As companies race to adopt generative AI, autonomous agents, and a new generation of data tools, leaders face a harder question than whether to embrace the technology at all: how to turn it into measurable value without creating fresh costs, risks, and organizational confusion.

That challenge framed the Fortune AIQ Summit, held at the New York Stock Exchange, where executives from the technology, finance, industrial operations, travel, and human resources sectors examined what separates genuine AI transformation from scattered pilot projects. A consistent pattern emerged: the strongest companies are not simply handing out AI tools. They are redesigning workflows, developing talent, strengthening governance, and building the data and operating discipline required to scale responsibly. The gathering reflected a broader moment for enterprise technology: leadership teams are under pressure to show returns on AI commitments that have moved past small-scale trials, a tension that ran through much of the discussion.

Discussions at the summit centered on five key trends:

• Moving from experimentation to enterprise transformation
• Building an AI-ready workforce
• Managing governance, trust, and risk
• Applying AI in physical and regulated environments
• Measuring ROI, controlling costs, and creating growth

From AI Adoption to Transformation

The most important AI question is not whether a company has deployed tools, but whether it has changed how work gets done.

Speakers repeatedly cautioned that broad access to AI does not automatically translate into productivity, business value, or organizational maturity.

Diana David, director of futures at ServiceNow, said the leading companies set themselves apart through their commitment to people and operational discipline rather than technology alone. ServiceNow's Enterprise AI Maturity Index found that 57% of "pacesetters" invest in ongoing AI upskilling, compared with just 4% of other organizations. The starkest divide involves attracting, hiring, and retaining AI talent: 68% of pacesetters do so, versus 10% of other companies.

"Talent investment. That was the biggest gap," David said.

She warned against treating widespread employee access to AI as evidence of success. "95% of everybody has X, Y, Z," she said. "But it is not translating into productivity, and it's often not with training, and that doesn't equal maturity." In her view, the real work begins with leadership, a clear vision, transparent systems, the right data foundation, and the iterative redesign of specific workflows.

Paul Fipps, president of global customer operations at ServiceNow, made a related point: the biggest gains will come not from running old processes faster, but from rebuilding them around new capabilities. "If I just automate or drive autonomous agents to drive the current processes I have, that's a choice that might get some productivity bump," Fipps said. "But the ones that are actually rethinking and reimagining the business processes … are really making the difference."

Fipps argued that companies should roll out AI horizontally—across employee-facing functions such as HR, legal, and technology—rather than confining it to a single department. He pointed to Standard Chartered, where AI has helped deflect 77% of employee requests. "If you get 25% to 30% deflection in those use cases, it's game-changing," he said. "If you get 75% to 80%, it is life-changing."

Amy Webb, founder and CEO of Future Today Strategy Group, was blunter about the distance between ambition and readiness. "Let's be fair, Fortune 500 companies are pretty late to the party on this," she said. "Artificial intelligence didn't just show up a couple of years ago." Webb cautioned that companies are committing substantial capital to pilots without adequate strategy, leadership, or planning—producing isolated projects that never deliver durable value.

Building the AI-Ready Workforce

AI transformation is, at its core, a workforce transformation.

The summit made clear that organizations need more than new software: they need employees and leaders willing to rethink established processes, learn new skills, and make room for new ways of working.

Stacy Dillow, executive vice president and chief people officer at HPE, said companies must resist the temptation to simply insert AI into existing systems. "We've got to force ourselves to think differently, because if we're not trying to change how we work, we're just trying to retrofit and put AI in a process that isn't going to give us a different answer," she said. For senior executives, she added, AI adoption and transformation have become central management responsibilities.

At Palo Alto Networks, AI is already reshaping the employee experience, according to chief people officer Danielle Gonzalez. A custom AI agent has cut IT tickets by 83% and now supports workflows across HR, finance, and legal. Rather than turning to a human HR partner or business contact, employees increasingly ask an agent for answers tied to their role and location.

Gonzalez said companies should hire and develop people who can adapt quickly—not only picking up new skills but questioning assumptions that no longer hold. "We're looking for curiosity," she said. "We're looking for agency, and we're looking for those people who can not only learn something at a fast rate … but also unlearn what they thought to be true, so that they can make space to move forward."

Drew Holler, chief human resources officer at homebuilder Lennar, stressed that the responsibility is shared. "We're going to give you all the tools, but it's your responsibility to upskill yourself as well," he said. "We're going to give you trainings, but you, as an individual, have to upskill."

The workforce implications extend beyond reskilling. Francis Shanahan, chief technology officer at Peloton, warned that AI-enabled creation carries its own challenge. Employees are "inundated now because they can literally build anything," he said, and "burnout" is becoming more of an issue.

Michelle Kwon, chief operating officer at Runway, urged leaders not to mistake volume for productivity. "Just because you are using, that is not a proxy for your productivity or for doing a good job at work," she said. AI, she argued, should improve how people work without generating extra work for colleagues or overwhelming them with output that has not been properly assessed and synthesized.

Governance, Trust, and Control

The case for AI must be matched by a case for oversight.

In regulated industries and high-stakes applications, leaders emphasized that successful deployment depends on choosing the right technology, understanding the risk, and keeping human judgment where it matters most.

Hari Gopalkrishnan, chief technology and information officer at Bank of America, cautioned companies against reaching for AI reflexively. "One of the biggest mistakes we see ourselves and others doing is rushing to AI as a solution, when deterministic models do a plenty good job," he said. The bank starts with client needs and a detailed inventory of the process in question; at times, a conventional mobile feature or real-time decision rule is the better tool.

