How leading companies are turning AI vision into business value

I’m fortunate that I get to speak with hundreds of customers every year. These enterprises span the technology adoption curve—some who thrive on the leading edge, and others who prefer being as risk-averse as possible. 

But the AI era—ever-churning and bursting with exuberance—has united nearly all of them at this moment. Experiment time is over. Enterprises are focused on moving beyond theoretical AI pilots to operationalizing it at scale, optimizing costs, and governing its actions.

Beyond those corporate mandates, which are important, I think what’s refreshing about my conversations with customers is that there’s an air of excitement about AI. About harnessing new ideas and seizing on opportunities not thought possible even a short time ago.

For example, at global bank BNP Paribas, their objective is to industrialize AI at scale—not just experiment with it. They have thousands of use cases already live, but they’re not stopping. They’re currently building an AI factory to deliver AI capabilities across the company with an emphasis on business value. This includes key elements such as fraud detection, “Know Your Customer” (KYC), client onboarding, and chatbots, plus capabilities for internal productivity, such as AI for developers. And they’re adamant that the focus has to be on measurable business outcomes.

What I love about BNP Paribas is how they demonstrate the power of adaptability combined with the requirement of good governance. 

This is exactly how automobile-maker Nissan is making the transition from software-defined vehicles to AI-defined vehicles, where AI workloads are a core product enabler. With compute, the most important requirement is flexibility, Nissan says. AI workloads will continue to evolve, so Nissan’s focus is on the ability to adapt compute—on-board, off-board, or hybrid—without locking software to a specific chip generation. And with data, AI raises the bar even higher. Data handling is critical not only to improve models, but to ensure accountability, trust, and continuous learning across Nissan’s AI-Drive and AI-Partner products—always with security, privacy, and governance built in.

Agents, agents everywhere

There’s a ton of interest and investment in agentic AI—those systems of AI agents that can reason, plan, and autonomously execute complex tasks. Let’s look at a couple of examples.

• BNP Paribas is exploring an internal project called Twin to deploy agents that observe GPU usage, detect anomalies, and automate remediation workflows in real time. 
• UPS, the global package-delivery and logistics company, views the future of AI through the lens of decision science—where agents analyze situations, determine probable next actions, and engage in automated decision trees to aid supply chain logistics.
• US telecom Verizon envisions using agentic AI to run complex triage loops that can detect network degradations for individual customers and autonomously execute actions to resolve issues before their customer even asks for help.

But we can’t forget about the humans! NASA, the US space program, aeronautics research, and space exploration organization, stresses that AI agents are meant to assist and act as a force multiplier rather than replace engineers and scientists. That’s the scale opportunity: Enterprises can increase AI’s impact because AI can take over rote but important maintenance and computational tasks so humans can focus on nuanced design and high-level decision-making. 

For example, EUROCONTROL, the European air-traffic control agency, is using high-performance AI systems to calculate complex trajectories in real-time, helping airspace users identify the most fuel-efficient and sustainable flight routes. NASA is looking to use AI models in space to rapidly visualize and process data, filtering insights so astronauts can focus on decision-making rather than data triage. And lastly, Motorola Solutions, which builds mission-critical communication devices, systems and software for public-safety agencies and enterprises, is integrating cloud-based AI workflows into on-premise emergency systems, providing 911 dispatchers with live transcripts, real-time translations, and automated case summarization.

Digital sovereignty and AI

With expanding regulatory oversight (like the EU AI Act), customers I speak with are asking for verifiable digital sovereignty over their AI. This breaks down into 3 requirements:

• Model sovereignty: The ability to choose, change, and run specific AI models anywhere without vendor lock-in.
• Data sovereignty: The data used to train and run models stays entirely under the organization's jurisdictional control and is never used to train external models without consent.
• Outcome sovereignty: The ability to audit, explain, and override any significant AI-generated decision. 

For example, regulated entities like Core42 and Telenor in Europe are building highly secure, multi-tenant sovereign AI clouds so data never leaves their control. 

These are just a few highlights from our customers about the future of AI at their organizations. I’m inspired by how they’re planning their moves from AI pilot to AI in production, seizing on the ideas their teams are dreaming up every day.