For the millions of organizations that rely on Microsoft Office 365 for productivity, Microsoft Copilot is likely to be used by some, but not all, of their users. Many of these users are also likely to have a personal or corporate iPhone, iPad, or Mac.
This article highlights key similarities and differences between Apple Intelligence and Microsoft Copilot and concludes with several predictions, including insights on the ability for Microsoft and Apple generative AI tools to work together.
In June, at its worldwide developers conference (WWDC), Apple creatively repurposed the “AI” acronym from the standard “artificial intelligence” to “Apple intelligence” -- although the company never seems to abbreviate it.
According to Apple, Apple Intelligence is “the personal intelligence system that combines the power of generative models with users’ personal context” helping users to “understand and create language and images, take action across apps, and draw from users’ personal context to simplify and accelerate everyday tasks.”
This is very similar to the expressed aspirations of Microsoft Copilot, which Microsoft describes as “AI for everything you do”, claiming that Copilot can help you “Work smarter, be more productive, boost creativity, and stay connected to the people and things in your life” by acting as “an AI companion that works everywhere you do and intelligently adapts to your needs.”
Apple has focused on the individual consumer and Microsoft on the enterprise, yet both are now competing to deliver AI solutions that assist the same end user. Comparing these solutions is challenging and increasingly important given their respective success in increasingly overlapping populations.
Security, privacy, data residency, and compliance are of critical importance to enterprises and increasingly understood and being evaluated by individual consumers.
Both Apple and Microsoft emphasized their AI products’ privacy and data security features.
Apple emphasizes that processing of personal data often happens locally, leveraging the more powerful chips in later-model phones and laptops. This approach is analogous to how Face ID biometric data is handled; the data never leaves your device and is never backed up to iCloud or anywhere else. When additional cloud-based processing is required to deliver “intelligence,” Apple claims that its “Private Cloud Compute” architecture will keep your data secure and encrypted. Your data is only used to process your request, never stored, and is not accessible to anyone, including Apple.
Apple does support ChatGPT through an OpenAI partnership; Siri and other Apple writing tools can tap into ChatGPT for certain requests. The user is asked each time before any information is shared with OpenAI. According to Apple, privacy protections are built in for users who access ChatGPT, IP addresses are obscured, and OpenAI won’t store requests; however, ChatGPT’s data-use policies apply for users who choose to connect their account.
Copilot for Microsoft 365 is compliant with existing privacy, security, and compliance commitments to Microsoft 365 commercial customers, including the General Data Protection Regulation (GDPR) and European Union (EU) Data Boundary. Copilot honors existing organizational permissions available in Microsoft 365 services, such as SharePoint. Copilot cannot access data that the user has not already been granted access to.
Prompts, responses, and data accessed through the Microsoft Graph (which captures collaboration information in an organization) are not used to train foundation LLMs (large language models), including those used by Microsoft Copilot for Microsoft 365.
Copilot for Microsoft 365 uses Azure OpenAI services for processing, not OpenAI’s publicly available services. (In contrast, Apple Intelligence appears to leverage OpenAI’s public services.) Copilot operates entirely within your existing Microsoft 365 tenant service boundary. Azure OpenAI does not cache customer content or Copilot prompts when using Copilot for Microsoft 365.
As a slight aside, but relevant to this summary: Microsoft announced a new class of Windows 11 AI PCs they are labeling as “Copilot+ PCs”.
Akin to the newest Apple devices, Copilot+ PCs leverage powerful processors and multiple state-of-the-art AI models, including several of Microsoft’s world-class SLMs, to unlock a new set of experiences you can run locally, directly on the device.
Both Apple Intelligence and Microsoft Copilot can create, refine, and summarize text. Both can generate original images based on text descriptions. Both can answer questions based on your personal information and web content.
Apple Intelligence announced capabilities to help manage notifications and to prioritize important messages. Additionally, Apple has focused on self-expression items such as generating custom emojis (Apple calls these genmojis because they use generative AI) or generating a custom movie based on a description and the content of your photo library.
Apple Intelligence will work on newer iPhones, iPads, and Macs.
