AI is evolving beyond experimentation, leaders say

Artificial intelligence is finally crossing the threshold. Businesses have long talked about AI as a force for change, but most activity has remained in sandboxes and innovation labs. In 2026, that balance will begin to shift.

When more than 300 corporate strategy leaders gathered in Munich this month for the World Economic Forum’s Industrial Strategy Conference, the question wasn’t “Will AI work?” “Where are we already seeing changes in how we operate at scale?”

The pressure to answer that question has never been greater. Discussions at the conference highlighted the scale of the challenge. Roughly three-quarters of companies are yet to generate meaningful value from AI, and many remain in the testing phase despite increased investment. As one executive articulated, 2026 is the year companies must prove that AI can return value.

Emerging technologies such as AI, automation, and advanced computing are being woven into the core of enterprises in a variety of areas, from decision-making systems and supply chains to finance, customer service, and engineering workflows.

At the enterprise level, we are seeing a decisive shift from pilots to platforms, with budgets increasing and 38% of organizations operating AI use cases, according to Capgemini research. In parallel, first-mover companies are beginning to report tangible results such as increased productivity, reduced cycle times, reduced costs, and new revenue streams.

However, the path from proof of concept to commercialization is rarely linear. Leaders must address issues ranging from trust and governance to skills and infrastructure, while keeping business value top of mind.

Here, executives from a variety of industries share how adoption is changing within their organizations and what it actually takes to make AI stick.

“In our business, agenttic AI for enterprise planning and execution has moved from experimentation to production deployment. However, careful design of agenttic AI solutions was required,” he said. Dr. Ashwin RaoExecutive Vice President of AI and R&D at o9 Solutions.

“Essentially, we combined the strengths of Neural AI (LLM) with the complementary strengths of Symbolic AI (structured enterprise data models and decision-making models based on deep enterprise knowledge).

“This allowed our agent to benefit from the scalability, adaptability, and learnability of neural AI, but also the inference, accuracy, and explainability benefits of symbolic AI. This allowed us to build a reliable agent that performs well in enterprise practice, as opposed to typical LLM-heavy agents that don’t move beyond compelling demos and POC.

“In terms of impact, for example, corporate planners are getting a lot of help with supply chain planning and root cause analysis of variances in results. This has allowed planners to reduce investigation time by as much as 80%.”

“We’re also seeing significant benefits from touchless execution of inventory and logistics, which saves even small businesses tens of millions of dollars in labor costs. We’ve also found that with AI’s assistance in driving integrated business planning, decisions across silos are now executed in a quarter of the time, making businesses significantly more agile.”

When Siemens introduced AI to its electronics factory in Amberg, he said there was some initial resistance. Dr. Gunter BeitingerSenior Vice President of Manufacturing. Siemens’ head of factory digitalization speaks at the Industrial Strategy Conference in Munich.

“In the early stages of AI implementation, there was a lot of concern, especially among shop floor and manufacturing staff, and a lot of anxiety about, ‘What is AI going to do for my job?’ … So what we did was slowly introduce what we wanted to do and involve people in the whole development.”

For example, the introduction of AI into X-ray quality assurance included someone training an algorithm to identify products that no longer needed AI quality assurance in the process to save time.

“We calculated that if 5% of our products were screened, that was already an economically viable scenario.”

Over time, factory teams have come to rely on AI to accurately identify products that do not require X-ray quality assurance, reaching 30% of products for which additional steps can be eliminated.

People are really confident. They designed the AI ​​from the ground up with experts to ensure it was safe, ethical, and explainable. ”

The combination of AI and data supports three main changes, Dr. Beitinger said.

“We’re moving from efficiency to resilience. That’s one of the things we’re looking for. Previously, there was global cost optimization. Of course cost is very important, but resilience is increasingly important. We’re also moving from automation to autonomy, from optimizing a single factory to the entire production ecosystem. AI is really helping us make those transitions.”

“The data and machine learning algorithms underlying the application of AI technology enable us today to manage our businesses better and more efficiently,” he said. Filippo RicchettiHead of Planning and Management and Insurance, Eni.

Among other things, Eni was able to reduce uncertainties in mineral resources, digitize business processes and industrial plants, and accelerate the development of new business and energy chains.

