evaluation playbook: public standards, eval frameworks, and governance signals.
- Siri's Gemini integration details and the balance between on-device and cloud workloads.
- CAPEX and partner agreements that hint at scale and model mix.
Action Checklist for Executives
- Run a build/partner audit across your AI roadmap; define where external models improve speed or reliability.
- Stand up an independent evaluation function with red-team, privacy, and cost gates.
- Refactor org structure to put model and product teams on the same release cadence.
- Create a vendor playbook with swap-out plans, data constraints, and financial guardrails.
- Set quarterly targets for model quality, latency, and unit economics tied to user outcomes.
- Upskill your leadership bench on AI strategy and operations; make it part of performance objectives.
If You're Building Your AI Capability
Need a clearer view of roles, skills, and training paths for your team? Explore structured learning options by job function.
AI courses by job function - Complete AI Training
Your membership also unlocks:
