While the world was downloading GTA VI or watching Anthropic go public, we Power Platform and Dynamics 365 geeks had more important things to do. We were following Ignite all week, of course. Right?
Three weeks after the Power Platform Community Conference, there weren’t big expectations for BizApps specific feature drops from Ignite 2026. Being the biggest Microsoft festival of the year, the most interesting bits are usually the surrounding technology that connects with Power Platform, rather than being inside the perimeter.
I’ve put down some thoughts and reactions on announcements across Copilot, Work IQ, Copilot Managed Runtime, Agent 365, Copilot Studio, and Planner. If you’re looking for something specific, Jussi Roine’s Ignite 2026 Recapped lets you browse the announcements by product area and go straight to the Microsoft posts.
Copilot
From “Copilot in X” to “X in Copilot”. That’s the main shift that is taking place. All the various sidecar experiences that Microsoft product teams were forced to build a couple of years ago are now gradually becoming obsolete. In-product assistants in places like the flow designer are getting removed in favor of working in the Big Chat of the main Copilot.
Features ranging from enabling in-app browsing by default to making full Office apps available inside Copilot all aim to give more tools and data for the LLM to work with. This makes sense because it’s the practical way to get AI to produce useful and accurate results. Waiting for a language model to achieve AGI through mere scaling laws would be less sensible as a strategy. Just train AI to use the same software we humans use instead and watch the digital coworker with infinite Excel skills and finite judgment emerge.
A notable exception to the “X in Copilot” framing came when Salesforce joined Microsoft’s commercial CEO on the Ignite stage to announce the launch of Copilotforce:

Judson Althoff, CEO of Microsoft’s commercial business, announcing Copilotforce at the Ignite keynote.
Really, Microsoft? You are happy to let a big business apps competitor take your Copilot brand and just casually slap a -force label on it? What’s next, Windowsforce? Teamsforce? Microforce?
There’s nothing special about using a bigger CRM vendor in a partner demo. Back when Satya Nadella became CEO, you could even say there was preferential treatment for Salesforce in certain areas while the in-house Dynamics CRM product was trying to convince customers of the “Better Together” story. At that scale, partnering with a competitor is routine. I’d just imagined Microsoft saw Copilot as a brand on par with Salesforce and would be more protective of it.
Work IQ
In the early days of Microsoft 365 Copilot, there was a lot of talk about the semantic index: how the MS Graph was going to magically allow AI to figure out the relevance of the data the employees have access to. Sure, we did get a search engine that can find relevant results without an exact keyword match. Yet the gap between how the vision was sold and what it did in practice remained pretty substantial. This Kai Lentit sketch from earlier this year captured it perfectly: “AI CEO: Identity access experience gateway, semantic discoverability membrane over high-quality data… Engineer: scattered across SharePoint.”
The idea of grounding Copilot in all of the tenant’s “knowledge” by making it read massive piles of documents and emails has not felt like a huge success to me. You’re ultimately just hoping the agent will pick the right piece of text from a sea of unorganized knowledge work artifacts. It can of course do a convincing job at producing an output that looks like it did just that. In practice, there have been ridiculously low, undocumented limits on how much Copilot can read from your mailbox or file folders.
At Ignite 2026, the Work IQ promise has now expanded to cover business processes, not just business data. Work IQ now includes context from Dynamics 365 and Power Apps. While the proof of how well this actually works will come from real-life tests, I’m a lot more optimistic about getting relevant context into Copilot from business apps than I was about getting it from the generic Microsoft Graph. This of course comes from my bias of having designed and built such business apps for a couple of decades now. If SharePoint documentation and Excel lists had been enough, why did anyone bother deploying structured systems like CRM to begin with?

Sharing settings for a Business Skill in the Power Apps maker portal.
Custom Semantic Model creation in Power Platform is now in preview, allowing you to take your Dataverse schema and explain the table relationships to your AI agents. Business Skills is GA and it offers solution-aware, skill files governed by security roles stored and versioned inside Dataverse. While there would have been nothing stopping you from writing Markdown skill files and feeding them to your agents before this, the fact that there’s a mechanism for keeping things connected across databases, app surfaces and automations is what I believe can make this scalable.
This is the part that makes the whole IQ story useful for a builder, not just the end user. I now see two sides to Work IQ. First, there’s the unstructured world with “I hope Copilot discovers the right version of the proposal document”. Then we have the structured world with “Copilot has a set of dynamic, detailed .md instructions about how our processes should (in theory) work”. Only the latter is a system where customers can directly control the instructions the next-token predictor works from. The collaboration layer is not insignificant, it’s just not been something where I see an opportunity for people like me to solve customer problems. Instead of just training users on “how to use Copilot”, I’d much rather be designing how make a business system with Copilot work for the users.
Copilot Managed Runtime
It’s GA! It’s got an icon! Wait, it’s got a new icon already. Oh, and the SDK repo and tooling were migrated/renamed.