Microsoft has released multiple versions of Copilot and has included “copilot” functionality in 20 or more of its products. This makes it challenging to describe the general features of “Copilot.”
Most business organizations would choose to license Copilot for Microsoft 365, as this version integrates Copilot’s generative AI capabilities into Word, Excel, PowerPoint, OneNote, Outlook and most notably Teams.
Microsoft has enabled Copilot for Microsoft 365 with application-specific capabilities. For example, Copilot in PowerPoint can create slides, including speaker notes, from an existing document, Copilot in Outlook can summarize a long email thread, Copilot in Word can draft a document based on multiple other documents, Copilot in Excel can help analyze and visualize information in a data table.
At present Copilot capabilities are best surfaced in Teams, providing meeting summaries, chat thread summaries, along with the ability to have the Copilot assistant in Teams pull together emails, meetings, and chats to summarize what is new or to help you prepare for an upcoming meeting.
The chart below summarizes three of the main Copilot versions.
Copilot capabilities are embedded into the web, desktop, and mobile versions of the Office 365 apps. There is also a mobile Copilot application that is available for both iOS and Android phones.
While the new generative AI models used by both Apple Intelligence and Microsoft Copilot are capable and have general knowledge based on their training dataset, it is through the process of “grounding” that AI tools can provide more personalized and relevant information.
Grounding refers to the process of providing Large Language Models (LLMs) with specific, relevant information that is not part of their pre-trained knowledge.
Apple Intelligence, as demoed at their WWDC, has access to all your on-device information to provide context. For instance, Apple says “[A user] could ask, ‘When is Mom’s flight landing?’ and Siri will find the flight details and cross-reference them with real-time flight tracking to give an arrival time.”
Where Copilot for M365 shines is in its ability to access a broader range of organizational knowledge via the Office Graph. The Office Graph provides Copilot with context across your documents, presentations, email, calendar, notes, and contacts. The Microsoft Graph enables Copilot to scan for the best meeting times for a group of people, examine the org chart to identify manager relationships, find the most relevant people related to a topic or a person, or reach other members of a particular group. The Graph is a conduit to all the organizational knowledge you have permission to access, not simply information that exists on one of your devices.
Importantly the Office Graph is extensible. Microsoft graph connectors allow you to bring external data into the Microsoft Graph, which in turn makes Copilot smarter because it has better context. External data could include a human resources database or product catalog, whether this information is hosted on-premises or in a public or private cloud.
Apple Intelligence will be included at no additional cost, provided you have or upgrade to one of the newest Apple devices. With Apple Intelligence, the optional access to ChatGPT, powered by GPT-4o, is free, and users will not need to create a ChatGPT account.
As illustrated in the table above, while there is a free version of Copilot, for medium and large organizations, adding Copilot capabilities to the Office 365 apps costs $360 per user per year.
In short, Copilot has been available for over a year while most will need to wait another year for Apple Intelligence.
The free version of Copilot was initially introduced in February of 2023, at the time it was called “Bing chat” and only worked when using the Microsoft Edge browser.
Microsoft Copilot for M365 was made available to enterprise customers starting November 1, 2023. It was then made available to any sized business in January 2024 and the paid Copilot Pro subscription for individuals was announced.
Copilot for M365 and Copilot Pro are available worldwide and currently support over 25 languages. (More accurately Copilot is generally available for purchase worldwide in public clouds; Microsoft has not announced a timeline for when Copilot features will be available in sovereign clouds)
Apple Intelligence will debut in beta state when iOS 18, iPadOS 18, and macOS Sequoia are released later this year, likely around September. The initial release will only support U.S. English. Expect Apple Intelligence to fully launch sometime in 2025, which should include support for additional languages.
Apple has had great success in creating consumer devices; albeit devices that are often also used for business. New AI features for Apple are intended to drive device upgrades/sales.
Eventually, all Apple users will have access to Apple Intelligence capabilities, as they replace their devices over time.
