I spent two decades learning how to run technology at scale. IBM. Kyndryl. 200+ engineers across the globe on a single account. I know the enterprise playbook cold. The governance boards, phased rollouts, RACI charts, steering committees, the whole apparatus.
I’ll tell you something uncomfortable. That same playbook is quietly strangling AI projects across the country right now.
Not because the discipline is wrong. Because it’s being applied to the wrong problem, at the wrong speed, by people who mistake the ritual of enterprise process for the substance of it.
The Playbook That Worked for Everything Else
Here’s why the enterprise playbook exists in the first place. When you’re deploying a new ERP system or migrating a data center, the technology is a known quantity. The vendor has done it a thousand times. The risk isn’t “will this work” it’s “will we screw up the rollout.” So you build process around coordination. You stage. You pilot. You get sign-off at every gate.
That’s rational. For that kind of work, it’s the difference between a clean cutover and a career-ending outage.
But AI isn’t that kind of work. Treating it like it is and that’s the mistake I watch smart, experienced leaders make over and over.
The Speed Mismatch Nobody Names
Enterprise process is built to slow things down on purpose. That’s the point. You add friction so that no single person can move fast enough to break something expensive.
Now drop that machine on top of an AI initiative.
The technology underneath you is changing every quarter. The model you scoped in January is obsolete by the time your governance board approves the pilot in June. By the time you clear security review and stand up the production environment, you’re building on a foundation that’s already two generations old.
I watched a mid-sized firm spend eleven months getting an AI project through their internal gates. Eleven months. The tool they finally deployed had been leapfrogged twice. They ran a flawless enterprise process and shipped an obsolete product. Everyone did their job. The project still failed.
That’s the trap. Enterprise discipline optimizes for not being wrong. AI right now rewards being fast enough to learn before the ground shifts again.
The Governance Theater Problem
Here’s the part that stings. A lot of what passes for enterprise AI governance isn’t governance. It’s theater.
Real governance answers hard questions. Where does this data live? Who can access the model’s outputs? What happens when it’s wrong, and who’s accountable? What’s our exposure if it leaks something it shouldn’t?
Theater answers easy ones. Did we hold the meeting? Is the committee chartered? Did legal sign the box? Are we compliant with the framework we adopted from a vendor slide deck?
I’ve sat in rooms where a team spent six weeks producing an AI governance document that never once addressed their actual data pipeline because nobody in the room understood the pipeline well enough to ask. They governed the paperwork. The infrastructure went ungoverned and that’s exactly where these projects die.
If your governance process would pass an AI project that’s sitting on duplicate records and disconnected systems, your governance process is theater. It’s checking that the ritual happened, not that the thing is actually safe to build on.
What the Enterprise Playbook Gets Backwards
The enterprise playbook front-loads the wrong things and back-loads the ones that matter.
It front-loads approval. Committees, gates, sign-offs, all the coordination machinery comes first, before anyone has learned anything.
It back-loads the infrastructure truth. The hard look at whether your data is actually usable, whether your systems can actually talk to each other, whether your security posture can actually survive an AI system reaching into it that gets pushed to “implementation,” which is to say, after the budget is committed and the political capital is spent.
That’s backwards. On the AI work I’ve done including the Kyndryl engagement that produced a 38% improvement in incident detection and over $2M in documented savings for the client the wins didn’t come from a tighter approval process. They came from doing the unglamorous infrastructure work first, before anyone got near a model. Data readiness. Integration. Security. The stuff no steering committee gets excited about.
The enterprise instinct is to govern the decision. The right instinct is to prove the foundation. Those are not the same activity, and confusing them is why seven-figure initiatives stall at ninety percent.
Why Smaller Companies Are Quietly Winning
Here’s the irony I keep running into. The businesses moving fastest and cleanest on AI right now often aren’t the enterprises with the mature playbooks. They’re smaller companies that never had one.
They can’t afford eleven months of gates. So they do the only thing that actually works. They figure out fast whether their foundation can support the thing they want to build, they fix what’s broken, and they ship. No steering committee. No governance theater. Just, is the data usable, can the systems connect, is it secure, go.
That’s not recklessness. Done right, it’s more rigorous than the enterprise version, because it puts the hard questions first instead of hiding them behind process. The small company that spends three weeks brutally interrogating its own data readiness is further ahead than the enterprise that spends eleven months interrogating its own approval flow.
The enterprise playbook made you feel safe. It didn’t make you ready.
So What Do You Actually Do
I’m not telling you to throw out discipline. I ran 200+ engineers and I’d be the last person to tell you process doesn’t matter. I’m telling you to point it at the right target.
Govern the foundation, not the calendar. Ask whether the data is real before you ask whether the committee is chartered. Compress the approval machinery and expand the infrastructure scrutiny. Move the hard questions to the front, where they can still change the outcome, instead of the back, where they only assign blame.
The playbook that migrated your data center will not deploy your AI. It was built for a world where the technology was the safe part and the rollout was the risk. AI inverts that. The rollout is easy. The foundation is where you live or die.
So here’s the choice, and it is a choice. You can run the process that makes everyone feel covered and ship something obsolete eleven months from now, or you can put the infrastructure truth first, move at the speed the technology actually demands, and build something that works.
Your old playbook can’t do both. Which one are you actually running?
Ready to find out if your foundation can support what you’re trying to build? Start with an AI Infrastructure Assessment — a readiness scorecard, data pipeline analysis, security evaluation, and a prioritized roadmap before you commit a dollar to a model.
See how we sequence AI work foundation-first on our Services page — assessment, strategy, implementation, and adoption, in the order that actually holds.
Want more like this? The rest of the Summit AI blog breaks down why AI projects really fail — and how to be one of the ones that doesn’t.
Russell Love is the Founder & CEO of Summit AI Business Solutions, based in Browns Summit, NC. With 20+ years of enterprise transformation experience at IBM and Kyndryl, Russell helps businesses build the foundations that make AI actually work.