I keep having the same conversation. A founder tells me they’ve rolled out AI across the company. Licenses bought, a couple of workshops run, people using it every day. Then I ask what actually changed on the P&L, and the room goes quiet. Mostly nothing. Maybe a few points.
The tools work. The adoption is real. And the results are a rounding error.
That’s not a tooling problem. It’s a change problem, and almost nobody treats it like one. I’ve started thinking about it as three gates a company has to walk through, in order. Gates, not stages, because you can’t skip one and you can’t fake your way past it. Most organizations are stuck at the first and think they’re done. The ones doing real damage to their own future are stuck at the second. And the third is where leadership teams quietly fail, because clearing it requires the one thing executives are trained to avoid.
Gate one: earn the right to have an opinion
The first gate is curiosity. Not the LinkedIn kind, where someone “is so excited about the potential of AI” and then goes back to their inbox. Real curiosity that ends in evidence.
Here’s the part people miss. Playing with a tool teaches you nothing useful. Running a real project through it teaches you everything. You don’t find the edges of what AI can do by asking it to write a haiku or summarize an article. You find them by handing it work you actually own, with a real deadline and a real definition of done, and watching exactly where it shines and where it falls on its face.
That’s the whole point of gate one. You’re not trying to be impressed. You’re building an accurate map of what these tools can and can’t do, so that when the bigger decisions come, you’re making them from experience instead of hype or fear. The person who has pushed twenty real tasks through these tools has a calibrated read on what’s possible. The person who read three threads about it has a vibe. Those are not the same input, and you should not let people with vibes write your strategy.
Most companies think they’re through this gate because a few employees use AI to draft emails. They’re not. Drafting emails is the haiku. The question is whether your organization has flexed these tools against its actual hard problems and built an honest, shared sense of where the line sits today. If you can’t point to the projects that did that, you’re still standing at gate one.
Gate two: be willing to destroy what works
This is the hard one. This is where I watch almost everyone get stuck, and it’s the gate that separates a 7% company from a 70% company.
The trap is seductive, and it sounds responsible. You take your existing process, the one that’s been working for years, and you go hunting for the steps you can automate. You find a few. You bolt AI onto them. You get a modest lift, maybe 5 to 10%, and you call it a win. Everyone feels productive. The process is faster.
You just paved the cow path.
The phrase comes from the early web, when companies took their paper forms and made them digital, faithfully reproducing every dumb step of the old process on a screen. They optimized the path the cows had worn into the field instead of asking whether the road belonged there at all. They got a faster version of the wrong thing.
That’s what bolting AI onto your current process does. The process you have was built around old constraints. That work is expensive. That a human is the only thing that can read a document, make a judgment, write a first draft. That coordination takes meetings and email. AI removes or reshapes most of those constraints, which means the right process on the other side doesn’t look like your current one with robots glued to it. It often looks nothing like it.
The blindness is completely natural, and that’s exactly what makes it dangerous. You’ve done it this way for years. It mostly works. Your team knows the steps. Of course your instinct is to protect the structure and improve the parts. But “it mostly works” is the sentence that keeps you parked at gate two forever.
The companies that break through stop asking “what can I automate in this process?” They ask “if I were building this function from scratch today, knowing what these tools can do, what would I even build?” Then they build that, and let the old process die. It’s an uncomfortable question, because the honest answer usually erases work people are proud of and good at. But you don’t get a step change by defending the staircase.
Gate three: fund the dip
Say you clear gate two. You’ve redesigned a core process around what’s actually possible now. Here’s the part nobody warns you about. It’s going to get worse before it gets better.
There’s a pattern in any serious transformation called the productivity J-curve. Rip out a working process, replace it with a genuinely new one, and performance drops first. People are learning. The new workflow has bugs. The old muscle memory is now wrong. The numbers dip. Only after the organization climbs out of that trough do you get the payoff, and the payoff dwarfs the incremental tweak you passed up.
The problem is the timing. The dip is real and immediate. The payoff is delayed and, at the moment you’re living through the trough, entirely theoretical. This is where leadership decides everything. A management team that flinches at the bottom of the curve pulls the plug, blames the tools, and retreats to the safe 7%. They’ll tell themselves they “tried AI and it didn’t move the needle.” What actually happened is they quit during the exact stretch where quitting was guaranteed.
Clearing gate three means doing something that cuts against every instinct of quarterly management. You have to choose to take a step backward to go forward. You have to protect the team in the dip from the company’s own antibodies, fund them through the ugly middle, and hold your nerve while the early numbers look bad. That’s not a tooling decision or a process decision. It’s a leadership decision, and it’s the one most leaders won’t make, because things are “going okay,” and okay feels safe.
Okay is the enemy here. Okay is what gate two hands you on its own. The jump from good to great lives entirely on the far side of a dip you have to be willing to walk into on purpose.
The part nobody wants to hear
Notice the gates get harder as they go, and they get more about people and less about technology. Gate one is curiosity. Gate two is intellectual honesty. Gate three is nerve. By the time you reach the gate that actually unlocks the value, the tooling is the easy part. Your own leadership team is the constraint.
That’s why most companies will get this era half right. They’ll clear gate one, mistake it for the finish line, harvest their 7%, and feel fine about it. The few who redraw the whole field instead of paving the cow path, and then hold the line through the dip, won’t be 7% better than those companies. They’ll be playing a different sport.
The gates are open. The only question is how far you’re willing to walk.
MR.TODD GAGNE,one of top Technology Thinkers,Author and Renowned Tech Startup Consultant,drawing on deep research,real life experience & actual data, identifies Change Management problem in AI—that nobody's naming—and shares deep insights on why—Most Companies Will Get AI Exactly Half Right——The tools work, adoption is real, and the P&L barely moves.
MR.TODD analyses the change management problem by using —The Three Gates Metaphor—Gate one is curiosity. Gate two is intellectual honesty. Gate three is nerve. By the time you reach the gate that actually unlocks the value, the tooling is the easy part. Your own leadership team is the constraint.—and Aptly opines that fragmented efforts and delay are becoming irreversible disadvantages because experimentation without direction or controls leads to fragmentation, fatigue, and risk.
MR.TODD advice is valuable & timely and is a framework of focusing on a defined set of value pools and creating mechanisms to scale what works while stopping what doesn’t.
Hmm although we are named Wildfire Labs we don't have anything to do with wildfires...