Enterprise AI Value Relies on Operational Workflows Over Model Upgrades, Domino Data Lab CEO Says

Enterprise artificial intelligence deployments are failing to deliver clear financial returns for a majority of organizations because companies remain focused on individual productivity tools rather than core business operations. In an interview with tech media platform theCUBE, first reported by SiliconANGLE, Domino Data Lab Inc. Chief Executive Officer Thomas Robinson noted that 57% of businesses currently report that the financial yields of their AI projects lag behind total expenditure.

According to Robinson, the disconnect stems from how enterprise IT departments initially deployed generative tools over recent years. Organizations prioritized personal desktop assistants designed to streamline tasks such as writing email drafts or generating marketing content, rather than embedding models directly into high-stakes, revenue-generating workflows like financial modeling, national security, or pharmaceutical discovery.

Robinson argued that focusing solely on desktop efficiency limits the total financial upside of technology investments. While organizations often measure success by tracking seat licenses, token consumption, or raw adoption metrics, these metrics record expenditure rather than generated revenue or long-term value.

Framing artificial intelligence initiatives purely as cost-cutting programs caps potential gains at 100% of existing costs, Robinson told theCUBE host Dave Vellante. High-performing organizations instead evaluate AI investments based on their capacity to fuel revenue growth and drive product development over a one-to-five-year timeframe.

Beyond deployment strategies, governance gaps continue to hinder operational scaling. Robinson cited data indicating that roughly 41% of organizations are currently running pilots or expanding agentic AI systems without establishing adequate operational guardrails to manage autonomous behavior.

To address those risks, Domino Data Lab advocates for a structured control architecture built around three core pillars: an upfront policy engine to gate projects from inception to deployment, continuous tracing and monitoring of live production environments, and final oversight by human managers accountable for outcomes. Robinson emphasized that human evaluation remains necessary, noting that current AI models have not reached reliable reasoning capabilities.

As a result of these structural hurdles, market value across the enterprise software sector is transitioning away from foundation model developers and toward integration software, specialized governance tools, and workflow management. Foundation model creators themselves are increasingly building out internal forward-deployed engineering divisions to assist clients with complex integration projects, signaling that technical implementation remains the primary barrier to realization of business value.