RIP old icon and repo for Copilot Managed Runtime.
The price of being early with Microsoft preview tooling is that packages you’ve installed and the slides you’ve made will be obsolete in a heartbeat. But I don’t mind these changes that much. Sure, it’s a terrible branding choice to talk about “apps on the Copilot Managed Runtime” instead of simply “Microsoft Managed Apps”. Not the first, not the last time stuff like this happens. “Managed Apps” still shows up in the UI and tooling, so people outside Microsoft product marketing will probably keep using it.
Anyway, CMR itself is now GA for the apps part. I’ve successfully built an app fully in Lovable while connecting it to Dataverse and publishing it to the M365 cloud — it was actually a smoother experience than trying something like that with Cowork. But of course business apps aren’t just about the UI. The preview for Workflows on CMR is going to be the next thing to try out, to see how the back-end logic for the app can be built and operated on the Microsoft cloud.

Agents + Workflows in public preview for Copilot Managed Runtime as of Nov 17th.
In many ways, CMR is now introducing an alternate reality for apps and I’ve seen the Dynamics 365 folks ask “why should I use this and not Code Apps?” This feels like a new canvas vs. model-driven moment that will split the crowd for reasons that aren’t purely technical. They can also be about licensing, governance and ALM. For someone like me who loves to write about these things, CMR is a gift that will keep on giving for months to come!
Agent 365
One year ago, when the Ignite 2025 keynote started, I gave Claude a task to go through the MS Learn and other materials published about the newly announced Agent 365 offering. By the end of the keynote, Claude had built the first version of the A365 FAQ site. That may not feel that special anymore, but let’s keep in mind that this was before the Claude Code agentic coding revolution had broadly arrived.
I’ve run a few updates since then to keep the FAQ up to date with the product changes and the May 2026 GA release. This week it was time to do a bigger update for it, given how A365 has moved from the “what is this thing even about” territory into the usual “what’s included with which license?” lifecycle stage of a Microsoft product. That’s what the revised A365 FAQ tries to answer:
When it comes to the actual functionality announced, hands-on experiments are still waiting for a better moment. For instance, I did not have time to dig deeper into the Business Analytics feature set. My own tenant opens the app with no visible data and the links in the licensing warning sent me to the Viva Insights configuration docs. I’d rather not touch any Viva products voluntarily, thank you very much.

“Minimum license requirement for this experience is not met” message when trying to access configurations for the A365 Business Analytics app.
Ryan Jones’s demo tenant was full of neatly categorized data on how employees at a fictitious medical company used Copilot. Wouldn’t it be neat if you could group your organization’s AI usage by domain, top % segments, job functions and all sorts of variables to produce beautiful graphs? LLMs can categorize volumes of log data that would be painful to sort by hand, so a dashboard like this may actually be possible.

Use cases view for Copilot Cowork in the A365 Business Analytics app, showing usage grouping by domain.
The question is, though: when you have that information, watcha gonna do? You’ll see the themes of token consumption, but can that raw data really tell you whether the output was useful for the business or not? Maybe that’s not a question that Microsoft has to answer. Maybe there’s still room and need for human judgment in doing “evals” in the real world instead of streams of tokens.
Copilot Studio
I’ve never felt as productive in the Copilot Studio GUI as I have in the Power Apps maker tools. I was therefore delighted to see that there’s finally an official CLI for building the new Copilot harness agents. I noticed this feature while having the Claude Code session for the A365 FAQ site open and decided to casually give it this prompt:

Giving my Claude Code the assignment to check out the msagent CLI and plan how to build an A365 Advisor agent from it.
A couple of auth commands later, I already had an agent up and running with the FAQ site as its knowledge source. The first time I opened the MCS GUI, the evals planned and implemented by Claude were already running:

