Enterprise AI Winners Focus on 'Boring' Automation, Not Just Models

On July 19, 2026, Motley Fool contributor Rachel Warren interviewed Adam Field, Chief AI Officer at Tungsten Automation, to discuss why enterprise AI transformation often fails and which companies are actually winning. Tungsten serves over 25,000 global organizations, including 40% of the Fortune 100, giving Field a unique vantage point on what separates successful AI deployments from expensive failures. The core of Field's argument is that most enterprises misunderstand AI's real value. Many treat advanced AI tools as magic solutions—handing teams the most powerful technology available and expecting automatic business transformation. As Field put it, "It's like handing someone the best camera and calling them a photographer. We would never do that." In reality, the difference between old-school automation (like robotic process automation, or RPA, which follows deterministic rules) and AI-driven workflow engines is fundamental. Traditional automation required humans to specify exactly what to do; new AI agents can accept a task, an output goal, and the available tools, then figure out how to accomplish the goal. However, this capability alone does not guarantee business success.

Field explained that the real competitive moat in enterprise AI is not owning the latest foundational model. With models from OpenAI, Anthropic, and open-source options widely accessible, the model itself has become a commodity. For 95% of Tungsten's daily work—processing invoices, extracting data from contracts, and handling document-heavy workloads—the specific model is nearly irrelevant. Processing an invoice does not require the most advanced model; it does not need to consume massive token budgets. What separates winners from budget-burners is the infrastructure, industry-specific logic, and data quality built on top of the model. This is where the real value lives. Field introduced the concept of "dark data"—information trapped inside enterprises that neither machines nor humans are currently extracting value from. Analysts estimate that 80% or more of organizational information is dark data; within that, approximately 80% is trapped inside unstructured documents: contracts, annual reports, emails, invoices, and call-center transcripts. These documents look different every time, making them unsuitable for traditional machine learning, which has historically excelled at structured, consistent data. Generative AI and large language models are now capable of reading these highly unstructured, multi-hundred page documents, breaking them into usable pieces, and combining that information with structured data to enable better decision-making. For a bank evaluating risk, an insurance carrier assessing claims, or a corporation managing supplier relationships, access to this dark data can radically improve outcomes and capital efficiency. Field emphasized that the right strategy is not to bet everything on the latest generative AI but to use the right AI for the right job. Traditional machine learning is often faster, more cost-effective, and more environmentally friendly than large language models, since it does not require massive GPU resources. The companies that are winning are those that can mix traditional machine learning and generative AI fluidly, choosing the best tool for each problem on the fly. This nuanced, infrastructure-heavy approach stands in sharp contrast to the flashy narrative around AI: huge capital expenditures on chips, big bets on foundational models, and the race for the most powerful model. The true cash cow for enterprise software businesses is the unflashy layer beneath—the automation engines that turn dark data into actionable revenue.

The enterprise AI landscape has shifted away from the chip manufacturers and foundational model companies that dominate headlines toward the infrastructure and automation layer that actually delivers value to the Fortune 100. As Field outlined, there are three layers: chip manufacturers (Nvidia, AMD), model companies (Anthropic, OpenAI, and open-source alternatives), and the application layer. Most investment focus has landed on the first two, but Field argues that the competitive advantage now lies in the third—specifically, in companies that can unlock the vast amounts of unstructured data trapped within enterprises' document systems. The distinction matters because access to foundational models is now democratized; nearly every competitor can use the same Claude, GPT, or open-source model. What separates winners from money-burners is not the model choice but the engineering, domain expertise, and data strategy layered on top of it. This reframing challenges the narrative that AI adoption is primarily a capital-expenditure story about chips and models. Instead, it suggests that returns on enterprise AI investment depend heavily on unglamorous but critical work: building automation engines that can parse contracts, emails, invoices, and call transcripts; integrating those engines into existing workflows; and converting that "dark data" into actionable business decisions. Field's emphasis on "boring AI"—traditional machine learning combined thoughtfully with newer generative models—suggests that the most valuable enterprise AI plays may be companies that solve specific, high-friction problems in document automation and data extraction rather than those chasing the latest, flashiest model capabilities.

No comments yet. Be the first to share your thoughts!