On 10 November 2026, IBM and Microsoft announced a joint product called Imagine Enterprise: a platform to help large companies build, run and govern AI agents that carry out real, industry-specific work. It is not a chatbot and not a single piece of software. It is a shared foundation on which banks, insurers and retailers can run fleets of AI workers that handle parts of their daily processes, under human supervision and inside the rules of their industry.
Why does this matter? For two years, companies have run AI pilots almost everywhere, yet very few have turned them into systems that work in production day after day. This article explains what the platform is, what each company contributes, and what is still unknown.
What "agentic AI" actually means
Most people know AI as a chatbot: you type a question, it types an answer. An AI agent is different. It plans a multi-step task, uses tools and company systems to carry it out, checks its own work, and hands the result to a human — or escalates when something is unclear. A claims agent, for example, might read a claim, pull the policy from the company database, check the customer's history, draft a response, and send it for approval. The human stays in charge, but the routine work is done by the agent.
An "agentic platform" is the layer underneath all of that: where agents are built, connected to company data, tested, monitored and governed. Without it, every agent is a one-off project wired up by hand, and nobody can say with confidence what the agent did or why. That is the problem Imagine Enterprise is aimed at, as AI interfaces become more capable and proactive — something we explored in our explainer on OpenAI's GPT-6 and its intelligent interface.
What IBM brings: 6,000 agents and the people who built them
IBM Consulting contributes the largest headline number in the announcement: more than 6,000 pre-built AI agents and assets, built on its AI-powered delivery platform, IBM Consulting Advantage. The wording deserves care. IBM counts agents and assets together, so this is not 6,000 fully finished, ready-to-switch-on AI workers. It is a library of agents, reusable components, templates and tools built up across client projects, plus the experience of building and running them.
Starting from that library is still far faster than starting from a blank page, and IBM says organisations can begin with capabilities already built for their industry and adapt them to their own business. IBM Consulting Advantage also includes three tools: Process Studio, which designs how an agent fits into a business process; Context Studio, which connects agents to the company's own data and knowledge; and Advantage Core, which deploys and governs the finished agents.
What Microsoft brings: the technology underneath
If IBM supplies the agents and the industry know-how, Microsoft supplies the ground they run on. Microsoft Foundry is where AI models and agents are built and managed. GitHub Copilot and Microsoft Copilot Studio coordinate agents with enterprise systems and the people who use them. Microsoft Fabric handles the data, Microsoft 365 is the productivity environment most of these companies already live in, and Microsoft Entra, Purview, Defender and Azure provide identity, governance, security and cloud infrastructure.
The governance part matters most to the platform's first customers. IBM says every agent action, decision and escalation is logged, with real-time data on latency, cost, token usage and run history, and that the platform is built with controls, observability and AI-compliance support. In a bank or an insurer, an agent that moves money or settles a claim must leave an audit trail a regulator can follow. The promise is that the trail is built in, not bolted on afterwards — a question becoming central as AI takes on more consequential work, including in security, where we recently covered how AI proved signals about encryption with mathematics.
Why two rivals are working together
IBM and Microsoft have competed for decades, in operating systems, cloud computing and enterprise software. So why join forces now? Each company has what the other lacks, and neither could easily build the whole thing alone.
Microsoft has the cloud, the AI tooling and the distribution — hundreds of millions of Microsoft 365 users and a huge Azure footprint — but no consulting army that can sit inside a bank for six months and rebuild its claims process around agents. IBM has that army and the industry expertise, but does not own the cloud platform or the productivity software where most enterprise work happens.
What "forward-deployed engineering" means
One phrase in the announcement deserves a plain explanation: "forward-deployed engineering support". It means engineers from both IBM and Microsoft are sent to work directly with a customer, on the customer's site or inside the customer's project, to help get the platform running in production. It is different from ordinary technical support, where you file a ticket and wait for an answer.
It matters because the gap between a demo and a deployment is where most AI projects fail. An agent that works in a controlled test can break when it meets a bank's real data, real approval rules and real legacy systems. Having the engineers who built the platform in the room while that happens is meant to close that gap faster. The announcement says this support is for "qualifying engagements" — the largest and most complex projects will get it first.
