IBM and Microsoft's 6,000 agents: what the release does not say

IBM and Microsoft announced Imagine Enterprise on November 10. It is an agentic platform, meant to let a company create AI agents, run them and keep them governed while they work through processes specific to its industry. The headline travelling with it is 6,000 agents. The number is real, but it does not mean what most of the coverage implies.

Read the release and the same number arrives in three different shapes. The opening bullet point promises "more than 6,000 pre-built AI agents and assets", available from day one. Lower down, the same total becomes a fleet that IBM says it has built on top of its Consulting Advantage delivery platform. A third bullet reduces it to IBM Consulting's "hands-on experience building and running more than 6,000 AI agents", which is a track record rather than a delivery.

Those are not the same claim. Pre-built assets you can adopt, a fleet somebody else operates, and experience accumulated across past projects imply very different things about what arrives in your tenancy. No catalogue has been published, no count is given per industry, and nothing says how many of the 6,000 are agents as opposed to assets. A buyer cannot yet tell what they receive.

There is no price because there is no product to price

There are three ways to buy it, per IBM. You can hand the platform to IBM Consulting to stand up and operate for you. You can build it jointly with IBM Consulting at enterprise scale. Or you can take a self-managed option. Both companies will put forward-deployed engineers on engagements that qualify, a bar neither has defined.

No price, no unit of pricing and no minimum engagement size appear anywhere in the announcement. That is consistent with what this is. Two of the three routes are consulting engagements, and consulting is quoted, not listed. It does mean the usual comparison, cost per seat or per agent against an alternative, cannot be run yet.

It is a Microsoft stack decision

The component list is specific. Microsoft Foundry, GitHub Copilot and Copilot Studio coordinate the agents and the systems they reach. Microsoft Fabric, Microsoft 365 infrastructure, Entra, Purview, Defender and Azure are grouped together as the layer supplying, in IBM's words, the "security, and cloud foundation" along with data, identity and AI governance. IBM Consulting adds its own delivery platform, IBM Consulting Advantage, and names Process Studio, Context Studio and Advantage Core as the pieces used to design and govern the agents.

For a company already standardised on Entra and Purview, that is a shortcut. For a company running identity or data governance elsewhere, it is a migration with an agent platform attached, and the governance layer is the part that will not port.

The industry frameworks available at launch cover three sectors: financial services, insurance and retail. Manufacturing, energy, the public sector and mobility are named as candidates to follow, with no dates attached.

The customer is exploring, not running

SEI, a provider of financial technology, operations and asset-management services that says it manages, advises or administers about $2 trillion in assets, is the only client named. IBM does not say SEI is using the platform. It says SEI is pursuing its own transformation by exploring agentic and emerging technology with IBM Consulting, backed by Microsoft. Sneha Shah, SEI's Chief AI Strategist and Head of SEI Next, said the company's focus is on applying AI "with intention", and described the work as exploring how the capabilities could support critical processes and scale responsibly over time.

Exploring is not deployed. No source published this week describes any organisation running Imagine Enterprise in production, and neither company disclosed commercial terms.

The executive quotes set the register. Mohamad Ali, Senior Vice President and Head of IBM Consulting, says what clients want is to become "an agentic business". Rodrigo Kede Lima, President of Microsoft Frontier Company, argues that enterprise AI will go to the organisations able to convert their own data and industry knowledge into action.

What IT teams should do

• Ask for the agent catalogue for your industry, split into agents and assets, with a count of what is reusable as-is against what is a starting template. The 6,000 figure is not an answer to that question.
• Get pricing in the shape you buy in, whether that is a fixed-scope engagement, a run-rate or a per-process fee, and ask what happens to the price after the first year.
• Establish who operates the agents in production and who is accountable when one acts wrongly. Two of the three routes leave IBM Consulting running your processes.
• Price the Microsoft dependencies separately. If Entra, Purview, Fabric or Defender are not already in place, that is a second project with its own licensing.
• Ask for one reference running it in production. If the answer is still SEI exploring, you are an early customer, and the contract should price that.

Sources: IBM Newsroom, Investing.com, StockTitan, Dataconomy.