Enterprise AI capabilities are improving almost everywhere, yet the returns still trail the spending. The technology is reaching production, but it often stops short of the business process where revenue, innovation and risk actually live — a gap that is now reshaping how enterprises measure AI success.
That gap is widest in industries where a wrong answer carries real consequences, such as pharmaceutical research, financial services and national security. Much of the past few years of adoption has been aimed at individual desktop productivity rather than core operations, and 57% of organizations still struggle to generate returns that outpace their spending, according to Thomas Robinson (pictured), chief executive officer of Domino Data Lab Inc.
“A lot of the application we’ve seen over the past few years has been about putting the tools in an end-user compute context, so giving users a desktop tool to draft emails or make marketing copy,” Robinson told theCUBE. “That’s not where companies make and lose their revenue. It’s not where they do their innovation, and it’s not where they drive cost cutting. At the end of the day, we think the AI needs to be applied in the high consequence, mission critical, core processes of the business.”
Robinson spoke with Dave Vellante during theCUBE’s coverage of the “Domino In The New Era of AI” exclusive interview. TheCUBE, SiliconANGLE Media’s livestreaming studio, explored the last-mile gap between AI in production and measurable business outcomes. (* Disclosure below.)
Rewiring enterprise AI around business processes
Part of the problem is measurement. Counting seats, tokens or raw adoption tells leaders how much they will spend, not what they earned, and framing projects purely as cost-reduction exercises caps the upside before work even begins, Robinson explained.
“I’m very fond of saying the maximum you’re going to save is 100% of your cost,” he said. “The leaders in our customer base are the ones who are thinking about what can AI do for their future innovation, the development of new products that’s gonna strike a year to five years down the road. They’re thinking about how to drive increased revenue with AI.”
Trust is the other constraint, particularly as agentic systems scale ahead of guardrails. About 41% of organizations are piloting or scaling agentic AI without the governance to manage it, Robinson noted. Domino Data’s answer is a three-layer model: a policy engine that gates projects from initiation through delivery, continuous monitoring and tracing in production, and a human accountable for the outcome.
“We still believe that judgment is better than hallucination,” he said. “Despite many of the models being called reasoning models, we don’t think it’s there yet.”
That is also why Robinson expects enterprise AI value to migrate away from the models themselves and toward integration, governance and workflow. Model providers building their own forward-deployed engineering arms offer evidence of that shift.
“They are not marketing that the gap is made up by better AI. They are building companies filled with people to go implement it,” he said. “It underpins a story of roll up your sleeves, get smart, capable people into your business to help drive digital transformation and use the new technology.”
Here’s the complete video, part of SiliconANGLE’s and theCUBE’s coverage of the “Domino in the New Era of AI” interview:
(* Disclosure: Domino Data Lab sponsored this segment of theCUBE. Neither Domino Data Lab nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)