The new A365 Advisor agent running evaluations in Copilot Studio before I’d clicked a thing in Copilot Studio.
This is beautiful! Or more specifically: this is how it’s got to be. In 2026, ain’t nobody got time to explore the studio experiences of Microsoft products that change every few weeks. LLMs can patiently read API contracts, inspect the inputs and outputs, and tell you when they actually need your judgment.
With the msagent CLI now available in preview, I might even build some Copilot Studio agents for my real work tasks and not just for customers. It’s now just another tool that I can access from my Herdr terminal and ask my other agents to work with.
Attio is the #1 fastest-growing CRM on Brex and Ramp
Built for the most ambitious companies in GTM, Attio is an end to end revenue platform that allows teams to run multiple plays across different revenue motions and lifecycles.
Teams at the frontier are already building on it: Granola, turbopuffer, Modal, and many many more.
Planner
You may recall how I’ve explored and criticized the way Microsoft has kept Planner Premium, formerly Project for the web, on a separate island from basic plans. The lack of Power Automate support in particular has been baffling to watch, given that upgrading a plan from basic to premium can cost customers core functionality.
At Ignite 2025, the roadmap promised a unification of the two sides. It didn’t happen in the real world, all the while Microsoft’s articles kept referring to “unified plans” without ever defining what that means in practice. Now, during Ignite 2026, the strategy was finally revealed. Unification will mean that more features will be built on the basic plan side and eventually the premium plans won’t be needed anymore.
In practical terms, unified plans means “we’re giving up on trying to figure out Dataverse and building the missing features outside it instead”. There will never be a connector for creating new Planner premium plans and tasks via Power Automate, for instance. The P4W elements on the Power Platform side became instant legacy, with no date for full feature parity.
There’s so much I have to say about this “unification” and the way it is being communicated that I’ll definitely create a future Perspectives issue dedicated to the topic. Once again, existing customers have been left in limbo by the Planner roadmap and they will surely have questions that the official MS comms aren’t able to answer. Meanwhile, if you have questions, thoughts, or need advice on how to deal with Planner Premium plans and automation, please do reach out to me.
One surprising casualty from the Planner unification strategy is To Do, which is retiring at the end of June 2027. Business users will be instructed to grab their data and start using Planner instead. For personal MS account users the message is blunt: “choose another task-management application and prepare for export.” It’s M365 subscription or GTFO.
Microsoft plus consumers just don’t get along these days. Xbox keeps shrinking, Windows is increasingly a platform for ads rather than apps, and Microsoft 365 Personal and Family subscribers get a forced Copilot upsell as a loyalty test of “let’s see if you’re really going to ditch Excel for Google Sheets”. Killing off a free service like To Do for personal MS accounts is in line with the positioning of Copilot as “AI for Work”. For personal life, there’s ChatGPT, Meta’s Muse, and all the rest, I guess.
Life after Ignite
We have to remember what the primary aim of these conferences is. Ignite is the biggest event of the year where Microsoft gets to paint a picture of “this is what we aim to achieve”. Much of the content is aspirational and many technical feature claims made during the conference sessions will never land as described. Plans change, and the pace of change is just getting crazier every year.
I wrote about the post-Ignite hangover and how a lot of those fancy agentic products were still missing in action two months after the event. We need to acknowledge that this was also the time when the Claude Code moment happened, showing how coding agents + services like Cowork were actually delivering on the earlier aspirational AI promises. In hindsight, it probably made sense for many of the Ignite 2025 plans to be scrapped or at least revised before Microsoft pushed ahead.
The technical sessions where product team members get to share their achievements and talk about the coming work are quite valuable forms of communication between Microsoft and the outside world. Much more so than the product marketing blog posts. While I’m not going to have time to watch all the interesting recordings from Ignite, it is luckily very easy these days to ask your coding agent to download the session transcripts and create personalized summaries out of them. That’s what I asked Codex to do this week:

Codex in my Herdr terminal, downloading selected Ignite 2026 transcripts and turning them into notes I can come back to.
The signals from Ignite will then turn into Markdown notes stored in my Obsidian vault, serving as the shared context for both me and the agents that have access to the vault. I’m then able to casually ask “what did Clay Wesener say about CMR Workflows integration at Ignite” and get a response grounded in what was actually said in the session. That’s how I’m putting AI into use for helping me with research and thinking, while still keeping these Perspectives as the outputs from my organic writing.







