Enterprise AI adoption is becoming universal, but value extraction from investment remains stubbornly weighed down by a type of “cognitive debt” as AI-generated systems quietly outpace the organizations meant to oversee them.

Research shows the gap between AI enthusiasm and AI outcomes is widening. A new study from Harvard Business Review Analytic Services, sponsored by Appian Corp., found that while 59% of organizations have AI in production, most are focused on incremental efficiency gains rather than top-line impact. The real obstacle for large enterprises deploying AI in mission-critical workflows is the cognitive debt created when AI-generated systems grow faster than teams can govern them, according to Matt Calkins (pictured), co-founder and chief executive officer of Appian.

“We need to find a way to make AI connect to the most important work in the world,” Calkins said. “This is the year where AI hits the enterprise, but it doesn’t happen automatically, we’ve got to make it happen. We have to reassure people that AI can be safe enough to be used in most enterprise cases and do the most valuable work. We’re achieving that — with guardrails, with structure, with process.”

Calkins spoke with theCUBE’s Dave Vellante and Allison Kosik at Appian World 2026, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed how process orchestration closes the enterprise AI reliability gap, the open-source parallel for AI economics and Appian’s role in the emerging AI stack. (* Disclosure below.)

Cognitive debt and the process layer enterprises are missing

The HBR survey Appian commissioned found that while the majority of leaders know they need rules-based guardrails for AI agents, yet fewer than half have actually defined them — a disconnect tied directly to cognitive debt accumulating at the point of deployment, Calkins noted. Appian’s vibe coding analysis frames the problem plainly: AI can make software creation feel effortless while quietly leaving teams with systems they cannot fully explain, audit or govern. There is a clear gap between awareness and action, but the situation is not necessarily intractable.

“They know what they need. They just haven’t done it yet,” Calkins said. “92% say they know they need rules-based guardrails, but most of them haven’t done it yet. It’s not hopeless. You can look across the economy and realize that people know what they need to make AI reliable.”

The current situation has an economic parallel to open source. Code becoming free did not collapse the software industry, but instead forced software to specialize at a higher level of trust and reliability, according to Calkins. The same logic applies to AI-generated code today, where unstructured AI-generated output too often produces throwaway, not-for-production software, a particular liability in regulated industries where error tolerance is near zero.

“Code may be cheap, but mistakes are expensive,” Calkins said. “When people tally up the reasons why it’s expensive to build an application, these days they’re already thinking more about whether it’s reliable than whether the code itself was expensive.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of Appian World 2026:

(* Disclosure: TheCUBE is a paid media partner for Appian World. Neither Appian, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE