Apple Adds an AI Assistant to Keynote, Pages and Numbers in Beta

Apple is rolling out an AI assistant to Keynote, Pages, and Numbers in beta, starting with devices that support Apple Intelligence. The feature works on devices running iOS 27, iPadOS 27, or macOS 27 or later. Apple's own newsroom frames the update as part of Creator Studio, its subscription bundle of creative apps, and 9to5Mac and MacRumors both name the feature Intelligent Assistant. Pages, Keynote, and Numbers make up Apple's iWork suite.

The assistant is built on AI models from Apple and OpenAI, according to The Verge. Its scope differs by app, and each one uses the open document as context.

In Keynote, the assistant can design, animate, and refine slides. In Pages, it can draft written content, restructure a document, and tweak existing passages. In Numbers, it can calculate and visualize data and organize it into structures. It can also suggest chart types and build custom tables. The more useful test will be how well it handles real documents with formulas, cross-sheet references (formulas that pull values from other sheets), and layout constraints that a casual demo does not exercise.

The most significant detail for developers is protocol support. Keynote, Pages, and Numbers support the Model Context Protocol (MCP), an open standard that lets applications expose their functions and data to language models, on macOS 27 or later. That allows third-party AI agents, programs that carry out tasks on their own, to plug into the apps. An external agent could read, edit, or generate content in a presentation or spreadsheet through a defined interface, rather than through screen automation or file export. Whether third-party agents gain meaningful control over document state (the live contents and structure of a file), and what permission model governs that access, will determine how much the feature matters beyond Apple's own assistant.

The same update extends to other members of Apple's creative lineup. Pixelmator Pro is introducing a Generate Fill feature that edits part of an image based on a text description. Logic Pro for Mac and iPad is receiving a Liquid Glass redesign and an AI assistant that can set up a vocal track from a single prompt. Apple is also building a Content Hub directly into Final Cut Pro, with stock videos, photos, and backgrounds. Updates to Creator Studio will be available to existing subscribers for free in November.

The assistant sits within a larger pattern. In June, Apple introduced Siri AI, which it describes as a more capable and personal assistant powered by Apple Intelligence. Apple Intelligence, announced in June 2024, is the personal intelligence system that places generative models at the core of iPhone, iPad, and Mac. In October, Apple introduced HomeView, a new way to enjoy entertainment and manage the home, and said Siri AI can pull up information from a user's Calendar app there. Siri AI is a separate product from the iWork assistant, and the two share a model stack rather than a feature set.

Apple has also stated that, due to the Digital Markets Act, it will not ship Siri AI in the European Union with the release of iOS 27 and iPadOS 27. That delay applies to Siri AI specifically. The announcements covered here do not address EU availability for the productivity-app assistant, so readers in the bloc should check Apple's current guidance before assuming any feature set.

In my view, the most important question for this rollout is less about the chatbot interface than about the MCP surface. A productivity suite that exposes its document model to outside agents changes the competitive shape of the category, because the assistant becomes one client among several rather than the only door into the application. Apple has not, in the material reported so far, detailed how it will govern those agents' permissions, and that is where the substance will sit.

The broader context here is a platform that has spent years keeping inference on-device where it can and routing heavier requests elsewhere. Using Apple's own models alongside OpenAI's in a single feature shows how that tradeoff is being handled in practice. For professionals who build slide decks, financial models, and manuscripts every day, the practical gain is less tedious formatting and restructuring. The risk is familiar: generated content still needs review, and a faster draft is only an improvement if someone checks it. The optimistic reading is that the time saved goes back into the thinking the documents are meant to carry.