The CollabTalk TweetJam’s 2025 M365 Mega Recap & 2026 Bold Bets

In case you missed it, we wrapped the year with our biggest “recap + predictions” TweetJam format: seven questions designed to surface what actually shifted in 2025, what surprised us in real-world Copilot usage, and what feels inevitable (even if it sounds a little unhinged today). Rather than rehash headlines, the panel kept coming back to what’s happening inside organizations: behaviors, governance reality checks, and the messy middle of adoption.

I was joined on the panel by fellow MVPs Ben Stegink (@benstegink), Adam Ball (@adamcball), Charles Lakes II (@the2lakes), and Shari Oswald (@shortcutshari). And thanks to the commenters following the live discussion, including our 6th Beatle, MVP Wes Preston (@idubbs) who has been a regular on CollabTalk TweetJams for the past decade or more (which began in January 2012).

You can watch the full-length panel discussion here. Below the livestream recording is my question-by-question recap of the discussion. Enjoy!

Q1: Looking back at 2025, what was the single most important shift in the Microsoft ecosystem, and why will we still be talking about it three years from now? Tech, culture, adoption, or business model.

The group largely converged on AI—especially Copilot—as the obvious “shift,” but with a more nuanced take: the real change wasn’t just a feature release, it was an organizational wake-up call. Copilot didn’t magically fix anything; it exposed everything. Teams are realizing they can’t just “buy their way out” of years of content sprawl, weak information architecture, and loose permissions. That realization is pushing governance back to center stage, and in three years, we’ll still be talking about this period as the moment organizations accepted that AI requires clean, intentional data to be useful.

Q2: Copilot is no longer new and is everywhere. What surprised you most about how people actually used Copilot in 2025 versus how Microsoft or the community expected it to be used?

The surprise wasn’t that people used Copilot. It’s how modestly they used it. Instead of “big transformation” moments, many users leaned on Copilot for daily friction removal: summarizing long emails, catching up after vacation, finding information, or acting like an advanced search tool with natural language. The panel also noted that most people didn’t become prompt-power-users; they kept prompts short and conversational, iterating like they would with a helpful colleague. And yes, the internet spent plenty of time generating AI images and novelty content, which is great for lowering the intimidation barrier, but not exactly the ROI story we’re all trying to tell.

Q3: Agents took center stage in 2025. From research to task automation, what is the first real business problem you believe Microsoft agents will truly solve at scale in 2026?

Agents sparked the most “potential energy” in the conversation. The panel floated several early wins that feel genuinely scalable: reducing SharePoint sprawl by helping identify data rot and bad practices; automating follow-up and relationship management (especially for sales and service workflows); and triaging overload across email and collaboration streams. A big theme emerged around trust: people want agents to take work off their plates, but they’re wary of handing off autonomous tasks if they must spend equal time validating outputs. The near-term value may be “assisted autonomy,” or agents that prepare, flag, and recommend, with humans approving the final move.

Q4: AI and agents exposed a lot of organizational cracks this year. What governance or risk issue caught the most teams off guard in 2025, and which one do you think we are still underestimating?

No contest: sharing. The panel cited real-world examples of massive numbers of organization-wide or overly permissive sharing links, plus OneDrive content effectively shared with thousands of people. Another underappreciated risk is public Microsoft 365 Groups and Teams: when groups are discoverable and joinable, access can expand quietly and rapidly, pulling SharePoint content along for the ride. The key point wasn’t “lock everything down” (that just fuels shadow IT), but to put automated governance in place so sharing is intentional, understandable, and aligned to how people work.

Q5: Not every role benefits equally from AI. Which teams or roles gained the most from Microsoft’s 2025 roadmap, and which struggled to keep up with the pace of change?

The discussion highlighted a growing squeeze on “the messy middle,” especially middle management. Individual contributors can move faster and build “good enough” solutions with Copilot and the Power Platform, but that can create a new mess: spreadsheets as systems, SharePoint as a database, and workflows that don’t scale. Leaders will push for efficiency; developers will push for proper architecture; middle managers will end up mediating both. On the upside, admins may be among the biggest winners as Copilot capabilities expand inside admin centers, making it easier to ask questions, find settings, and operate complex environments without memorizing every role and control.

Q6: Fast forward to the end of 2026. What Microsoft 365 or Copilot capability will feel obvious in hindsight but sounds bold or unrealistic today?

The panel’s favorite “bold” idea: a unified Copilot (or a Copilot for the Copilots). Today’s Copilot experience can feel fragmented across portals and products. A more coherent future is one interface that can reach into security, admin, developer, and productivity contexts seamlessly. Another bold bet: rationalizing overlapping tools (particularly the Loop/OneNote situation) so users aren’t forced to pick between competing “work surfaces.” And on agents specifically, the panel predicted a shift toward smaller, specialized agents that orchestrate together (like good software functions), rather than one giant agent trying to do everything and becoming unreliable.

Q7: Looking three to five years out, what is your bold prediction for how Microsoft’s AI strategy reshapes how organizations define productivity, value, or work itself?

The longer view came back to fundamentals: data hygiene becomes unavoidable. AI will force organizations to purge outdated content, improve metadata and structure, and invest in real governance. Not because it’s trendy, but because the alternative is unusable output at scale. The panel also raised an uncomfortable but likely comparison companies will make: the cost of agents/compute versus the cost of humans, and the emerging “AI-to-human ratios” that will influence staffing models. At the same time, there was optimism around Copilot becoming more like a true assistant; anticipating needs, nudging at the right moments, and shifting performance conversations away from “busy-ness” toward actual output and quality.

—

If 2025 was the year Copilot stopped being “new,” this panel discussion made one thing clear: 2026 is shaping up to be the year organizations stop pretending governance is optional. AI has many talents, but being polite about your data chaos is not one of them.

What surprised me most is that many of us expected 2025 to be the year Microsoft would step up with a stronger catalog of real-world Copilot stories. Not hype, not feature demos, but credible, repeatable examples of business value. Instead, it often felt like a punt, with the implied message being that 2026 is when the ROI narrative really comes together.

Maybe that is not entirely a bad thing. If there is a business value gap, the community is in a perfect position to fill it. We are the ones inside real projects, navigating messy environments, measuring outcomes, and learning what actually works. Microsoft can publish reference architectures and success stories, but practitioners translate that into reality, and reality is where the proof lives.

So my bold prediction for 2026 is simple: more content, more focus on ROI, and more focus on governance. We will see a shift from talking about Copilot as a capability to documenting it as an outcome. What did it replace, what did it accelerate, what did it prevent, and what did it cost to get there? The organizations that win will be the ones that treat governance as the foundation for value, not the paperwork you scramble to assemble after something breaks.