In early June, Microsoft held its annual Build conference. The conference is aimed at developers, and much of what was presented was technical. But behind the technology was a message that every leadership team with Dynamics 365, or plans to implement it, should take seriously.

Microsoft's own summary of the conference can be found here: Microsoft Build 2026: Be yourself at work (Official Microsoft Blog).

The message is this: Microsoft is now rebuilding its platform around AI agents. Not assistants that answer questions, but agents that perform work. And for the first time, Microsoft is doing this with its own AI models, built from scratch, rather than renting technology from other providers.

What does this mean for you? Not tomorrow, but in the coming years. That is what this article is about.

From assistant to colleague

In recent years, AI in enterprise systems has mostly been about Copilot: an assistant that summarises, suggests and answers questions. Useful, but passive. Humans do the work; AI helps out.

What Microsoft is building now is something else. Agents that receive a task and execute it: registering the order, following up on delivery, preparing the customer meeting, handling the service case. Humans move from executing to deciding and controlling.

This is a bigger shift than it sounds. It affects what your processes look like, what skills you need, and how you measure efficiency. And it will not arrive as one major release on a specific day, but gradually, feature by feature, into the system you already use.

Three things that become more important

Our assessment is that three issues are now rapidly rising up the priority list for any organisation already using Dynamics 365 or considering doing so.

1. Your data determines the result

An agent is never better than the data it works with. If your customer register contains duplicates, if item data is inconsistent or if prices live in a spreadsheet alongside the system, then the agent will make wrong decisions. Faster and at a larger scale than a human would.

Data quality has always been on the to-do list. The difference now is that the cost of postponing the work becomes visible: organisations with proper data foundations can let agents take over routine work, while others will see their investment yield significantly less in return.

The question for your own organisation is simple: would you dare let a new employee make decisions based solely on what is currently in your systems today? If the answer is no, you have found your first project.

2. Someone must decide what agents are allowed to do

When software begins acting autonomously, questions arise that no IT department can answer alone. Which decisions can an agent make without human approval? Who is responsible when things go wrong? How do we detect if an agent is doing the wrong things, albeit efficiently?

These are not technical questions. They are leadership and accountability questions. Microsoft provides the tools to monitor and limit agents, but the decisions on where to set the boundaries must be made by your organisation itself.

Our advice: do not wait until the first agent is live. The organisations that manage this well have deliberated on responsibilities and boundaries in advance, at leadership level, rather than after something has already happened.

3. Lock-in to the platform deepens

Part of what Microsoft demonstrated involves agents that learn your specific business: your processes, your exceptions, your way of working. The longer they are operational, the better they become at your daily routine.

That is valuable. But it also means that this knowledge will not transfer if you switch platforms one day. The tie to Microsoft, which is already strong through data and customisations, becomes stronger.

This is no reason to abstain. But it is a reason to enter with open eyes, and to factor it into contracts and architectural choices right now, while flexibility still exists.

What it means for your partner selection

Here we see perhaps the biggest practical consequence, and this is also where many evaluations are still asking the wrong questions.

When more of the technical standard work is automated, an implementation partner's value shifts elsewhere: understanding which of your processes are suitable for agents and which are not, helping you establish a solid data foundation, and building governance so that agents do the right things securely.

This means that classic selection criteria - number of consultants, number of completed projects, and certifications - say less and less about what you actually receive. The questions that separate the wheat from the chaff in 2026 look more like this:

  • Have you put agents into production for customers, and what was the outcome?
  • How do you approach data quality as a prerequisite, rather than as a post-implementation fix?
  • How do you help the client decide what agents are allowed and not allowed to do?

A partner who answers those questions concretely has done the work. A partner who responds with a slide deck on the possibilities of AI has not.

What the leadership team should do now

None of this requires hasty decisions. But three things deserve a spot on the agenda this autumn.

Make an honest assessment of your data foundation. Not a technical audit, but a business question: do we trust what is recorded in our systems? If not, start there. Everything else builds on it.

Decide how you want to govern agents before you have any. Which types of decisions can be automated, and which always require a human? A short principle document from leadership goes a long way and saves difficult discussions later.

Update the requirements for your partner. Whether you are facing a procurement process or have an established supplier: ask the agent questions. The answers say a lot about who you are dealing with.

Conclusion

Microsoft Build 2026 was the starting pistol for a platform where AI agents perform work within your business processes, learn how your operations function, and need to be governed more like employees than traditional software. The change will come gradually, but the direction is clear.

What matters for you is not the technology itself. It is whether your data holds up, whether governance is in place, and whether your partner actually masters this beyond presentations and promises. The organisations that tackle these questions calmly now will have a clear advantage when AI agents become a natural part of everyday work.

Are you facing partner selection, or do you want to test your current supplier's capabilities in this area? That is precisely the kind of decision that we help you structure on d365 Guide.