The Mac Observer reported earlier this month that a “performative AI” trend is spreading at work. According to the survey it covered, 48% of full-time employees admit to faking or stretching the truth about their AI skills, and around 60% say their employers have no clear technology plan. The pattern is easy to see in the Apple ecosystem, where AI has arrived as a built-in feature of tools many knowledge workers already use every day, well ahead of any guidance on how to use it.
AI as a system feature, not an app
Apple Intelligence puts Writing Tools — rewrite, proofread and summarize — inside text fields across supported iPhones, iPads and Macs, so assisted writing happens in Mail, Notes and Pages without anyone opening a separate app. Notes has also been able to record audio and produce a written transcript since iOS 18, and iOS 18.1 added call recording, with the recording and transcript saved to Notes. Summaries depend on Apple Intelligence hardware, while other participants on a recorded call are notified, and local consent laws still apply.
Usage data suggests people are touching these features. In a March 2025 Morgan Stanley AlphaWise survey of 3,300 US consumers, nearly 80% of eligible iPhone owners said they had tried Apple Intelligence, though that subgroup was only about 450 people. Tim Cook said on Apple’s fiscal first-quarter 2026 call that most users with supported iPhones engage with the features, without defining what counts as use. Interest in upgrading for AI is more modest: a 2026 UBS survey of more than 7,500 smartphone users in five countries found only about 24% would upgrade sooner because of Apple Intelligence, while about 31% said it made no difference.
Don’t miss the best of The Mac Observer
Set us as a preferred source and our Apple reporting ranks higher in your Google Search results and Discover feed — one tap, no account changes.
Taken together, the features reach users whether or not they sought them out. That is the setup for the skills gap PwC describes globally. Its 2026 survey of 49,364 workers found 64% had used AI at work in the past year, with daily generative AI use up from 14% to 22%, while the share able to access the learning resources they need fell from 59% to 51%.
Meetings are where Mac and iPhone users meet AI first
Recording and summarizing conversations is an obvious entry point. Estimates of weekly time in meetings range from roughly 11 to 15 hours depending on the source, and Asana puts unnecessary meetings at about 103 hours per person each year. The AI note-taking segment is projected to pass $740 million in 2026, inside a broader note-taking app market growing from about $11 billion in 2025 to $13.3 billion this year.
Apple’s built-in options sit alongside web-based ai note taking services that work from uploaded recordings and run the same way on a Mac, an iPad or an iPhone. Whichever route people choose, accuracy has limits. A 2023 JMIR Mental Health study found the best transcription service reached a median word error rate of 8.9%, near professional human transcription at 7.6%, while another built-in service reached 19.2%. A review of clinical speech recognition found error rates above 50% in real multi-speaker conversation, and a 2024 study by Kirstein and colleagues, as cited in industry analyses, found hallucinated content in up to 37% of AI-generated meeting summaries. Output is a draft to verify, not a record.
When system-wide writing meets detection
Because Writing Tools work everywhere on the device, AI-touched text moves easily into emails, reports and school assignments. An Ahrefs analysis of about 900,000 web pages found 74% of newly published pages contain AI-generated content, and schools and employers increasingly screen submissions. Services such as GPTZero, Turnitin and standalone ai text detector tools are widely used, though text detectors have a record of wrongly flagging human writing, so a flag is best treated as a prompt for review rather than proof.
Watermarking is a different approach. Google opened its SynthID Detector to the public on October 7, but it only finds marks from Google and partners, and its own site says it is not a general AI detector. Apple is reportedly next to join the system, according to Phandroid. Watermarks cover images and video from participating labs; text remains far harder to verify.
The gap Apple can’t close alone
Apple can make capable tools ubiquitous, but it cannot supply what the survey respondents say is missing: a company plan, concrete use cases, and permission to admit they are still learning. Until employers provide those, built-in AI will keep outpacing the confidence of the people using it.
Discussion