Which industries get it first, and why
Imagine Enterprise launches with industry frameworks for three sectors: financial services, insurance and retail. Manufacturing, energy, the public sector and mobility are named only as possible future additions, with no dates attached. The choice of the first three is deliberate.
Financial services, insurance and retail share three traits. They are full of process work — claims, onboarding, fraud checks, customer queries, stock and pricing decisions — that is repetitive enough for agents to help with, but important enough that mistakes are expensive. They are heavily regulated, so they need the governance and audit features the platform leads with. And they are big buyers of both IBM Consulting and Microsoft technology, so the two companies already have relationships and credibility there.
Retail is the most interesting of the three. Banks and insurers are the obvious regulated-industry customers, but retail shows the platform is broader than compliance-heavy work: a retailer might use agents to manage stock, answer customer questions, or coordinate promotions across hundreds of stores. AI models are gaining new senses and new ways of working with the world, as we explained in our guide to Tencent's HY4, an AI model built to see and understand the world.
SEI: the early example, with no results yet
The announcement names one early customer: SEI, a global provider of financial technology, operations and asset management services, which manages, advises or administers approximately $2 trillion in assets as of 30 September 2026. SEI is described as exploring emerging and agentic technologies with IBM Consulting, with Microsoft's support.
The wording is important. SEI is "exploring" — not deploying, not reporting results. Its chief AI strategist, Sneha Shah, said the company is focused on applying AI "with intention in ways that strengthen trust, enhance human decision-making, and create meaningful value". That is the language of an early engagement, not a finished rollout, and the announcement contains no quantified results: no cost savings, no time saved, no productivity figures.
What is still unknown
A fair reading of the announcement has to list what it does not say. No pricing was disclosed, so there is no way to judge what Imagine Enterprise costs or how it compares with building directly on Microsoft's own tools. No general-availability date was given, so nobody outside the first engagements knows when they could buy it. No client count and no revenue figures were shared. The platform is available by engaging IBM Consulting to deploy and operate it, by co-innovating on enterprise-scale implementations, or through self-management options — but the details of those options, and what each costs, are not public.
The announcement also ends with a standard caution that statements about future direction "are subject to change or withdrawal without notice, and represent goals and objectives only". That means the future industries — manufacturing, energy, the public sector, mobility — are ambitions, not commitments. A showcase is expected at Microsoft Ignite, from 17 to 20 November 2026.
A very busy November for enterprise AI
Imagine Enterprise did not arrive in a vacuum. The most striking parallel came from Cognizant, which has built three separate delivery groups around the three big AI model makers: a Gemini group of more than 18,000 professionals announced on 19 October, an OpenAI group of more than 25,000 announced on 30 October, and a Claude group of more than 25,000 announced on 10 November — the same day as Imagine Enterprise. Together, the three groups hold more than 68,000 people, nearly a fifth of Cognizant's workforce, each assigned to only one group at a time.
The data side of the industry moved on the same day too. Commvault and Databricks announced Data Activate for Databricks, a way to share governed enterprise data with analytics and AI applications without rebuilding connections to every source. Denodo signed a multi-year strategic collaboration agreement with AWS to give agents live, governed data through a unified AI data layer. And GoodData.AI launched its Agentic Serving Plane, a governed execution layer between enterprise data platforms and the agents that consume them. The shared idea is simple: agents are only as good as the data they can reach, and enterprise data is still locked in backups, warehouses and silos.
What to watch next
Three things will tell us whether Imagine Enterprise lives up to its announcement. The first is Microsoft Ignite, from 17 to 20 November, where the platform is expected to be showcased — watch for a live demonstration on real industry data, not just slides. The second is a named customer deployment with measured results. The third is pricing, or at least a clear statement of how the platform is sold, because until buyers can compare it with the alternatives, they cannot make a rational decision.
For now, the sensible way to think about Imagine Enterprise is as a serious, well-resourced attempt to solve the hardest problem in enterprise AI: getting agents out of the pilot stage and into production, inside the rules of regulated industries. IBM brings the agents and the industry knowledge. Microsoft brings the technology, the cloud and the security. What neither company has yet shown is the evidence that it works.