If your bank, insurer or retailer gets pitched Imagine Enterprise this quarter, the offer is an IBM Consulting engagement built on Microsoft software you license separately, and neither company has published what it costs. Evaluate it the way you would evaluate a systems integrator's bid: get the Microsoft licence assumptions, the ownership of the agents IBM adapts for you, and the exit terms in writing before you compare it with building on Foundry and Copilot Studio yourself.
IBM and Microsoft announced Imagine Enterprise on November 10, 2026, in matching releases from Microsoft and IBM. They describe it as a platform for building, running and governing AI agents that carry out industry processes, launching with frameworks for financial services, insurance and retail. Neither release gives a price, a general availability date or a list of regions, a gap MLQ.ai also noted.
What Imagine Enterprise Actually Contains
Imagine Enterprise is IBM's consulting delivery platform layered on top of Microsoft's AI, data and security products. The IBM release splits it into two halves:
- The Microsoft half: Microsoft Foundry, GitHub Copilot and Copilot Studio for building and coordinating agents, plus Fabric, Microsoft 365 infrastructure, Entra, Purview, Defender and Azure for data, identity, governance, security and cloud.
- The IBM half: IBM Consulting Advantage, including Process Studio, Context Studio and Advantage Core, plus "more than 6,000" prebuilt agents and assets from IBM Consulting.
IBM Consulting Advantage is IBM's own AI delivery platform for its consultants. IBM said in January that it had supported more than 150 client engagements. Two of the named components were not finished products as of IBM's May update: that Think 2026 release called Context Studio "now available" and Process Studio "coming soon." The November releases do not say whether Process Studio has shipped.
The 6,000 figure counts agents and assets together, and neither company breaks it down by industry or type. Runtime Wire framed the open question as whether IBM's prebuilt assets "cut deployment time enough to justify buying the bundle" compared with buying the Microsoft services and the implementation help separately. Nobody has published data on that yet. The one named customer, SEI, is described as "exploring" agentic technology with IBM Consulting, which is a long way short of a production reference.
How You Buy It Decides What You Own
The releases offer three ways in, and each one moves ownership and cost in a different direction. Per Microsoft's announcement:
- IBM-managed: IBM Consulting deploys and operates the platform for you.
- Co-innovation: you build enterprise-scale implementations alongside IBM Consulting.
- Self-managed: you run the platform yourself.
"Self-managed" is the option to question hardest. Consulting Advantage was built as IBM's internal delivery tool, so ask what you actually receive under that model: software you license and keep running after the engagement ends, or access that lapses when the statement of work does. The releases do not say.
Forward-deployed engineers come from both companies, but only "for qualifying engagements," and neither release defines what qualifies. Microsoft's side of that is its Frontier Company unit, which Microsoft funded with $2.5 billion and 6,000 forward-deployed engineers in July under Rodrigo Kede Lima, the executive quoted in this launch. We covered that unit when it launched in our piece on Microsoft's 6,000 engineers. Ask in writing whether your deal qualifies, how many engineers you get, for how long, and who pays for them.
The Licence Bill Sits Outside the Bundle
Every product named on the Microsoft side carries its own meter, and nothing in either release says Imagine Enterprise includes those licences. Assume it does not until a quote says otherwise.
The published rates give you a floor. Microsoft's Copilot Studio pricing page lists a capacity pack at $200 a month for 25,000 Copilot Credits, a pay-as-you-go option that requires an Azure subscription, and Microsoft 365 Copilot at $30 per user per month paid yearly (prices checked November 12, 2026). Foundry's agent runtime has its own billing, which we broke down in our Foundry Copilot harness piece, and Purview, Defender and Fabric each price separately.
That matters for your negotiating position. If you already hold an Azure commitment, Foundry and Copilot Studio consumption can draw it down, which is the logic we walked through in Bedrock vs Vertex AI vs Foundry. IBM's fee is a separate line on top of that. Ask IBM for a quote that lists every Microsoft SKU the design assumes, at what volume, and which side of the table each one bills to.
IBM's own commercial model points to where the services fee will land. HFS Research, in a May 2026 analysis of IBM Consulting's strategy, reported that IBM Consulting is moving toward "outcome-based contracts with 15% gain-share of contract value" and "pod-based pricing." That is an analyst's summary of IBM's direction, not an IBM price sheet, but it tells you to expect outcome language in the proposal. Outcome pricing only protects you if you define the baseline and the measurement before signing.
