Your Ways of Working Are the Bottleneck, Not Your AI Tools

So your organization has moved past the foundational work of preparing for AI. SOPs written and in place. Governance well defined. Information security and technical guardrails set. Tools standardized and deployed. Use is on the rise.

Ready to accelerate, right? Maybe not.

AI tool adoption moves lightning fast in the hands of experts. And I'm not talking about surface-level use. Put any of these tools (Claude, Copilot, Cursor, Replit, etc) in front of strong engineers, product leaders, UX, or other technology professionals and you'll see productivity accelerate almost immediately. That's one of the beautiful things about what AI brings that's unlike prior technical revolutions: you can impact productivity right away, vertically. When I say vertically, I mean within the function or silo.

The harder question, and the one I've watched teams struggle with, is how you accelerate what matters most: the outcome. McKinsey's 2026 State of AI survey puts a number on the gap. Eighty percent of respondents say AI has improved their individual productivity. Thirty-seven percent report that AI has contributed to their organization's EBIT, essentially unchanged from a year ago.

The vertical is working. It just doesn't always translate to the bottom line. Here's why: productivity is immediately available to the function. The outcome isn't. Optimizing a silo for productivity is not the same thing as accelerating the outcome. A function can double its throughput and the enterprise can feel nothing, because the outcome doesn't live inside any one function. It typically crosses many of them. Speed up a step that was never the constraint and you haven't moved the finish line. You've just piled up work in front of the next handoff.

The fix is a leadership problem before it's a process one. The impact on your organization's ways of working needs to be understood first. That means answering hard questions inside your team, and many more that sit outside it. It takes leaders who can explain what AI changes about your job and how you work. This is human. This is hard. Those of us who have led large-scale transformation know the change management dynamics involved in translating plans and vision into outcomes. AI is no different in the change it demands, which is exactly why leadership is so critical. For an individual or small R&D team, a startup, or a small firm, this may be less of an issue. With a handful of people, there are barely any handoffs to get wrong. But if you're part of a large enterprise, especially a highly regulated, process-intensive one, it's not that simple. It can be, though.

Working horizontally, across functions to deliver a particular value stream, is not a new concept. Rummler and Brache were writing about the white space on the org chart more than thirty years ago in Improving Performance, arguing that the critical interfaces live in the white space between the boxes, and that most process failures happen at those handoffs rather than inside any one department. What's different now in the age of AI is the stakes. Get this right and the acceleration available to us has no precedent in our history.

What's also different is what's moving into that white space. As organizations climb the AI maturity curve and agents begin running workflows, who coordinates decisions and knowledge across functions, if anyone? Writing in Harvard Business Review in "How AI Agents Orchestrate Work Across Silos," Ferreira and Tong call this the hidden bottleneck: the decisions that matter most span multiple functions, carry interacting constraints and data from many sources, and depend on judgment built from experience rather than from a spreadsheet. And that judgment often has to be applied in situations where the downstream effects are genuinely uncertain. Their answer points to agentic AI orchestration, systems that connect the tools to the workflows and pull in what only humans know. But I digress. Whether the handoff is human to human, human to agent, or agent to agent, somebody has to understand how the work actually moves. That doesn't get easier with agents in the mix. It gets harder, and more expensive to get wrong.

To get there, leaders need a vision of how their teams' roles change and how the organization's ways of working are affected (i.e., how they deliver value). These are handoff questions, not tooling questions. And they are critical.

A fair objection:

As organizations move up the maturity curve, capability shifts left. Doesn't that collapse the handoffs on its own?

Not quite. Shifting work left can shorten time to value, but only if you're able to answer who owns the decisions. Technical capability moving into a role doesn't mean the expertise moves with it. A designer who can ship code has not become an architect. They can build the thing. They may not know what it breaks three systems downstream. The capability crossed the line. The judgment didn't.

If your organization is deliberately consolidating responsibilities, that's a strategy, and a separate conversation. I'd just want to know it was a decision and not a drift. But if you intend to keep decisions with the experts who are accountable for them, you have to define where capability ends and authority begins. Otherwise the handoff doesn't disappear. It just stops being visible to anyone who could manage it. The evolving AI-Driven Development Lifecycle, or AI-DLC, is one way to organize these cross-functional workflows. More on that next time.

Here's my point in all this: the tools will deliver productivity on their own. Your ways of working determine whether that shows up in the outcome. And in the end, that's what it's all about. The vertical is easy, and it will happen almost organically. The outcome won't. What are you doing to help your organization move horizontally?