In Brazil, a truck driver hauling ethanol across a treacherous stretch of road starts to swerve. Inside fifteen seconds, a camera in the cab flags it. An AI agent cross-references the telematics, decides he's fatigued, and a second agent radios him in Portuguese to talk him through getting to safety. The driver is back on course before a dispatcher could have dialed his number.
That system runs in production today, across hundreds of trucks, in 32 languages. It was built by Roger Rohatgi's team at Chai. And it was one of three perspectives on a stage in Chicago that kept arriving at the same place from completely different directions.
I was moderating a panel at The Future Of conference. Roger was there. So was Anthony Wolf, who runs AI activation at Canadian Tire, Canada's largest retailer. And Carly Price, a design leader at ServiceNow, a company where AI is woven into every workflow. John Gleason and the team behind The Future Of had built a room worth being in: P&G, IBM, Whirlpool, KraftHeinz, Colgate, Mondelez, 3M all working through the same questions on the same day. The hallway conversations alone were worth the trip.
These three kept arriving at the same place from different doors. And the place is something most companies are getting wrong.
Anthony told the room that Canadian Tire doesn't have an AI strategy.
I watched people's faces shift. In a room full of leaders building AI strategies, roadmaps, and governance frameworks, someone running AI activation at one of Canada's largest companies just said they don't have one.
AI is in service of their corporate strategy. That's it. And the more I've sat with it, the more I'm convinced the act of creating a separate AI strategy is itself the mistake. The moment you separate it, you've turned it into a project. A thing to manage. A workstream with a sponsor and a timeline and a definition of done. That's exactly what most companies did with digital transformation. Separate team. Separate budget. Separate initiative. Most of them never finished.
If you have an AI strategy, you've already lost the plot. AI isn't a workstream. It's the operating environment.
Roger sees the evidence. His team walks into companies that say they're ready for AI, and here's what they actually find. Data that's a mess. Processes that are disconnected. Cultures that are confused. Legacy systems nobody replaced. Digital transformations that got 80% done and moved to the back burner. AI doesn't clean any of that up. It inherits it. Then it scales it at a speed your organization has never operated at before.
The problem isn't AI. It's everything you didn't finish before AI showed up.
Then Carly complicated things in the best possible way.
AI was already woven into everything at ServiceNow when she arrived. The question wasn't whether to use it. It was how fast. And the breakthroughs she's seeing aren't coming from a strategy. They're coming from people.
Hungry, proactive individuals who are just figuring it out. A designer who built a prompt template that feeds directly into Figma Make. Another who went into Terminal and started building. Carly organized a two-day workshop that used AI to facilitate synthesis across all their portals. Twenty prototypes came out of it.
That workshop is worth pausing on. It compressed what would have been six to eight weeks of work into two days. But the insight isn't speed. The old timeline existed because synthesis was expensive. AI made it cheap. The six weeks wasn't the work. It was the cost structure the process was designed around.
Think about what that means. Almost every slow process in your company was designed for a world where certain things cost more than they do now. The work isn't hard. The economics are obsolete. If you don't redesign the process around the new unit cost, you're paying six-week prices for two-day work. Compound that across a few hundred workflows and you've described why most enterprise productivity numbers haven't moved.
Her team has since built custom Claude skills for heuristic evaluation. Junior designers running mini usability audits before they share their work. Demo scripts built from PRDs, turned into coded components, connected through to engineering. The whole concept-to-testable-product pipeline compressed in ways that would have been hard to imagine a year ago.
And Carly isn't overseeing this from a conference room. She uses voice memos on walks for meeting notes. Builds prototypes herself. When a designer showed her something impressive, her first question was "show me the steps." Then she started doing it.
Roger says companies aren't ready. Anthony says the readiness problem is organizational. Carly is showing what actual readiness looks like, and it looks nothing like a transformation program. It looks like a leader in the tools, creating space for people to figure it out. No roadmap. No separate AI strategy. Just conditions for experimentation and a leader who goes first.
Anthony's analogy keeps rattling around in my head. AI isn't like bringing a power drill to replace a screwdriver. It's like a die casting that replaces the entire need to screw something together in the first place.
This is the recasting.
Canadian Tire has millions of loyalty program users. Deep first-party data on what customers are buying, when, and in what combinations. They could have used AI to optimize how they sell products. Faster recommendations, better targeting, the obvious stuff. Instead, they reimagined what they're selling around. Not products. Occasions. What's the weekend going to look like for a store operator near a lake during fishing season versus an urban store near a music festival? 1,500 micro occasions, defined from first-party data, scaled store by store, week by week.