“Over the years, Eni has used AI tools in various sectors, in around 300 use cases from exploration to operations. The use of AI solutions has increased efficiency by reducing plant downtime, optimizing production and reducing emissions. For example, in the drilling sector, a 35% reduction in drilling time has been achieved. Automation and machine learning have led to these improvements.”

Looking to the future, the company is building an AI agent based on its own corporate data and knowledge assets. This is a deliberate choice to maintain control of the most important intelligence. Initial applications have already been launched throughout the process, managed by a responsible AI framework designed to ensure transparency and accountability at every stage.

“Every technology resets the world. The arc is always the same: experimentation, adoption, dependence, irreversibility. What’s different this time is the speed and the stakes,” he said. Hala ZeineSenior Vice President and Chief Strategy Officer at ServiceNow.

“Organizations are not waiting for evidence; they know that hesitancy is a structural flaw, so they acted early.

“But experimentation is just the first step. The real transformation will happen when AI is built into the flow of work by rewiring how companies sense context, make decisions, and execute at scale. That’s when adoption becomes cognitively dependent and AI in workflows becomes not only helpful, but non-substitutable. Organizations that cross that threshold first will have a growing advantage that latecomers simply can’t get away with.”

“At ServiceNow, we are built for this very moment. AI is not bolted into our platform, but built natively into the infrastructure that already orchestrates 80 billion workflows and 6.5 trillion transactions annually for 85% of Fortune 500 companies. Our architecture is designed to move organizations from experimentation to irreversibility, which means enterprise It senses context, makes business-responsible decisions, operates autonomously within managed workflows, and governs every step with audit-grade controls.

“The results are already getting worse: AstraZeneca recovered 30,000 hours a year; Pure Storage resolved cases seven times faster; Siemens autonomously processed 210,000 tickets per month. These are early signs of cognitive dependence at scale, not efficiency gains. AI in workflows is autonomous enterprise execution. This is where the competitive advantage becomes permanent.”

“S&P Global is at the forefront of artificial intelligence, transforming data analysis and decision-making across industries.” Swati SaudiyaniSenior Vice President, Strategic M&A and Ventures, S&P Global.

Since acquiring Kensho Technologies in 2018, S&P Global has leveraged AI to enhance workflows and drive value for customers.

“We are committed to providing our customers with access to trusted S&P Global data no matter where their workflows occur, ensuring that we meet their rapidly evolving needs in the AI space. This means enabling our customers to seamlessly integrate our data and insights into their Gen AI workflows. One example of these efforts is through our LLM-enabled Application Programming Interfaces (APIs), which allow our customers to access various S&P Global data through any Gen AI application. Seamless access to global datasets.

S&P Global is focused on driving AI innovation, including partnering with leading companies such as Anthropic, OpenAI, and Google’s Gemini to develop new AI capabilities.

“These new capabilities enable our clients to gain connected insights, increase automation, and make faster decisions.

“AI will fundamentally change the way the financial services industry works. The strategic priority is not just to deploy the technology, but to help people across the organization become proficient with it. That means building a culture that embraces AI, along with the training and tools to make that change happen.”

According to Cohere’s chief AI officer, it’s helpful to think about three generations of AI. joel pinaultHe gave a speech on the sidelines of the Industrial Strategy Council.

“There is something called predictive AI, which is the ability to use data information to do fairly specific predictions, weather forecasts, classifications, etc., which almost every company relies on today. If you have a digital footprint as part of your business, part of it is powered by predictive AI.

“Generative AI is more recent, going back to around 2022-23. We’re starting to see companies experimenting with this. Given the big technology changes of the past, it took decades to actually materialize.

“Agent AI is actually going to burst onto the scene in 2025. It seems pretty optimistic to think we’ll see widespread adoption across industries directly from the lab within a year. There are a lot of very positive predictions about AI, including agent AI. What’s surprising is that we’re already starting to see the benefits and benefits of new technologies like this.”

The transition from experimentation to deployment is real and accelerating, but it’s not primarily about technology. This is a story about organizational redesign, cultural trust, and a solid data foundation.

The companies that move the fastest are not the ones with the best models. They are the ones who have done the more difficult work of preparing the organization to receive them.

Quotes have been lightly edited for length and clarity.