When a product pushes AI or new interaction models before it gets the basics right, usability suffers. There is a consistent pattern in recent updates by Microsoft of elevating AI features to prominent positions in the interface while core functionality becomes harder to find or less stable. This challenge is not unique to one product in Microsoft 365, and it’s not without consequences.
When an interface adds extra steps, hides navigation, or forces guesswork, it doesn’t matter how modern it looks because people stop relying on it when it doesn’t intuitively and efficiently support their day-to-day needs. End users create workarounds while confidence in the platform quietly (or in my case, loudly) erodes. My first product ownership role with Microsoft 365 was at a biotech where navigating SharePoint Online and Teams efficiently truly did impact our ability to save lives. If I did not structure the solutions effectively, shadow IT moved to the Google stack or other competing products outside our purview and more importantly, security boundaires.
I am increasingly concerned that consistently prioritizing AI features front and center, while measured user needs around stability, clarity, and discoverability remain unresolved, will have long-term consequences. There are many competing products in the market, and I already see organizations beginning to reassess whether Microsoft 365 still supports the way they actually work.
I am vocal about UX because I believe deeply in this ecosystem. I have seen what Microsoft 365 can enable when it is intuitive, reliable, and thoughtfully designed. If Microsoft’s mission is to “empower every person and every organization to achieve more”, then the foundation must consistently support how people truly work. This has been lost in the race of Artificial Intelligence (AI) to showcase what is newest.
If you guessed by the title… yes, I am calling AI flair right now and I know that is a bold statement. AI has transformative potential for many organizations. I use it in my personal and professional life. I speak about it at conferences and train on it with my clients. That is not the problem we are seeing in Microsoft 365’s direction right now. I am talking about the prioritization of AI, and surfacing it in the user interfaces, ahead of ensuring stability of the product, clarity of functionality, and ahead of navigation that people depend on daily. This is when it becomes flair.
Nielsen Norman Group’s usability heuristics are not new. They are decades old. And they endure because they are grounded in human behavior, not product cycles.
Recognition rather than recall.
If users must remember where something moved because space was cleared for AI entry points, we have increased cognitive load. See more on why I think OneDrive and Microsoft 365 Copilot Search isn’t ready to replace the workday start experience and how the SharePoint Document Library UI update misses the mark.
Visibility of system status.
If the system feels unpredictable because features shift to accommodate the newest intelligence layer, confidence erodes. Again, what is happening in these SharePoint document libraries?
Consistency and standards.
When core navigation patterns change repeatedly to elevate AI experiences, muscle memory breaks. For anyone navigating multiple systems in your life… wait, that is all of us…. you understand muscle memory is not trivial as it makes enterprise software feel dependable under pressure. This is not theoretical. In the biotech, when we were moving quickly (let’s be real… that was every day), collaborating across regulatory, clinical, and research teams across the globe, there wasn’t time to relearn where something lived when a new feature was released. Especially if it wasn’t one we were using or the business relied on. We did not have margin for navigation experiments by Microsoft (that have been proved ineffective by Apple iOS 17 & 18 referenced in Microsoft Teams app bar changes: When “clean” design undermines usability). The platform needed to be stable and intuitive to prioritize the work being done within it. That is the standard enterprise software must meet.
In Is It Really Social Permission or Should It Be Product Restraint?, I wrote about the responsibility of companies to exercise restraint because at the end of the day, simply because they can ship something does not mean they should ship it. Especially not front and center and the same applies here. AI does not need to dominate the interface to prove its value.
In Context for Systems Is Not the Same as Context for People, I argued that system intelligence does not equal human context. Just because an AI model can summarize, predict, or suggest does not mean it understands the workflow moment someone is in. This is where I see the growing tension with where AI is today and how tech companies, specifically Microsoft, are deploying system changes.
Microsoft is optimizing for system intelligence visibility. Users are optimizing for task completion clarity. These are wildly different objectives.
One of the pillars of user experience is Hick’s Law: the more choices presented, the longer decision-making takes. We cannot lose sight of this core usability rule when adding AI buttons, prompts, suggestions, and new interaction models so it seems Microsoft’s response is to remove existing functionality, assuming expertise for all users though we have new users in Microsoft 365 every day. This does not reduce friction and instead increases decision complexity.
If AI is layered on top of already strained navigation, we are not simplifying work but instead we are multiplying touchpoints. For example, in SharePoint lists and libraries, content types are a foundational element of the information architecture. They sit above columns (and preferably site columns for reuse) in the hierarchy. Columns define metadata but content types define the structure those columns belong to and therefore what metadata we need for varying types of content. That relationship matters to drive effective search across the platform as well as custom search experiences.
For the past few years in a SharePoint list or library, this hierarchy is not reflected accurately. To add or manage content types in the modern front-end, users must select “Add column” to find content types which is effectively positioning content types as if they are subordinate to columns. They are not and it completely reverses the mental model. In the UI update, users are now prompted to “Add or remove fields” with the concept of content types entirely lost. I will save my rant on creating these items locally in lists and libraries, which prohibits reuse, for another blog though it is another opportunity missed by Microsoft to empower us as end users to “achieve more.”
When the UI contradicts the architecture, confusion always follows and countless opportunities are missed. Information architecture should be reinforced by the interface, not obscured by it, especially in enterprise environments where governance, compliance, and structured metadata matter. Clarity in hierarchy is not optional.
This is exactly the kind of foundational usability detail that should be resolved before additional AI layers are surfaced. If the basics of structure are misrepresented in the UI, adding intelligence on top of it does not fix the underlying friction. It amplifies it.
If Microsoft wants AI to be transformative for all organizations across the globe, it must be integrated in a way that honors foundational user needs (found through usability research which I plea for more transparency on in Microsoft Teams app bar changes: When “clean” design undermines usability), respects cognitive load, and demonstrates product restraint when necessary. I genuinely love Microsoft 365 and have built a career around it as a user experience practitioner and advocate because I believe in what it can be when it truly centers people. At its best, this platform fades into the background and lets meaningful work take center stage. AI should strengthen that foundation, not compete with it. It should reduce friction, not introduce new decision points. It should enhance clarity, not require users to relearn navigation patterns they depend on daily. Innovation does not require destabilization.
AI is not the problem. Prioritization is. The long-term health of this platform depends on getting that order right.
Emily Mancini 