On November 10, IBM and Microsoft announced Imagine Enterprise, a platform to build, run and govern AI agents for industry-specific processes. The first frameworks target financial services, insurance and retail, with plans to extend to manufacturing, energy, public sector and mobility.
The split is clear. IBM brings its consulting arm and more than 6,000 prebuilt agents and assets from its Consulting Advantage platform. Microsoft brings the foundation: Azure, Foundry, Copilot Studio, Fabric and, for security and governance, Entra, Purview and Defender. Clients can have IBM run it, build with IBM, or manage it themselves.
Why it matters
It shows where the market is. A year ago the conversation was about which model to use. Today the big vendors sell business processes solved with agents, with governance and security included. For a bank or an insurer, that can shorten the path from proof of concept to production considerably.
It also confirms something I keep saying: in regulated industries, the differentiator is not the model but the identity, controls, observability and compliance around the agent. It is no accident that both companies put that at the center of the announcement.
What to assess before signing up
A platform with thousands of prebuilt agents is an attractive offer. That is exactly why a few questions are worth asking first:
Which process do I want to change, and how will I measure it? A large catalog invites you to start with what is available rather than what matters. The right question is which process has the most business impact and which metric will show whether it improved.
Where does my data live, and who touches it? A useful agent needs context: customer, product and operational data. You need to understand what data the platform will use, where it is processed and how existing permissions are enforced.
How does it integrate with my core systems? In banking, the value lies in connecting the agent to core banking, payments and customer management. That integration is usually the most expensive part and the one that decides whether the project scales.
How locked in will I be? Building on a single vendor’s stack makes the start easier but constrains future decisions. Know from day one which parts are portable (prompts, flows, evaluations, data) and which are not.
What does it cost to run at scale? The cost of an agent in production is not the cost of the demo. It depends on volume, on how many model calls each task makes and on how much context it carries. Estimate it per use case before committing.
My take
Imagine Enterprise is good news for the industry: it lowers the barrier to entry and puts AI governance front and center. But a platform does not replace a strategy. Organizations that first define what they want to achieve, with which data and under which controls, will get far more out of it than those that start from the catalog.
I help organizations define their AI strategy and assess platforms before committing to them. See services.
Sources: IBM announcement and Microsoft announcement, November 10, 2026.
