I turned 50 in January 2025.
Most guys in that situation buy a sports car. I signed up for a triathlon. Two of them, actually.
A sprint in June (in Welland, Ontario), and an Olympic distance in August (in Wasaga Beach, Ontario). For those unfamiliar with the distances, a sprint is very manageable. An Olympic is the kind of thing that makes a 50-year-old seriously question his judgment. I had zero triathlon experience, didn’t know how to swim, and was a horrible runner. What I needed was a training program.
The $20 Plan
After some googling, I ended up on a site called TrainingPeaks, a platform where coaches sell structured programs for various sports. For twenty dollars, I bought a plan built by a personal trainer, designed to carry me from couch to Olympic triathlon over the exact number of months I had before the Olympic distance race. Twenty dollars for a roadmap that understood something I didn’t yet appreciate: the gap between where I was and where I needed to be could only be closed one week at a time.
The program started embarrassingly small. Short swims and easy bike rides. Runs that felt more like walks, frankly. Nothing about those first few weeks felt like real training, and I remember wondering if I’d wasted my money on something too basic.
Then it built. Week by week, the distances grew and the combinations got harder. My body adapted, slowly enough that I almost didn’t notice until I looked back at month one. By race day, nothing the course threw at me felt overly difficult because the plan had already put me through all of it in pieces across months of quiet, boring, necessary work.
Start small. Finish big. That’s all it was.
It worked. Later, I realized it was a business lesson in disguise.
The Rollout Problem
Companies and entrepreneurs love reaching for the stars. They love the “go big or go home” mentality. But AI rollouts, and most massive initiatives, usually do go home if you skip the structured work of getting from zero to capable.
I use AI rollouts in this article because they are the clearest example right now. “We need to roll out agentic AI across our organization.” No, you don’t. You’re not ready. You don’t understand your own workflows, let alone how to apply large language models to them.
Agentic AI is the Ironman triathlon of today’s AI technology, far beyond the Olympic distance. It makes decisions, runs multi-step workflows, and operates without a human checking every stage. Yes, this is where the technology is going. But it will wreck you if you skip the training.
Companies that fail at AI rollouts, or any major project, almost always fail the same way. They attach advanced tools to workflows they don’t fully understand or that haven’t been tested, staffed by people who haven’t learned the basics. A lot of money gets spent. A lot of frustration follows. Six months later, there’s nothing to show except a cautionary tale and a team that’s become skeptical of the whole thing. Oh, does the hair on the back of my neck stand when I hear them say “we tried AI but it made too many mistakes”. Sorry sir, it was you who made the mistakes. AI didn’t make the mistakes, it just scaled yours. AI is an amplifier of potential.
You don’t go from zero to hero. That’s true in triathlon and it’s true in business.
The Training Plan for AI
Before your organization is ready for agentic AI, some questions need answers.
Do you understand your workflows? Where does information actually move? Where does it stall? Who makes the decisions, and based on what? If you can’t draw the map, you can’t plan the route.
Do you know your real pain points? The repetitive tasks eating up time, the handoffs that break down. The spots where a human is doing something a well-prompted AI could handle in seconds.
Next to that, there’s the people question. Have your employees actually learned to use AI? Used it, I mean. Sat with it, struggled with it, figured out where it helps and where it falls short.
There are stages to this, and they matter. First, AI as an assistant. Asking it questions, having it draft content for you, letting it summarize, and doing basic research. This is the sprint triathlon. Accessible, genuinely valuable if done well. Second, AI as a collaborator. Giving it context, working back and forth, letting it challenge your thinking, building on its output rather than blindly accepting whatever it produces. That’s the Olympic distance. You need the sprint behind you first.
Only after people have gone through both stages does agentic AI make sense. At that point, you know what AI can and can’t do, where guardrails belong, and what “success” would actually look like for an autonomous agent.
Skip the stages and the cost goes beyond wasted money. You erode trust in the technology before anyone has given it a fair shot.
It’s All About the Plan
I finished the triathlons and beat my expected time. The waves in Wasaga caught me a bit off guard because my training was in a pool. Other than that, my races were flawless because the plan had prepared me for exactly what race day would demand.
The twenty-dollar training plan worked because it was structured. It respected the process and it did not try to skip ahead.
Learn how to walk before you try running. That’s the lesson.
The most advanced AI strategy will fail if you don’t build the foundation underneath it first. The limiting factor is usually not the technology. It is the organization.
By the way, if you’ve never done a triathlon, try one. There are distances for every fitness level, and the atmosphere on race day is one of the best feelings you’ll have at a sporting event.
And if you’re thinking about an AI rollout, I have one question for you: what week of the training plan are you actually on?