Joel, thank you again for the opportunity to collaborate on this.
One of the things I appreciate most about your work is that you do not treat AI adoption as a toy box or a panic button. You keep bringing the conversation back to leadership, judgment, implementation, and what teams can actually do on Monday morning.
That made this piece a real pleasure to contribute to.
My hope is that the “front door” framing helps leaders pause before the next AI pilot quietly becomes permanent scope. Because the issue is rarely whether people are excited enough about AI.
The issue is whether the organization has enough structure to turn that excitement into accountable decisions.
At the innovation ends of the cycle. Pilots are totally disposable and meant to fail. It's better to title then as "experiments" but finance hates that name as it sounds frivolous and wasteful so pilots it is. 😉
Chris, I agree with the spirit of this, especially the need to distinguish exploratory work from committed delivery.
Where I would sharpen it slightly is that I do not think pilots are necessarily meant to fail. Some absolutely should be disposable. Some should be killed quickly. But some are really MVPs in disguise and can grow into something marketable if the learning is strong enough.
That is why I like the “experiment” framing, even if finance sometimes hears that word and starts looking for the adult supervision.
The key, at least to me, is making the boundary explicit up front:
What are we trying to learn?
What would make us stop?
What would make this graduate into real scope?
Without those answers, “pilot” becomes a politically safer word for “commitment we have not admitted yet.”
After setting up AI systems for dozens of small business owners, I have seen all three of these failure patterns play out. The one that hits hardest is treating AI as a one-time installation rather than an ongoing practice. Which structural failure do you see most often in the businesses you coach?
Ikram, yes. The one I see most often is intake collapse, or more honestly, the absence of any legitimately formalized intake process.
The example from the piece is the VP asking for an AI feature to summarize customer calls. In the unhealthy version, that request becomes “strategic” because it came from above. By the time anyone asks basic questions, the pilot has already started behaving like permanent scope.
The healthier move is to route it through intake first:
Who owns the data risk?
What does success look like?
What is the experiment boundary?
What gets delayed if we say yes?
Your point about AI being treated like a one-time installation lands for the same reason. Once the system goes live, new requests keep arriving. Without intake discipline, the AI tool slowly becomes another overloaded process with better branding.
This “front door is the governance moment” framing really lands. So many AI pilots go sideways not because the technology is unworkable, but because enthusiasm gets converted into scope before anyone has named the owner, success threshold, risk, or kill criteria.
The issue is not enthusiasm itself. Teams need that. The danger is when enthusiasm hardens into scope before anyone names the owner, success threshold, risk, or stop condition.
That is why I keep coming back to the front door. Sometimes governance is the moment a shiny idea is required to become a real decision before the team inherits it.
And for that to happen, you need a champion who either out ranks or is a direct peer in order to assure they comply with the funnel. Not always an easy ask but essential.
“The intake system is supposed to be the filter, not the rubber stamp” is exactly the line leaders need to sit with before the next AI request turns into accidental scope.
That is the whole danger of the mandate trap. Urgency shows up wearing strategy’s jacket, and unless the front door holds, the team inherits the mess.
Much like the human side of working with AI, the mission-critical skill is not just moving faster. It is exercising judgment intelligently before speed turns a bad decision into an expensive one.
Great one Mark. I have first hand seen how AI amplified what I was already weak at. I like the idea of fitting things into existing rituals for a start!
Chintan, thank you! That is exactly the part @Joel Salinas and I were trying to name in this piece.
AI does not usually create the weakness from scratch. It makes the existing weakness louder, faster, and harder to ignore.
That is why fitting the work into existing rituals matters so much. If the team already has a planning meeting, a review cadence, or an intake conversation, the goal is not to invent a giant new governance machine.
The goal is to add enough structure to the moment where decisions already happen.
A five-minute intake check inside an existing ritual will beat a beautiful process nobody uses every time.
Yeah such a useful and actionable way to go about it. Reminds me of habit formation, and this is what’s often recommended for starters. Because adding completely new things from scratch tends to face a lot of resistance!
Exactly, Chintan. Habit formation is the right analogy!
The more “new process” energy something has, the more the organization’s immune system reacts to it.
That is why I think AI governance has to start small and attach itself to moments the team already recognizes: planning, review, intake, prioritization, approval.
A tiny decision check used every week beats a grand operating model nobody has the stamina to maintain.
@Juan Salas-Romer Thanks for saying so! There are many ways to build an intake funnel. I find this five-step one to be the most universally successful. Which is why it’s one of the centerpieces of my upcoming book!
Joel, thank you again for the opportunity to collaborate on this.
One of the things I appreciate most about your work is that you do not treat AI adoption as a toy box or a panic button. You keep bringing the conversation back to leadership, judgment, implementation, and what teams can actually do on Monday morning.
That made this piece a real pleasure to contribute to.
