Apple adds AI to Pages, Keynote, and Numbers

What did Apple announce?

According to The Verge, Apple is adding an AI assistant built on models from Apple and OpenAI to Keynote, Pages, and Numbers. The assistant can use a document’s context when carrying out tasks.

In Pages, Apple says it can create drafts, reorganize text, and revise writing. In Keynote, it can help design slides, add animation, and refine presentations. In Numbers, it can perform calculations, organize data, create visualizations, recommend chart types, and build custom tables.

The assistant is rolling out in beta for devices compatible with Apple Intelligence, running iOS 27, iPadOS 27, or macOS 27 or later. Keynote, Pages, and Numbers also support the Model Context Protocol on macOS 27 or later, allowing third-party AI agents to connect to the apps.

What could this mean for a small business?

For a team already working in these apps, the potential benefit is less manual drafting and formatting within everyday documents. A manager could test it on a presentation outline, a rewritten proposal, or a table for an internal report. These are sensible trial tasks, not promises of better results.

The important distinction is between producing an output and approving it. A polished slide can still contain a weak argument. A clear paragraph can still misstate an offer. A useful-looking chart can still be based on the wrong calculation. Keep a person responsible for the final business decision.

What should managers check before a trial?

Start with device eligibility. Check both Apple Intelligence compatibility and the operating system requirements before planning a rollout. Since the assistant is in beta, avoid making an essential customer deadline depend on it.

Set a data boundary before staff begin experimenting. Use sample or non-sensitive documents first. Review the applicable data-handling terms before entering customer details, employee records, or confidential financial information. The report describes capabilities, but does not establish whether they meet your business’s privacy requirements.

Treat third-party agent access as a separate decision. Model Context Protocol support is not a reason to approve every agent. Assess the proposed task, access permissions, and review process before allowing an agent into business documents.

What practical steps should your team take next?

Choose one recurring task and define what a good result looks like. For a proposal draft, check factual accuracy and editing effort. For a spreadsheet task, verify formulas and totals against a known answer. For slides, review the message as well as the appearance.

Compare the assisted workflow with your current method. Record the time spent prompting, checking, and correcting, not just generating. Expand the trial only if the complete process is useful, and keep a manual fallback.

For a broader approach to delegated work, Terabot gives businesses digital workers: AI workers with their own computer, a private virtual machine, that work in the apps a team already uses and work around the clock. It is in private preview, by invitation only. The same evaluation rule applies: start with a defined task and check the result.