What to expect, what they’re really testing, and what a strong answer looks like — scored.
User experience quality, hardware/software integration, privacy, and premium brand positioning. Apple PMs must think about the full end-to-end experience (not just software), understand Apple's constraint-based design philosophy, and reason about how features interact across the Apple device ecosystem.
The question below was asked by Apple interviewers. The answer is graded on the five dimensions real PM interviewers use: structure, specificity, reasoning, decision quality, and delivery.
“How would you improve Siri to better compete with ChatGPT?”
I'd reframe the question slightly: the goal shouldn't be to make Siri 'beat' ChatGPT — it should be to make Siri the best assistant for things Apple's users actually do on their devices. ChatGPT is a general-purpose reasoning tool; Siri's advantage is deep OS integration. Competing directly on reasoning quality is a losing battle in the short term.
The gap I'd target: Siri is excellent at single-turn commands ('set a timer,' 'text Mom') but fails at multi-step tasks that require understanding context. If I ask Siri 'what's a good restaurant near where I'm meeting Sarah tomorrow?' it can't cross-reference my calendar, Messages, and Maps in one answer. ChatGPT can't do this either — but Siri should be able to, because it has the data.
Proposed improvement: expand Siri's context window to span on-device app data — calendar, messages, contacts, health, photos metadata — and allow multi-step queries that reference this data. 'Siri, based on my usual commute, what time should I leave for my 3pm meeting?' is an example of a query only Siri can answer, because the commute data is on-device.
This aligns with Apple's privacy positioning — all inference happens on-device, no data leaves the phone. It's a Siri-specific moat that ChatGPT cannot replicate without device integration.
Success metric: multi-step query success rate — measured by whether the user explicitly says 'thanks' or follows through on Siri's answer without rephrasing. Guardrail: don't increase Siri's error rate on single-turn commands, which is the current reliability baseline.
Reframes the competitive question, identifies a specific capability gap, proposes a solution that aligns with Apple's constraints.
The cross-referencing calendar/Messages/Maps example is concrete; the commute example is instantly relatable.
The 'competing on reasoning is a losing battle' reframe is courageous and correct; the on-device privacy moat is non-obvious.
Commits to a specific capability expansion; the guardrail metric shows awareness of unintended consequences.
Well-paced; the reframe at the top sets up the rest cleanly.
The strategic reframe from 'beat ChatGPT' to 'be the best assistant for Apple users' is the answer's biggest strength — it shows competitive thinking and honesty about Apple's short-term position. Identifying multi-step on-device context as the Siri moat is correct and well-argued. The weakness is that the success metric (implicit 'thanks' or follow-through) is hard to measure reliably and the answer doesn't acknowledge that.
Propose a more measurable success metric than implicit thanks — for example, rate of query rephrasing within 30 seconds, which indicates the first answer failed.
Paste any Apple PM interview question and your answer. Get scored on the same five dimensions — instantly, free, no signup.
Grade my answer free →First grade is free. No account needed.