On-device AI runs language and image models directly on a phone or laptop instead of in a cloud data centre. Apple Intelligence on recent iPhones, iPads and Macs, and Gemini Nano on supported Android devices and in Chrome, are the best-known examples. For B2B marketers, the change is that buyers increasingly read summaries of emails, notifications and pages written by their own device before they read the original.
What is on-device AI and how is it different from cloud AI?
Cloud AI, such as a chat assistant in a browser, sends your request to a remote server where a large model processes it. On-device AI uses a smaller model stored on the device and the device’s own chip, often a neural processing unit, to handle the task locally.
The trade-offs are clear. Local models are smaller, so they handle shorter and simpler tasks: summarising, rewriting, sorting, smart replies, transcription. In exchange they work without sending data off the device, respond quickly and can run offline. Apple also uses Private Cloud Compute for heavier requests, and both Apple and Google can hand some tasks to larger cloud models, so “on-device” is a spectrum rather than a strict line.
Which features matter most for B2B buyers?
The features that touch marketing are the ones that sit between your message and the reader. At the time of writing, these include:
- Notification and email summaries in Apple Mail and on the lock screen
- Priority sorting of messages in Mail
- Writing Tools that rewrite, proofread or summarise text in many apps
- Summaries of web pages in Safari
- Gemini Nano features on Android such as message summaries and smart replies
- Built-in AI APIs in Chrome that let websites call Gemini Nano locally for tasks like summarising or translating
Availability depends on device, operating system version, language and region, and features roll out gradually. Check current support for the markets you sell into before planning around a specific feature.
How do on-device summaries change B2B email?
A buyer scrolling a lock screen may see a one-line AI summary of your email instead of your subject line and preview text. That summary comes from the email body. If the main point is buried under a long greeting and company news, the summary may say something vague, or the wrong thing.
The practical response is to write emails that summarise well. State the purpose in the first two sentences. Keep one main ask per email. Put dates, prices and deadlines in plain text, not in images. These habits were good practice before, but now a machine reads first and the cost of ignoring them is higher.
Does this affect deliverability and open tracking?
Apple’s Mail Privacy Protection already made open rates unreliable for many recipients, because Apple preloads images. On-device summaries add another layer: a buyer can understand the email without opening it. Clicks, replies and downstream pipeline are better measures of email success than opens.