My hope is that the “front door” framing helps leaders pause before the next AI pilot quietly becomes permanent scope. Because the issue is rarely whether people are excited enough about AI.
The issue is whether the organization has enough structure to turn that excitement into accountable decisions.
At the innovation ends of the cycle. Pilots are totally disposable and meant to fail. It's better to title then as "experiments" but finance hates that name as it sounds frivolous and wasteful so pilots it is. 😉
Chris, I agree with the spirit of this, especially the need to distinguish exploratory work from committed delivery.
Where I would sharpen it slightly is that I do not think pilots are necessarily meant to fail. Some absolutely should be disposable. Some should be killed quickly. But some are really MVPs in disguise and can grow into something marketable if the learning is strong enough.
That is why I like the “experiment” framing, even if finance sometimes hears that word and starts looking for the adult supervision.
The key, at least to me, is making the boundary explicit up front:
What are we trying to learn?
What would make us stop?
What would make this graduate into real scope?
Without those answers, “pilot” becomes a politically safer word for “commitment we have not admitted yet.”
After setting up AI systems for dozens of small business owners, I have seen all three of these failure patterns play out. The one that hits hardest is treating AI as a one-time installation rather than an ongoing practice. Which structural failure do you see most often in the businesses you coach?
Ikram, yes. The one I see most often is intake collapse, or more honestly, the absence of any legitimately formalized intake process.
The example from the piece is the VP asking for an AI feature to summarize customer calls. In the unhealthy version, that request becomes “strategic” because it came from above. By the time anyone asks basic questions, the pilot has already started behaving like permanent scope.
The healthier move is to route it through intake first:
Who owns the data risk?
What does success look like?
What is the experiment boundary?
What gets delayed if we say yes?
Your point about AI being treated like a one-time installation lands for the same reason. Once the system goes live, new requests keep arriving. Without intake discipline, the AI tool slowly becomes another overloaded process with better branding.
This “front door is the governance moment” framing really lands. So many AI pilots go sideways not because the technology is unworkable, but because enthusiasm gets converted into scope before anyone has named the owner, success threshold, risk, or kill criteria.
Alex, exactly. You named the danger cleanly.
The issue is not enthusiasm itself. Teams need that. The danger is when enthusiasm hardens into scope before anyone names the owner, success threshold, risk, or stop condition.
That is why I keep coming back to the front door. Sometimes governance is the moment a shiny idea is required to become a real decision before the team inherits it.
For this to work, you also need to convince senior leaders that their quick asks must go through this funnel. Just sayinnnggg
100%!
And for that to happen, you need a champion who either out ranks or is a direct peer in order to assure they comply with the funnel. Not always an easy ask but essential.
The mandate arrives with urgency already attached the intake system is supposed to be the filter, not the rubber stamp.
John, I love how cleanly you named that.
“The intake system is supposed to be the filter, not the rubber stamp” is exactly the line leaders need to sit with before the next AI request turns into accidental scope.
That is the whole danger of the mandate trap. Urgency shows up wearing strategy’s jacket, and unless the front door holds, the team inherits the mess.
Much like the human side of working with AI, the mission-critical skill is not just moving faster. It is exercising judgment intelligently before speed turns a bad decision into an expensive one.
Very well said, John!
Great one Mark. I have first hand seen how AI amplified what I was already weak at. I like the idea of fitting things into existing rituals for a start!
Chintan, thank you! That is exactly the part @Joel Salinas and I were trying to name in this piece.
AI does not usually create the weakness from scratch. It makes the existing weakness louder, faster, and harder to ignore.
That is why fitting the work into existing rituals matters so much. If the team already has a planning meeting, a review cadence, or an intake conversation, the goal is not to invent a giant new governance machine.
The goal is to add enough structure to the moment where decisions already happen.
A five-minute intake check inside an existing ritual will beat a beautiful process nobody uses every time.
100 percent. Thanks for breaking it down in detail :)
You're welcome, Chintan. Happy to oblige
Yeah such a useful and actionable way to go about it. Reminds me of habit formation, and this is what’s often recommended for starters. Because adding completely new things from scratch tends to face a lot of resistance!
Exactly, Chintan. Habit formation is the right analogy!
The more “new process” energy something has, the more the organization’s immune system reacts to it.
That is why I think AI governance has to start small and attach itself to moments the team already recognizes: planning, review, intake, prioritization, approval.
A tiny decision check used every week beats a grand operating model nobody has the stamina to maintain.
Joel. The mandate trap is so real. The same thing happens to solo operators with client requests. The five-step funnel fixes both.
Thank you Juan! It really is a trap
@Juan Salas-Romer Thanks for saying so! There are many ways to build an intake funnel. I find this five-step one to be the most universally successful. Which is why it’s one of the centerpieces of my upcoming book!