Bank of America evaluates AI projects through 16 risk pillars, including privacy, bias, workforce impact, and intellectual property. It is also moving deliberately on autonomous systems. "There is so much juice to be squeezed right now with assistive agents are actually working with humans in the loop," Gopalkrishnan said. For banks and other regulated institutions, frameworks like these have become the gate between building AI agents and putting them into production.

Sally Moore, chief client officer and co-head of market intelligence, Kensho data and platforms, at S&P Global, stressed that AI's usefulness rests on trustworthy data. "Data is the currency within AI," she said, noting that clients in regulated workflows require accuracy, citations, auditability, and traceability back to the source. Her broader point: companies should start with the problem to be solved, not the technology they want to deploy. "What am I trying to solve for here? And where can this technology take me?" she said.

Fipps contended that governance should not be framed as an obstacle to innovation. "I see governance and trust driving speed," he said. He cited the example of a regulated bank that had built dozens of agents but could not deploy them because its leaders lacked confidence in the controls—a vivid illustration of the cost of weak oversight.

Webb likewise pressed executives to take a practical approach to liability and risk. Leaders need to "stop anthropomorphizing" AI, she said: failures typically stem not from technology that has become independently sentient, but from people deploying poorly designed systems without adequate safeguards.

AI in the Real World

Physical and regulated environments require a different standard of AI deployment.

In buildings, factories, water systems, and financial services, reliability, safety, cost, latency, and data sovereignty shape what companies can responsibly put into production.

Suresh Venkatarayalu, chief technology officer at Honeywell Technologies, said industrial customers "really demand 99.9999%" reliability, while frontier models "could be at 85%." That gap makes deployment in safety- and mission-critical settings fundamentally different from dropping an AI assistant into a desktop workflow.

Venkatarayalu described the near-term direction as "a semiautonomous world" rather than a fully autonomous one. "Autonomy is also not about removing people," he said. In commercial buildings and industrial facilities, operators must remain involved because downtime and system errors carry immediate operational consequences. "You cannot afford to have a building shut down for one and a half hours," he said.

AJ Wijesinghe, chief AI officer at Ecolab, made the economic case for a more selective approach. When the company used "the best model that's out there" for high-volume tasks, "the tokenomics go off the roof," he said. "Sometimes it's more expensive than having humans." Ecolab responded by optimizing its use of models and cutting token costs by roughly 70% to 80%.

Companies do not always need the most advanced model available, Wijesinghe said: "Sometimes you don't have to have the fastest car." He advocated a "human in the lead" approach for agentic AI and argued that results depend on keeping data foundations, process readiness, and cost discipline in balance. "If one is heavier than the other, then you don't get the value," he said.

Value, Cost, and Growth

The next test for AI will be whether it delivers measurable returns.

The summit's finance and operating leaders agreed that companies must pursue productivity and growth while maintaining discipline on model costs, data infrastructure, and outcome quality.

Ewout Steenbergen, executive vice president and CFO of Booking Holdings, said uncertainty over AI's returns is widespread—even among the largest technology investors. "I even think that the hyperscalers that spend hundreds of billions on the development of their large language models, they don't really know what is going to be the ROI," he said. "They have maybe some assumptions and some hypotheses, but the whole point is, you don't want to fall behind."

For Booking, the path to value runs not through indiscriminate spending but through redesigning processes "from a white sheet." Steenbergen said the company's customer-service costs are slightly down even as bookings grow at a high-single-digit rate, while customer satisfaction has risen. He added that Booking's roughly 9,000 engineers are getting about 30% more code into production, counting merge requests that clear the company's testing and quality-control standards.

"The returns will be there for those processes that are being redesigned end to end," Steenbergen said, "but you also have to make sure that those costs on the other hand are not going out of control." Booking manages that tension through "effective model cost routing," assigning lower-cost models to simpler tasks and reserving more expensive models for work that demands deeper reasoning. The approach offers a concrete look at how usage-based model costs—an expense line that grows with adoption rather than behaving like a fixed software license—are being managed inside a major public company.

Steenbergen suggested AI's larger potential may lie in customer growth. The technology will not change the fundamental nature of travel—"the hotel is still the hotel, and the airline is still the airline, and the rental car is still the rental car"—but it can make planning, booking, and responding to disruptions easier for travelers. By linking the parts of a trip and anticipating a customer's needs, he said, AI can help create "more loyalty, more trust, and more brand value."

The Bottom Line

The Fortune AIQ Summit made clear that AI leadership is not defined by how many tools a company has purchased, the size of its pilot program, or the volume of content it can generate. The organizations pulling ahead are pairing technical ambition with human development, process redesign, financial discipline, trustworthy data, and accountable governance. The summit's own data points provide markers for tracking that shift: whether the talent and training gaps exposed by ServiceNow's maturity index narrow, whether industrial deployments close the distance between frontier-model reliability and the near-perfect uptime physical operations demand, and whether ROI clarity improves for the hyperscalers spending hundreds of billions to build the underlying models. As AI moves deeper into the enterprise, the leaders who succeed will be those who use it not simply to automate existing work, but to build more resilient, productive, and responsive businesses.

Generative AI was used for research assistance and/or transcription of this article. The reporter independently reported and verified the factual claims in this article, and a human editor reviewed it before publication.

This story was originally featured on Fortune.com.