That's not a faster version of Canadian Tire. That's a different company.
But here's the part of the recasting that nobody said out loud. If AI replaces the need to screw things together, then the expertise people built around screwing things together is suddenly worth less. The 70% people problem Anthony talks about (10% models, 20% data and tech, 70% people) is not just about adoption and training. It's about identity.
What happens when the thing that made you an expert is the thing AI just collapsed? You don't lose your job. You lose the part of the work that made you feel competent. That's a harder thing to manage than headcount.
Anthony named the emotional weight of this more honestly than I've heard anyone do it. There's a maturity curve with AI, he said. And it doesn't have an ending. It just keeps going. You can't cross a bridge and tell your board you're on the other side.
There is no other side.
Every leader going through this is feeling the same excitement, the same fear, the same apprehension, all at once. That's just what leading in the age of AI feels like. Every change management playbook that assumes you're moving from state A to state B breaks when there is no state B.
Roger pulled the camera back.
He shared a framework from his work with MIT. Work has always been done human to human to human. H, H, H, H in the chain. Now you introduce an agent. H, H, H, A. Then another. Eventually a single human leading a network of agents. Or an agent leading agents and humans.
The question stops being about tools and starts being about decisions. Which ones are you willing to hand over? Roger believes that's the biggest decision organizations will face. Not which AI to buy. Which decisions to trust it with.
Most companies don't have a framework for that. They're trusting AI with everything or trusting it with nothing. The ones who build trust architecture first, with clear principles for what goes to humans and what goes to agents, will move fastest.
The Brazilian biofuels company shows what that architecture looks like in production.
An email comes in. An AI agent reads it, extracts the order, runs a quote.
Legal review and human approvals happen where they're needed. A work order goes to the warehouse. Inventory gets counted, picked, packed.
The system schedules delivery. Calls the freight forwarders. The truck driver gets monitored by the safety AI already described. Cameras detect damage at the yard, manage dock maneuvering, verify inventory.
Social media monitoring closes the loop on customer experience.
Roger calls this the agentic workforce. But it's only legible as a single coherent thing because someone designed the experience. The technology is a stack of agents. The product is a service. The difference is the experience layer that holds it together.
Which brings us to the most important argument Roger made, and the one I think most companies are going to miss.
Jensen Huang's AI stack has five layers. Energy. Infrastructure. Chips. Models. Applications. Roger argues there's a missing sixth: experience. Everything else is being built. Everyone has access to roughly the same models. The chips are commoditizing. The applications are converging. But people aren't adopting any of it. They're confused by it. The experience layer is what's missing.
Every dollar your company is spending on the first five layers is a dollar spent on something that won't differentiate you. The compute is the same. The models are the same. The applications are converging. The only spend that compounds is on the layer closest to humans.
Design leaders, the people who understand experience, who understand how humans actually interact with systems, are sitting on the most valuable real estate in the AI landscape. Most of them are still waiting for someone to hand them a strategy.
Don't wait. Build the experience layer. That's where the advantage compounds.
An audience member asked a question that proved Roger's point. He's inside an enterprise where the internal AI tools aren't as capable as what's available externally. Use the company tools and fall behind. Experiment on your own and create risk. That gray space is the experience layer breakdown in real time.
Anthony's answer stuck with me. Governance should be guardrails, not speed bumps. Guardrails let you go faster because the rules of engagement are clear. Speed bumps just slow everyone down equally. Most companies are building speed bumps and calling it governance.
I told the room what I believe. Get your hands on the tools. Personally. Even if it's outside the enterprise firewall. If you haven't built something with AI, you don't fully understand what it can do and you don't understand where it falls short. Both matter. Corporate IT shouldn't define your understanding of what's possible.
Nobody in that room is as far ahead or behind as they think. The maturity curve doesn't end. If you approach it with curiosity, to learn from each other instead of measuring yourself against each other, we all move faster.
The competitive advantage isn't having AI. Everyone's going to have AI. Same models. Same tools. Same infrastructure. Same agents available off the shelf within a year.
The advantage is how fast your organization can absorb it.
Absorption is a function of culture, trust, and leadership. Not technology procurement. Speed of decision. Speed of synthesis. Speed of handoff between humans and agents. Speed of recasting the processes that no longer match the new unit cost of intelligence.
Speed compounds. The companies that internalize that first will end up running a different business than the ones still building AI strategies.
There is no other side. Just the next casting.
Jason Hauer is the founder and CEO of HauerX Holdings, where he backs and builds a portfolio of AI-native companies that accelerate how businesses grow, operate, and compete. From mid-market to Fortune 500.