Microsoft has focused on broad-based software and gained traction primarily in the commercial space. AI features increase ARPU (average revenue per user) by incenting organizations to purchase additional licenses, beyond the Office/Microsoft E3/E5 licenses. (Effectively the add-on Copilot, Pro, Premium, and Plus licenses have replaced the imagined E7 bundle.)
Microsoft has chosen to leverage its collection of “copilot” generative AI tools, and leverage its estimated $10+ billion investment in OpenAI by creating a plethora of add-on paid licenses. (In February 2024 I counted 8 Copilots; by March 2024 at Enterprise Connect, I counted 11 Copilots; others have suggested there are now as many as 20 Copilot add-on licenses.)
Microsoft is charting a unique path where organizations will need to decide, justify, and then pay more for users that require AI capabilities. As no sizable organization is likely to spend $360/user for thousands of users, it is almost certain that businesses standardized on the Office suite will have two classes of users, those with generative AI capabilities, and those without.
Driving adoption, sustained usage, and measurable business outcomes will be key to organizations acquiring and then renewing Copilot licenses. (Recognizing this, Microsoft has created an extensive Copilot Success Kit that provides both guidance and customizable materials that Microsoft claims will help “achieve rapid value with Copilot while enabling your progressive skilling journey with AI tools.”)
Of Note: In the unified communications space, both Zoom and Cisco Webex have decided to include their AI Companion and AI assistant capabilities at no additional cost.
AI (artificial intelligence … not Apple Intelligence) for years has been working behind the scenes on Apple devices and within Microsoft applications.
AI has been eliminating background noise from calls, helping transcribe voicemail messages, allowing us to “talk” to our devices (speech to text). AI has also been helping us find specific people or objects in images, blur or change the background during video calls, or remove undesirable objects from photos.
This “behind the scenes” AI has been well-adopted and for most users is appreciated as new “magical” application features.
In contrast, most of the initially released generative-AI features require user invocation.
With current Microsoft Copilots, you, as the pilot most often need to invoke the AI-powered features. In fact Microsoft has “leaned into” the term copilot by emphasizing that you are always in control.
Apple Intelligence, although not yet available for hands-on confirmation, appears to be positioned both as an as expert, via an enhanced Siri, to respond when invoked and as a helpful assistant, working in the background, to surface and prioritize without an explicit invocation.
At the Build 2024 conference Microsoft discussed two new extensions to the Copilot concept. Team Copilot which is described as a personal assistant working on behalf of a team to improve collaboration and project management, and custom copilot agents that work independently and can be triggered by events to orchestrate and automate business processes.
The most successful AI is likely to embedded itself directly and transparently within the business process or task, much how current spell and grammar checkers simply alert us as we compose text. Expect both Microsoft and Apple to need time to experiment to optimize current AI approaches.
Will Microsoft and Apple work together to deliver “Copilot Intelligence” or “Apple Copilot” or will each fight for a knockout blow of the other?
If we look to the world of unified communications, it has taken years, arguably decades, for leading vendors (Microsoft, Zoom, Cisco, etc.) to better interoperate. Similarly after many years, Apple only recently agreed to support RCS (Rich Communication Services) which bridges the gap between the iOS and Android messaging experience, as part of the still-to-be-released iOS18. Apple’s capitulation arguably being driven in part by regulatory issues and scrutiny.
History suggests several “rounds” where AI approaches between Apple and Microsoft compete. Perhaps after a multi-year “slugfest” we will eventually see more of a “tag team” approach, Apple and Microsoft working together to battle task inefficiency and information overload.
In the shorter-term, I expect both Apple Intelligence and Microsoft Copilot to deliver wonderful and magical capabilities, except when they don’t. Ultimately both products are very similar, they are brand new and still experimenting and evolving as they seek to provide the most value for their users.
This three-part series provides an inside look at the generative AI inside Microsoft Teams, Zoom, Cisco Webex, and Google Meet.
You can also check out Kevin Kieller and Brent Kelly's session "DEEP DIVE: GEN AI-BASED PERSONAL ASSISTANTS: STRAIGHT TALK ON VALUE & USE CASES" at Enterprise Connect AI on October 1-2 at the Santa Clara Convention Center.