Who Owns the Agents IBM Adapts for You
Neither release says who owns an agent once IBM adapts one of its prebuilt assets to your processes and data. For a bank or insurer this is the clause that matters most, because the adapted agent encodes your underwriting rules, your claims logic or your pricing.
There are three positions you might see in a draft contract. IBM keeps the base asset and licenses the adapted version to you for the term. You own the adaptations and IBM keeps the base, with a perpetual licence to run it. Or the language is silent and IBM's standard services terms apply by default. Push for the second, and get the licence term to survive the end of the engagement.
Microsoft's stated data stance when it launched Frontier Company was that it will not use proprietary customer data to train its models "in ways that commoditize what differentiates them in their industry," per CX Network's report. That is narrower than a flat no-training commitment. Ask both companies for a plain clause that your data, prompts and adapted agents are not used to train or improve anything offered to other clients, with no qualifier.
The Case for Buying the Bundle
The strongest argument for Imagine Enterprise is that both companies are selling engineers whose stated job is moving customers "from experimentation to production," per Microsoft's release, on a stack your security team may already have approved. If you are already on Microsoft 365, Entra and Purview, the identity and governance layer is in place before the first agent ships. Our Copilot Studio agent accounts piece covers how those Entra agent identities behave.
The argument against is lock-in on two fronts at once. In January IBM pitched its Enterprise Advantage service as working across AWS, Google Cloud, Azure and watsonx without forcing clients to change cloud providers. Every component named in Imagine Enterprise comes from IBM or Microsoft. If you buy this, your agent runtime is Microsoft's and your delivery tooling is IBM's, and leaving either means rebuilding. We made the same point about exit cost in our Palantir AIP alternatives guide: score the exit before you score the features.
If you have your own integrator, or engineers who already build on Foundry, you can buy the same Microsoft services directly and hire implementation help on a normal SOW. The bundle is worth paying for only if IBM can show you a shorter path to production in your specific process, with numbers from a client that is past the exploring stage.
What to Do Before You Sign
Treat the first Imagine Enterprise proposal as a services bid and run it through your normal integrator evaluation, with four extra questions.
This Week:
- Ask your IBM and Microsoft account teams for the list of prebuilt agents and assets that apply to your industry and process, broken out from the 6,000 total. If they cannot name the ones for your claims or onboarding workflow, the number does not help you.
- Ask whether Process Studio is generally available today, and get the answer in writing.
This Month:
- Get a quote that lists every Microsoft SKU the design assumes, its volume, and who bills it. Price the same design built directly on Foundry and Copilot Studio with your current integrator, so you can see what IBM's fee buys.
- Put the "qualifying engagement" question to both companies: how many forward-deployed engineers, for how long, and at whose cost.
- Ask for one reference client in your industry that is running an Imagine Enterprise agent in production, not exploring.
Before Signature:
- Write the ownership clause: you own adaptations of IBM assets, with a perpetual licence to the base asset that survives termination.
- Define what "self-managed" delivers on exit: which code, configurations, prompts and evaluation sets you keep, in what format, and whether any IBM tooling stops working when the SOW ends.
- If the price is outcome-based, fix the baseline metric, the measurement method and the dispute process before signature. The guidance in our agent orchestration selection framework on exit criteria applies here too.
The Bottom Line
Imagine Enterprise is the latest version of a deal enterprises have signed many times: a consulting firm's accelerators on a hyperscaler's platform, sold as a product. The forward-deployed model behind it is now standard: Google, AWS, Microsoft and IBM have all started programs, MLQ.ai noted citing CIO Dive, and our coverage of EXL's Aimpoint Digital deal and Anthropic's Claude Frontier Academy tracks the same shift among integrators and labs. What decides whether it pays off is the contract: the licences it assumes, the ownership of what IBM builds on your data, and what you keep when the engagement ends. None of those terms are public. Get each one in writing before you compare it with building on Foundry yourself.
Continue Reading
- Why Microsoft Is Paying 6,000 Engineers to Fix Your AI
- Microsoft Starts Billing for Copilot Harness on December 1
- Copilot Studio Agents Keep Shared Memory, No User Opt-Out
- Pick Databricks Over Palantir AIP if Your Team Can Build
- EXL Acquires Aimpoint Digital; Ask About Staffing in Your Next SOW
- Agent Orchestration Platforms: Score Exit, Not Features