In the first post of this series, I asked whether humans and AI might merge into something new, a composite organism shaped by selection pressure and mutual dependence. That question matters because the merge is already happening. Not as science fiction, but as organizational reality. The question becomes: what kind of merger are we choosing?
2024 was the year of AI pilots. 2025 was supposed to be the year of AI value.
For most organizations, it wasn't.
Not because many pilots failed, but because they were never designed to graduate.
A familiar pattern
I've watched this pattern before.
In the late '90s, everyone knew the internet would change everything. We sold dial-up connections to farmers in Eastern Washington who met us at the door with cookies, thrilled to send mail without stamps. Within a decade, we'd built fiber networks passing 65,000 homes, and I was fielding angry calls from people complaining about "only" getting 98% of their 100-meg connection.
The technology delivered. But the gap between promise and value was always organizational, who owned the implementation, who was accountable for outcomes, who did the hard work of integration.
AI is following the same arc, compressed into months instead of years.
The annual reports are in. Deloitte, OpenAI, Anthropic, Google, Microsoft, the UK AI Safety Institute, everyone who tracks enterprise AI adoption has published their findings. The consensus is clear: the question is no longer whether AI will reshape how organizations operate. That debate is over.
The new question is harder
Why did so much investment produce so little operational change?
The reports point in the same direction. Organizations ran pilots that were really more like demonstrations. They demonstrated capabilities. They impressed stakeholders with what AI could do. Then they moved on to the next pilot.
The problem wasn't the technology. The problem was that most were designed to demonstrate capability, not prove scalability. They were built to impress, not to integrate.
A pilot without graduation criteria isn't a pilot. It's an experiment with no hypothesis.
AI amplifies organizational maturity
Here's what the reports make clear: AI amplifies organizational maturity. It doesn't create it.
You can see this pattern playing out in real time. LinkedIn is full of comic strips parodying CEOs who chase AI while neglecting the organizational foundations that make technology work. The punchline is always the same: executives excited about capabilities, oblivious to, well, human resources. The tragedy is that while we're laughing, people inside these organizations are trying to make impossible projects succeed with inadequate support.
My dad used to say, "There's no right way to do a wrong thing." He was talking about integrity, about how you can't fulfill your responsibilities to family and community if you've compromised the foundation everything else rests on.
The same principle applies to organizations trying to scale AI. You can't automate your way past broken trust, unclear accountability, or fragmented decision flows.
AI doesn't fix those problems. It amplifies them
Organizations where decisions already flow smoothly, where accountability is clear, where handoffs are clean, where trust exists between functions, those organizations scaled AI quickly. The technology accelerated what was already working.
Organizations where decisions get stuck, where ownership is ambiguous, where information decays between handoffs, where functions protect their own data, those organizations found that AI amplified the dysfunction. The pilots looked impressive. The integration failed.
The bottleneck was never the model. It was the org chart. More precisely: it was the disconnect between how the org chart says decisions flow and how they actually flow. Between formal authority and practical influence. Between who's supposed to own outcomes and who actually does. AI doesn't navigate these contradictions. It exposes them, brutally.
This isn't a technology problem
It's an operating model problem.
The organizations that captured value from AI didn't have better models or bigger budgets. They had clearer decision rights. They knew who owned what, who handed off to whom, and who answered when something broke.
They also approached pilots differently. Every pilot had success criteria defined before it started. Every pilot had a scale path. Every pilot was connected to a real workflow with real outcomes that someone was accountable for measuring.
Pilots that matter have three things: an owner, a measure, and a scale path. Everything else is exploration.
The reckoning
The reckoning isn't about whether your organization experimented with AI. Nearly everyone did. The reckoning is about what you have to show for it.
If your AI investments still live in slideware, impressive demos, compelling proofs of concept, enthusiastic presentations to the board, then 2024 was expensive practice. The question for 2026 is whether you'll practice again or build something that lasts.
The shift required isn't technical. It's organizational.
It means designing pilots for graduation, not applause. It means clarifying decision rights before automating decisions. It means measuring outcomes, not activities. It means building trust, between functions, between organizations, between systems, as deliberately as you build technology.
The organizations that scaled AI fastest weren't the ones with the best technology. They were the ones where trust already flowed. Where decisions moved without friction. Where accountability was clear before automation made it essential.
Where the gap closes
I work in fresh supply chains, produce, proteins, dairy, the cold chains where perishability doesn't negotiate and margins are thin. It's an unforgiving environment for technology pilots. If the system doesn't work in the real flow of product from farm to shelf, it doesn't matter how impressive the demo was.
What I've learned there applies everywhere: the gap between AI potential and AI value is organizational. Close the gap, and the technology works. Leave it open, and you're running pilots forever.
The experimentation phase is over. The question now is: do you have the organizational foundations to actually build? Or are you running another year of "impressive" AI projects that lead nowhere?
The next post in this series examines what those foundations look like, not in theory, but in the architecture of systems designed for integration rather than demonstration.
The question isn't whether AI will reshape your operating model. It's whether you'll do it intentionally, or have it done to you.