AI adoption continues to accelerate at a breakneck pace. But the actual returns remain uneven at best. Alexander Wurm, a principal analyst at Nucleus Research, believes solving this paradox ultimately comes down to one simple word: context.
Wurm was a featured speaker at the recent Boomi World conference, and he turned the usual stuck-in-pilot-purgatory AI narrative on its head. He said the “enterprise AI gap” isn’t a problem with the agents or models themselves. Instead, the real culprits are fragmented systems and manual connectivity.
Those hidden barriers are why AI projects stall, with no measurable ROI to show for the time, energy, and expense organizations invest. True AI maturity, Wurm explained, requires a connected architecture that delivers context-specific business data, giving agentic systems what they need to accurately reason and act.
It’s why, he added, the concepts of integration and context are inseparable.
“If you want to get value out of AI, you’ve got to figure out a way to provide the context, and the way you do that is to connect and integrate the systems,” Wurm told the audience. “What I try to get people to realize is that those are synonyms. Context and integration go hand-in-hand.”
Integration Is Having a Moment
Part of Wurm’s role at Nucleus Research is to keep his finger on the pulse of the integration platform as a service (iPaaS) category. He authored the “iPaaS Technology Value Matrix 2026“, which positions Boomi as a Leader for the sixth consecutive year.
IT organizations have always understood the need for connecting their digital architectures, Wurm said. At the same time, the iPaaS category has long had something of an image problem. Businesses sometimes struggle to recognize the value of integration because it’s hidden behind the scenes. So, they assign other technology categories a higher priority with their budgets. But as Wurm wrote in the market overview portion of the Value Matrix, “Agentic AI has introduced a new dimension” that has raised the stakes for integration.
Connectivity has been elevated to a fix-this-or-else issue because it’s the difference between AI success and failure. Wurm described it as the only way to provide the proper business context, spanning an untold number of data sources, from legacy green-screen systems to cutting-edge cloud applications – all with continuous, real-time accessibility.
“The question has always been, ‘Why are you buying an iPaaS?’” Wurm said. “It’s been a challenging space because with so many integration patterns, it’s often hard to nail down why you’re getting it. Before 2023, the idea of the connected enterprise was an esoteric goal. Now, AI has made the explanation a lot simpler. It’s much more tangible for CIOs and CFOs that agents can run everything. We’ve certainly had more interest in this Value Matrix because everyone’s moving to agent management as the new definition of iPaaS.”
The Expanding Capability of iPaaS
During his session, Wurm joked that back in the good ol’ days – like the fall of 2023 – businesses were wrestling with whether these new models and agents could really make decisions and do stuff autonomously. Those questions have been answered. Now, the pilot-to-production bottleneck has shifted. Wurm said businesses need to consider three key structural areas related to AI: context, tools, and the emerging “harness” concept for providing guardrails and governance.
Where exactly does iPaaS fit?
“This has always been the middle layer at the enterprise, connecting data, connecting applications, and powering workflows,” he said. “Well, iPaaS is very similar to an AI ecosystem. It connects the context, connects the tools, and connects the sub-agents across these different systems. It can also be a harness, an interesting new term for layered software around these models. An iPaaS is very good at serving as a portion of the harness that manages and orchestrates agents across your enterprise. It already has most of what you need.”
Wurm added that several product announcements at Boomi World, such as Boomi Orchestrate, add additional functionality to the platform, serving as that “harness” infrastructure and giving businesses accessibility, control, governance, and observability over agents.
He cited one example where a life sciences organization wanted to use OpenAI’s ChatGPT, but was concerned about security within their large enterprise. So, they stood up the AI model using Amazon Bedrock, and then relied on Boomi to integrate, pull data, and execute across different systems.
“Now, the organization can take information from across multiple systems and use AI to synthesize that into consistent reports that otherwise would have taken their teams multiple days to put together,” he added. “The important note here is not that ChatGPT was the value in this use case. It was the harness and the infrastructure around the model that created the value.”
Properly Measuring ROI From AI
A driving philosophy of Wurm’s analyst work is that the ROI of any solution going unused is zero. In these early days of AI, he noted that there is no shortage of dire reports and surveys citing poor ROI due to a broad range of issues, including security, change management issues, and cost.
But another wrinkle in calculating the ROI of agentic initiatives is that value cannot be attributed entirely to AI. Just like in his life sciences case, the entire architectural framework that supports AI plays a vital role in every project.
“A lot of ROI is completely outside of the model,” he said. “What people should recognize when it comes to enterprise AI is that we shouldn’t be throwing away software in favor of AI. Rather, we should be using software as that structured layer to control AI.”
That idea is also part of his take-home message for this year’s Value Matrix. Businesses should understand that agents and integration are inextricably intertwined. Connecting systems is the only way to ensure they operate within the proper context for decision-making and action.
“I can tell you for certain that we will be sitting here two or three years from now talking at length about what software is orchestrating your enterprise agents,” he added. “And iPaaS is definitely the prime candidate to be that orchestrator. Integration has never been more important, and it’s because context has never been more important.”
Learn more about how Nucleus Research views the integration category with this free copy of the iPaaS Technology Value Matrix 2026.