Prebuilt AI agents became a catalogue business on November 10, 2026. IBM and Microsoft announced Imagine Enterprise, a platform to build, run and govern AI agents that execute industry-specific processes, starting with more than 6,000 prebuilt AI agents and assets from IBM Consulting Advantage, IBM Consulting’s AI-powered delivery platform. The offer removes the blank page. It does not remove the work that decides whether an agent acts on the right company fact, in the right version, under the right approval.

What 6,000 prebuilt AI agents actually contain

The headline number needs one correction before it travels. The 6,000 figure includes agents and other assets, rather than exclusively deployable agents. Microsoft frames the launch around three advantages: Instant Expertise, starting with prebuilt agents and industry expertise from day one; Fits the Enterprise, a platform designed to connect with existing data, processes and systems; and Intelligence Multiplier, putting the knowledge of the business behind people and AI agents.

The stack is a division of labour. IBM Consulting Advantage contributes Process Studio, Context Studio and Advantage Core to design, deploy and govern tailored agents. Microsoft Foundry, GitHub Copilot and Microsoft Copilot Studio coordinate agents and enterprise systems, while Microsoft Fabric, Microsoft 365, Microsoft Entra, Microsoft Purview, Microsoft Defender and Microsoft Azure provide the data, identity, governance, security and cloud foundation. Industry frameworks cover financial services, insurance and retail first, with manufacturing, energy, public sector and mobility named as a future blueprint. Forward-deployed engineers from both companies support qualifying engagements, and clients can have IBM deploy and operate the platform, co-develop it with IBM, or manage it themselves. The announcement gives no standard pricing and no general availability date.

The one named customer is a study in careful wording. SEI, which manages, advises or administers approximately $2 trillion in assets as of September 30, 2026, is exploring emerging and agentic technologies with IBM Consulting and Microsoft support. Its role is exploratory, without quantified production results for Imagine Enterprise. On governance, IBM’s launch overview states that every agent action, decision and escalation is logged, with real-time data on latency, cost, token usage and run history, and that governance controls set which actions an agent may take autonomously, which require human approval and when escalation is mandatory.

Read-only is the current setting for prebuilt AI agents

The catalogue arrives in a market that still rations authority. PYMNTS Intelligence, in its November report The Agentic Edge Index, found agentic AI in use at 89 percent of large United States enterprises, while 78 percent would restrict agents to read-only access to company directories, 18 percent would allow agents to perform tasks with human approval, and 2 percent would permit independent execution. The same report found 51 percent of executives consider organizational barriers, including staffing, business processes and corporate culture, greater obstacles to adoption than technology. The workforce side lags further: only 4 percent of CIOs have implemented plans to adjust roles for a human and agent workforce, according to Gartner figures reported by CIO Dive.

IBM’s own chief executive stated the constraint on the October earnings call. Organizations are recognizing that AI is only as effective as the system around it, Arvind Krishna told investors, naming the data that provides context as part of that system. Mohamad Ali, Senior Vice President and Head of IBM Consulting, put the demand side in the launch release: clients want to build an agentic business, not a faster version of the same work. Both statements point at the same gap. A prebuilt agent carries an industry process. It does not carry a company’s current policy, its exception history, or the decision that superseded last quarter’s rule.

The Adaptation Test for a prebuilt agent

A catalogue agent earns production authority through adaptation, and adaptation is a record problem before it is a model problem. Three checks settle it.

First, the local fact check. For each fact an adapted agent reads, name the source, the date, and the version it supersedes. An industry template is a starting draft. The company record is the authority, and a correction to that record must reach the agent, its derived summaries and any queued actions. This is the machinery set out in How Jilu Works and in the Context Engineering pillar.

Second, the bound check. Record, before production, which actions the agent may take alone, which need a person’s approval, and which must escalate, function by function. The fragmented enterprise knowledge evidence shows why the bound stays narrow where the record is weak, and the operating knowledge cases show what changes when an agent stops answering and starts acting: a wrong answer costs a correction, and a wrong action changes a payment, an order or a claim.

Third, the receipt check. Save the log of what the agent did, on which facts, and under whose delegation, in a form a person can inspect and reverse. A platform log that records latency and token cost is an operating receipt. An adaptation receipt also names the facts used, so a later correction can find every action that rested on the old version.

The caveats belong in the record. Imagine Enterprise is a launch announcement, not a deployment result: no pricing, no availability date, a headline count that mixes agents with assets, and a named customer whose engagement is exploratory. None of that makes the catalogue small. It makes the next question precise.

A catalogue scales the supply of agents. A record scales the trust placed in them. The companies that gain from 6,000 prebuilt agents will be the ones that treat each adaptation as a governed change to their own record, with source, date and superseded version attached, so that when the underlying fact changes, the agent changes with it and the receipt shows where the old fact acted.