Apple's on-device AI approach is finding real-world applications beyond generic chatbots, with Singapore-based developers demonstrating how Apple Foundation Models can make productivity tools more practical while protecting user privacy. Rather than bolting on AI features that feel disconnected from an app's core purpose, companies like Notewise and Basil are integrating Apple Intelligence directly into workflows, allowing users to get AI assistance without sending personal data to external servers.

Why Does On-Device AI Matter for Productivity Apps?

Productivity applications often handle sensitive information that users would prefer to keep private. A note-taking app might contain lecture notes, research papers, meeting notes, or client feedback. A finance app tracks spending habits, payment patterns, and budget information. These are not necessarily details users want transmitted to the cloud. Apple's on-device AI push gives developers an alternative route: make AI genuinely useful while keeping as much processing as possible on the device.

This privacy-first approach addresses a fundamental tension in modern app design. Many developers feel pressured to add AI features simply because the technology exists, resulting in features that feel tacked on without meaningfully improving the user experience. By contrast, apps built around Apple's Foundation Models can integrate AI into existing workflows in ways that feel natural and necessary.

How Are Developers Implementing Apple Intelligence in Real Apps?

• Visual Understanding: Notewise, an Apple Design Awards-nominated note-taking app with close to 1.3 million monthly users, uses AI to understand handwritten notes, diagrams, drawings, and imported files. Users can chat with their notes, ask for quick recaps of page content, or select specific portions using Magic Slab and request explanations.
• Audio Transformation: Notewise's Guided Podcast feature turns notes into audio explanations that behave like guided lessons rather than simple text-to-speech readings. As the podcast plays, the app scrolls to relevant sections and focuses on the material being explained, making it useful for students reviewing 100-page lecture notes or researchers working through long papers.
• Automated Data Processing: Basil, a finance-tracking app built by Swapnil Bapat, uses Apple Foundation Models to automate expense logging from common payment methods including Apple Pay, PayNow, PayLah!, GrabPay, and ShopeePay, reducing manual entry friction that causes users to abandon tracking altogether.

What Makes These Apps Different From Typical AI-Enhanced Tools?

Notewise founder Fan Weiguang, who graduated from NUS Computer Science in 2019 and later worked at Google as a software engineer on Search, explains that the difference comes from deep product understanding rather than simply having access to AI technology. "AI has made a lot of things much easier for everyone," Weiguang stated. "The advantage now comes from your understanding of the space you are in, and whether you are able to keep up with the pace of innovation. Are you able to come up with features like the podcast feature? Are you able to deliver something before people even think of that kind of feature?"

"One thing that is very different from typical notes is that it is visually rich. It is not just text, tables or structured information. It can include people's drawings. Not every app in the market is capable of understanding this properly," said Fan Weiguang.

Fan Weiguang, Founder of Notewise

Notewise uses a hybrid approach, relying on Apple Intelligence and Foundation Models where possible because they offer low latency, availability, and stronger privacy protection. If a document exceeds what Apple's on-device models can handle, Notewise may route the task to online models. The aim is to keep the experience responsive while preserving local processing whenever it makes sense.

Basil's approach reflects similar priorities. Swapnil, who works as an AI data product designer with an engineering background from NTU and service design training from the UK, built Basil because he was frustrated with the laborious work involved in tracking expenses. Early test users responded to two things: the cleaner interface and the fact that Basil removed much of the hassle from expense tracking.

How Do These Apps Handle Privacy Concerns?

Privacy considerations drove both developers' technology choices. Swapnil explained his decision to avoid integrating OpenAI or Anthropic models into Basil: "Because I promised users privacy, nothing leaves your device at all. I am not going to integrate OpenAI models or Anthropic models into the app. I do not want your sensitive information to be sent to a server just so I can get an AI response that sounds better. To me, that is not privacy".

Swapnil

"Because I promised users privacy, nothing leaves your device at all. I am not going to integrate OpenAI models or Anthropic models into the app. I do not want your sensitive information to be sent to a server just so I can get an AI response that sounds better," explained Swapnil Bapat.

Swapnil Bapat, Founder of Basil

This privacy commitment also shapes Basil's business model. Swapnil says he does not want to charge users just so they can save money, nor does he want to hide useful tools behind a subscription. "I made a promise that Basil will always be free. The features will always be free. I will never gatekeep them behind subscriptions, because that is not what I want," he stated. Using Apple Foundation Models allows Basil's AI features to run on-device, work without an internet connection, and avoid API costs, which matters for an app he intends to keep free.

Swapnil

What's Next for On-Device AI Development?

Weiguang is particularly interested in how Apple's new generation of Foundation Models could expand on-device capabilities. The multimodal nature of these models, which now support image and audio input beyond text, could enable more visual understanding to happen offline. He is also exploring Siri's onscreen awareness, App Intents, and app actions, which could allow users to ask Siri for a recap of notes from a specific class or start a note by voice without opening Notewise first.

These Singapore-based examples suggest a broader shift in how developers approach AI integration. Rather than treating AI as a feature to bolt onto existing products, successful implementations treat it as a tool that can make core workflows lighter and more intuitive while respecting user privacy and maintaining the app's original purpose. As AI tools become easier to build, the competitive advantage increasingly comes from thoughtful product design rather than simply having access to the latest models.