I have been thinking quite a bit about the AI direction for Microsoft partners lately, as the landscape over the past year has been exploding.

The question for a Microsoft partner should not only be: which model is best this month? That will continue to change.

The better question is: where can we build repeatable customer value, reusable IP, recurring revenue and a delivery model that actually fits how enterprise customers buy, govern and operate technology?

That is why I still believe the Microsoft stack is the right place to build. Not because it is always the coolest answer, and not because every Copilot demo is convincing. They are not. But because the stack is already where many of our customers run their business. The users are there, the roles are there, the security model is there, the data is there, the integrations are there, and the commercial relationship is already there.

A smart AI demo outside the business platform is easy to like. I like many of them myself. But after the demo, someone still has to own the architecture, data access, security, lifecycle, support, integrations, cost control and process fit. This is where I think many customers will get into trouble. One department starts using Claude, another uses OpenAI, someone builds something in Vercel, someone connects a spreadsheet, and suddenly there are small agents and assistants everywhere.

For a while, it looks like innovation. After a while, it can start to look like AI anarchy.

Who owns it? Who supports it? Who knows what data it used? Who checks whether it fits the ERP process? Who explains it to IT? Who controls cost when usage grows? Who cleans it up when the person who built it leaves? And who decides whether this is a product, a prototype, a customization, or just another clever shortcut?

This is why I think we need to separate AI curiosity from AI strategy. We should absolutely test Claude, OpenAI, Google and whatever else comes next. We should learn fast, compare honestly, and use the best model when that is the right technical choice. But that does not mean we should build our partner strategy around every new tool that looks impressive.

For us, the more interesting opportunity is the agent direction around Dynamics 365 and the wider Microsoft platform.

If AI can help profile a D365 implementation, detect project risk earlier, support Success by Design reviews, monitor telemetry, prepare documentation, reduce manual follow-up and help customers operate D365 better after go-live, then we are no longer talking about a chatbot. Then we are talking about changing how ERP and CRM projects are delivered and operated.

AI does not remove the need for consultants. But it changes where the value is. Less time should be spent chasing information, rewriting the same documentation, checking the same lists, preparing the same status material and reacting too late. More time should be spent on architecture, business process understanding, decision support, customer trust and actually improving how the customer runs their business.

This is also where Microsoft has an advantage that is easy to underestimate. Not necessarily model by model, but platform by platform. Dynamics 365 gives the business context. Power Platform gives the extension and automation layer. Azure gives architecture, scale and consumption. Copilot Studio gives a way to package repeatable agent patterns. Teams, Outlook, SharePoint and Excel are where much of the daily work already happens. Entra and Purview help with security and governance through Work IQ.

My concern is that some Microsoft partners may spend too much energy chasing what looks most exciting right now, while someone else builds the reusable IP, owns the customer architecture, strengthens the Microsoft relationship and captures the recurring revenue. So my view is fairly simple. Use the best model when needed, but build the business on the platform where we already have customers, competence, trust and commercial leverage.

For us, that is still Microsoft we trust. And for a Dynamics 365 partner, that is where I think the real opportunity starts. In the future when the entire cost model is tokenized, a significant part of partner revenue will be based on the consumption of these services, and as far as I have seen, Microsoft model and partner ecosystem is what lasts through this rapid innovation phase.

Consider the following table — think about the areas business owners need to weigh.

Area	Anthropic	Microsoft Copilot Ecosystem
Developer productivity	Very strong	Strong and improving fast
Large-scale code understanding	Very strong	Strong, especially with GitHub and enterprise context
Agentic software development	Strong advantage today in some scenarios	Advancing rapidly
GitHub and DevOps integration	Strong	Market-leading
Enterprise governance	More limited and more dependent on architecture choices	Strong native advantage
Microsoft ecosystem integration	Limited	Native
Security, compliance & audit	Good, but narrower	Broad enterprise coverage
Enterprise-scale deployment	Emerging	Mature
Business process context	Must often be integrated separately	Native through D365, M365 and Power Platform
Commercial leverage for Microsoft partners	Possible, but less direct	Strong
Reusable IP potential for D365 partners	Possible, but more fragmented	Very strong
Recurring revenue potential	Depends on chosen architecture	Stronger alignment with consumption, services and platform adoption
Long-term customer ownership	Risk of fragmentation	Stronger platform ownership

Be smart and invest in the Microsoft platform. That’s how we as Microsoft partners will be able to continue to serve and bring long-term value. Also remember that Microsoft is multi-model so you get many of the same models. You just get it through Microsoft instead of directly to Anthropic, OpenAI and Google….. or Grok.

Microsoft is not the only place AI innovation happens, but it is where Microsoft partners are best positioned to turn AI into long-term customer value.

(Copilot Cowork assisted in writing this article, but it is still my words.)

And a small dinosaur joke: