In 2024, I named my company for a disruption I couldn’t prove yet.
The name came at the end of a decision, not the beginning of one. Two weeks abroad, four countries in, and mentally detached from the real-world, I solidified my plan to leave the grind and build my own company. I was sitting on a balcony in Croatia overlooking the Adriatic, brainstorming business models. My head was clear but, it was a short session, because I could already see where AI was heading and the disruption about to hit every business.
This left me with the one question that mattered most:
How the hell do you build a company around a single business model at a moment of complete unpredictability?
That question consumed my until the answer took shape. I didn’t need a business model as much as I needed a moldable one, a company built to adapt at the speed of AI as it evolved and as I learned. Two months after deciding to bounce from corporate, the name hit me. I bought the PerformanceLabs.AI domain immediately, and got to work.
The logic behind the name was simple, even if the conviction behind it wasn’t. I knew whatever this company became, it wouldn’t be a single product. If AI went the way I believed it would, the future of business wouldn’t look like a roadmap. It would look like a laboratory: iterative, experimental, willing to kill what doesn’t work. That’s the Labs.
The Performance part came from 25 years in RevOps leadership roles. That career taught me one durable truth: technology never matters on its own. The only thing that matters is whether you can operationalize it to improve performance.
Everything else is a demo.
Here’s the honest part. I thought it would all happen faster. I expected the disruption to hit organizations like a wave in 2025. Instead, the models moved at light speed and the organizations moved at the speed of human behavior change, which is to say, the speed organizations have always moved. Two years later, in 2026, the wave is finally landing. Companies are recognizing this moment for what it is and racing to capitalize on it.
That gap, between how fast the technology moved and how slowly organizations absorbed it, turned out to be the entire story of the last two years. And it’s the story I want to tell on our second anniversary, because I’ve had a front-row seat.
Looking out vs. looking in
Since launching PerformanceLabs, I’ve had more than 300 conversations with operators and leaders about AI. Contacts and contracts, prospects and clients, workshops and war rooms. Across all of them, one pattern explains more than any other.
For two years, companies learned about AI by looking out. At vendors. At competitors. At headlines and conference keynotes. The shift happening right now, the one that separates the companies pulling ahead from the ones still circling, is that the serious ones have started looking in. At their own workflows. Their own data. Their own people.
Every company I talk to is in a different place on that journey, and that’s fine. But the journey itself has been remarkably consistent. Here’s what it looked like, year by year.
2024: The year of talking
In 2024, AI was a topic, not a practice. The conversations I had that year were full of tool names and empty of workflows. Leaders could rattle off five AI products they’d read about and couldn’t name one process in their own business that had actually changed.
One moment stands out. I was sitting with a CEO who was rambling about how his company used Microsoft Copilot and was therefore ahead of the AI curve. It was the first time I asked the question I would go on to ask dozens of times over the next two years: “What revenue impact do you expect AI to contribute next year?” Blank stare, defensive posture, and meandering non-answer.
The data backs up the memory. In 2024, only about half of enterprises had even a single AI workload in production, according to McKinsey's global survey work, and Gartner found the average AI project took eight months just to move from prototype to production. AI was a line of conversation, not a line item. The gap between the noise and the deployment was enormous, and confusion filled it.
I don’t say that to dunk on anyone. Confusion was the rational response. The technology was changing monthly, the vendors were overpromising, and nobody had a map. But it meant that 2024’s defining activity was looking out, hard, at everyone else.
The Operator’s Hour
A working session on applied AI. Real tools, real numbers, and real-time builds.
2025: The year of shadow adoption
In 2025 I ran roughly 30 AI workshops, and they taught me more than anything else I did that year. Walking into those rooms, the pattern was unmistakable: AI adoption was already happening. It just wasn’t happening on the org chart.
Teams were using AI every day. There was no oversight, no budget, no policy, and often no awareness at the leadership level. The infancy of organizational AI was on full display: enthusiastic, ungoverned, and bottom-up.
In one of those sessions, a sales rep showed me a prospecting tool she had built to generate key prospect lists. Her CRO had no idea it existed. Neither did the rest of her team. Here was a rep who understood the benefits of AI and was using it to create real efficiencies for herself, with leadership completely out of the loop.
Again, the numbers match the rooms. Surveys that year found 78 percent of employees using AI tools without IT approval, while only about 15 percent of organizations had any governance policy in place. One study found that 68 percent of security leaders admitted to unauthorized AI use themselves. The people responsible for the rules were breaking them.
Here’s the charitable read, and I think it’s the correct one: employees weren’t rogue. They were ahead. Adoption went bottom-up because top-down hadn’t shown up yet. Your team didn’t wait for a committee because the value was sitting right there in the browser, and they had work to do.
The lesson I took from 2025 is one I still repeat to clients: shadow adoption isn’t a compliance failure. It’s a demand signal. The companies that treated it as data got ahead. The ones that treated it as disobedience lost a year.
2026: The year of budget and whiplash
This year, the money showed up. AI now averages roughly 15 percent of enterprise IT spend, up from 4 percent two years ago, and about two-thirds of enterprises increased their AI budgets again this year. The question stopped being whether to invest and became how.
And with the money came the whiplash. Three patterns define the 2026 conversations I’ve had so far.
First, companies got the hires wrong. The instinct was to hire technical: engineers, data scientists, people who could build. What organizations actually needed were operators, the people who know where the exceptions live in a workflow and can point the technology at the right problem. I wrote about this a few weeks ago in “Your Best AI Operator Isn’t an Engineer,” and nothing in my client work has contradicted it since. Building is now cheap. Knowing where to point the build is the scarce input.
One example stands out: watching a company move its most “AI-capable” employee into a newly created Head of AI position. This person was talented and understood AI better than most, but had never operationalized technology inside an organization. So they built some genuinely cool things, in a vacuum, with no focus on the work that actually impacted the organization at large.
Second, the AI committees dissolved. And honestly, good. Committees were a looking-out structure, built to study a phenomenon happening elsewhere. Ownership is a looking-in structure, built to change work happening here. When I see a company disband its AI committee and hand the mandate to a named operator with a budget, I read it as maturity, not failure.
Third, urgency is compounding, and it’s changing build-versus-buy math. Teams bought tools all year, but as the year has gone on I’m watching a quieter movement: companies starting to detach from parts of their SaaS stack because they can now build the workflow themselves, shaped exactly to their business.
This is already happening, but the bigger the company, the slower the movement. Which makes sense, given the hooks SaaS providers have into their organizations. The ones moving fastest are in the mid-market: nimble, forward-leaning, and not mired in enterprise SaaS contracts. The speed is palpable.
That third pattern is the one I’d watch most closely, because it hints at where 2027 goes.
2027: The philosophical year
Here’s what I believe comes next, and it isn’t just another new model release.
The technical era of this transition is ending. From 2024 through 2026, the operating question was “can we?” Can the model do it, can the tool connect, can the team learn it. In 2027, the operating question becomes “should we, where, and who decides?” That’s not a technical question. It’s a philosophical one, and it lands on every leader’s desk, not just the CIO’s.
Three forces drive the shift. Workflow builders will keep democratizing creation, which means more people building more things that touch more data. The IT security and compliance layer will move from afterthought to center stage, because it has to: Gartner projects that 40 percent of companies will experience security incidents related to unsanctioned AI use by 2027. And the failures will keep teaching the same lesson. Research from IDC and Forrester finds that roughly 88 percent of AI pilots still never reach production, and the failures cluster on governance, data readiness, and observability. Not model quality.
Read that again, because it’s the most important finding of this entire two-year arc: the constraint is no longer the technology. It’s the organization.
Ethics, data boundaries, accountability, disclosure. In 2027 these stop being the compliance team’s problem and become operating decisions inside every workflow. The companies that treat that as a burden will slow down. The companies that treat it as design criteria will build things their competitors legally and culturally can’t.
And here's my contrarian call for 2027: New AI leaders will shift from technologists to org-wide operators. The Head of AI, the AI czar, the Chief AI Officer titles companies are racing to create right now. I watched organizations get this hire wrong all through 2026, and the root cause wasn't the person. It was the premise. AI is not a department, and 2027 is the year the org chart figures that out. Nobody employs a Chief Internet Officer anymore, because the internet stopped being an initiative and became the way business works. AI is on the same path, just faster.
The companies that get 2027 right won't have an AI team off to the side. They'll have operating leaders accountable for AI inside the workflows they already own, sitting on top of a hardened security and compliance layer.
Back to the lab
Two years in, it turns out I got the name right for a reason I didn’t anticipate.
I chose “Labs” because I thought my company would need to be a laboratory. What I’ve learned from those conversations is that every company now has to become one. Iterative. Evidence-driven. Comfortable running experiments on its own operations. Willing to kill what doesn’t work and scale what does, without waiting for the market to hand down a playbook, because there is no playbook. There’s only your workflows, your data, and your people.
That’s the real meaning of looking in.
What encourages me most, two years into this, is simple: nobody is just talking anymore. Every company I meet is implementing something. They’re in wildly different places, and the honest ones admit it. But the direction is set.
I got the speed wrong in 2024. I got the prediction right. And if the last two years taught organizations to look in, the next one will test what they find there.