IBM's Governed Context Strategy: Making Real-Time Data Work for Production AI Agents

As organizations deploy artificial intelligence agents to handle business decisions, those systems require access to information that stays current with actual operating conditions. Moving past consumer-focused productivity tools means building data foundations that can feed agents making real decisions across enterprise systems. IBM Corp. frames this challenge as a core production concern: the reasoning power of any large language model depends entirely on what information reaches it. According to Shaun Clowes, general manager of Confluent and Data at IBM, this is fundamentally about matching the right context to the right task.

You wouldn't give your accounts receivable work to a nuclear physicist that worked on your energy production side of your business. So, why would you take a very, very smart model, give it all of the possible information about your business and say, 'Figure out accounts receivable?' You need the right context for the job.

Shaun Clowes, general manager of Confluent and Data at IBM

Clowes shared these insights during an exclusive broadcast on theCUBE, SiliconANGLE Media's livestreaming studio, in conversation with theCUBE Research's Dave Vellante and John Furrier. The discussion covered real-time data delivery, agent specialization and governance in production environments, timed ahead of IBM TechXchange 2026.

Building governed context into the data flow

Traditional enterprise data architectures evolved to handle storage, computing and application needs that have shifted repeatedly over time. Introducing AI agents creates a new requirement: pulling together information from multiple operational systems—travel bookings, baggage tracking, rental cars—while that data remains actionable for immediate decisions. Clowes described the architectural shift this demands.

Now we need a new type of architecture, a type of architecture that can bring together all the state of our business right now and present that directly to LLMs. We call that the streamhouse architecture.

Shaun Clowes

IBM Confluent combines streaming, connectivity, processing, governance and querying functions to enable this model. Rather than forcing all data to move to a central location first, the platform allows data products to be prepared and governed where business information originates. This approach moves governance closer to the source, accelerating the production of high-value data and then connecting it to all systems that need it.

Our argument would be that you can actually move all of that very, very close to where the data is first produced, as close as you possibly can. The earlier you govern it, the better. The quicker you produce the high-value data, the better. Then you need to just connect up to all the relevant places.

Shaun Clowes

Giving specialized agents room to reason

Supplying current, governed data solves only part of the problem. Agents also require context tailored to their specific responsibilities. Business terminology often carries different meanings across departments, so preparing an agent's context means establishing the definitions relevant to its particular task.

You can describe, 'Hey, in this business process, what I mean is X.' Then, you can have these governed metrics and these governed datasets that effectively represent the foundation upon which the model will reason.

Shaun Clowes

Organizations can start by assigning agents to high-value work and refining individual agents as they accumulate operational experience. Once agents consistently execute established business processes reliably, the scope expands to questioning whether those processes themselves need restructuring. At that point, agents can begin proposing fundamental changes to how work is organized.

Then you can say to the supervising agent, 'Do I even need these processes? Should I just do something totally different?' The best bit about this whole system is it can learn at every layer.

Shaun Clowes

Making business context available in production

As agents assume greater responsibility, having access to current business information becomes increasingly critical. The shift from human-assisted decisions to fully autonomous operations fundamentally changes how enterprises must deliver data. Agents operating continuously require immediate access to operational information without waiting for a person to compile it.

People used to talk about data as the new oil, but what they really meant was [that] data, in some sense, existed and was used by humans to make good decisions. Businesses were run by humans … [and] that's how you win.

Shaun Clowes

Business data originates from scattered locations—oil rigs, cruise ships, remote facilities—and enterprises must integrate these sources into a coherent operational picture. No single centralized system can contain everything. Instead, organizations must build the ability to answer a fundamental question: what is the current state of my business right now?

https://www.youtube.com/embed/DWZbOcNRmIM?feature=oembed

You should always be asking yourself, 'How can I know the state of my business right now?' Because that is the world in which you make decisions. That is the world in which humans would make decisions, too.

Shaun Clowes