IBM has used its latest earnings call to detail how the company does not want to win the AI race by building the biggest model.
Instead, it is betting that enterprises will care more about what happens after the models arrive, including how agents are governed, where they run, and whether they can access the data needed to do something useful.
That was the clearest message from IBM’s Q2 2026 earnings call. And while the company did not announce a CCaaS deal, customer service deployment, or new CRM product, its strategy has a clear CX implication.
Indeed, during the call, Arvind Krishna, CEO, President, and Chairman of IBM, said:
“Value will increasingly shift towards the orchestration and data layers so that clients can optimize outcomes, cost and governance across multiple models and agents and keep control of their proprietary data."
Contact center, CRM, and customer experience leaders know better than most that AI agents can only be as useful as the customer, operational, and product data they can access.
If that information is stale, fragmented, or poorly governed, the result is often a fast answer that is wrong, irrelevant, or impossible to explain.
IBM Wants to Be the Control Layer for Enterprise AI Agents
IBM positioned watsonx Orchestrate as the “control plane” for enterprise AI agents.
According to Krishna, the platform helps customers “build, manage and govern agents” across multiple models, cloud environments, and on-premises infrastructure. It is also designed to provide observability, evaluation, governance, identity management, and security.
While this may be a fairly vague claim, it does speak directly to a problem many enterprises are now encountering as AI pilots progress beyond a single chatbot or internal copilot.
Early deployments can be relatively contained: a virtual agent answers frequently asked questions, an agent-assist tool summarizes calls, and a sales copilot drafts follow-up emails.
Yet, as organizations introduce more agents across service, sales, IT, and marketing, the task becomes less about deploying a single AI feature and more about managing a growing network of automated decision-makers.
Which agent has access to which data? Which model is being used? How are outputs evaluated? Can the business identify why an agent took a particular action? And can it stop that agent from exposing sensitive customer information?
IBM is aiming to answer those questions with watsonx Orchestrate. Krishna argued that the vendor’s differentiation lies in its “neutrality and enterprise-grade operational control,” rather than locking customers into a single model or cloud ecosystem.
That could appeal to large, regulated organizations, particularly those trying to deploy AI across legacy systems, private environments, and public cloud platforms.
Real-Time Data Is the Other Half of the Equation
However, governing an AI agent is only half the battle. The agent also needs context.
