The AI Capability Gap You Already Identified
Every leadership team I talk to is asking some version of the same question this year. We bought the AI tools. We funded the AI training. Why aren’t we seeing the productivity gains in the room?
Most of the answer being offered is more AI. More licenses. More prompt training. More custom GPTs. A few companies have gone the other direction and put capability work ahead of any structural change at all, betting that if they get their people ready, the structure follows.
I think both moves miss the same thing.
The capability that decides whether your AI investment pays off isn’t a new capability. It’s the one already on your L&D roadmap. The one your CHRO has been trying to build for a decade. Framing a problem worth solving. Holding a point of view that survives a senior stakeholder pushing back. Influencing people across functions who don’t report to you.
In recent discovery conversations, I kept asking L&D leaders what was actually blocking AI adoption inside their companies. Nobody said the interface was too complex. Nobody said the model wasn’t good enough. Nobody said cost or access. What came up every time was people. People who couldn’t frame a problem worth solving. People who couldn’t land a point of view with the team that needed to act.
People who can’t frame a problem in a meeting can’t frame it in a prompt either.
That’s the argument. The rest of this post unpacks it.
You’re paying twice for the same capability
In most companies right now there are two parallel workstreams. One is “human skills” or “leadership development” or whatever this year’s version of the manager curriculum is called. It includes things like critical thinking, communication, stakeholder influence. It has been on the plan for years.
The other is “AI enablement.” It includes prompt training, tool rollouts, AI literacy modules, sometimes a custom GPT or two.
These two workstreams have separate budgets, separate vendors, and often separate owners. The L&D team owns one. A digital or transformation team owns the other.
The skills inside them are almost identical.
Framing a problem clearly is what makes a prompt useful. Holding a point of view is what lets a person evaluate model output instead of accepting the first plausible-looking answer. Influencing across functions is what turns an AI-generated insight into a decision the rest of the team will act on.
If you separated those workstreams, you are paying twice for the same capability and getting it from neither.
What the org chart was actually solving for
The org chart was never really about power or control. It was built to solve an information problem. In a large organization, how do you move knowledge to the right people and get decisions back out? Structure existed because information was scarce and routing it was expensive. Management theorists have been making this argument for decades.
AI centralizes intelligence. Anyone in the organization can now access what used to require escalating three levels. The routing function, which is most of what middle management did, is becoming redundant.
Org chart compression isn’t a strategic choice. It’s where intelligence is flowing.
The work that remains is the work that can’t be automated. Building trust up to leadership while holding your team’s reality. Influencing peers across functions. Developing the people whose growth you’re responsible for.
That’s relational and cognitive work. Not technical. And it’s the work organizations have been trying to build capability around since long before any of us heard the word transformer.
Two companies that stopped buying AI training
An L&D leader at a mid-market tech company. His CFO announced a company-wide AI efficiency mandate at all-hands. His response wasn’t to roll out tool training. His observation was simple: the tool landscape moves too fast. There’s a new platform every few months. Training people on Tool A guarantees retraining on Tool B by next quarter.
So he shifted the entire curriculum. Teach motions, not tools. How to frame a problem. How to ask a good question. How to evaluate the output you get back. Those capabilities survive every platform change, and they happen to be the capabilities that determine whether anyone gets value from the AI investment in the first place.
A fast-growth tech company. Seventy-five percent of managers had been promoted from IC roles in the last three years. When the company went through a significant strategy shift, the managers who struggled weren’t missing AI skills. They were still operating as information routers. Escalating concerns upward rather than stewarding them. Translating strategy into “here’s what leadership said” instead of “here’s what I believe we should do.”
The intervention wasn’t AI training. It was reframing the manager role from information routing to relationship stewardship. The same capability shift that would let those managers get more out of any AI tool you handed them.
Both organizations stopped treating AI capability as a separate workstream. They folded it into the work they were already doing on people.
The bottom line
Companies that treat the human capability work and the AI enablement work as one investment will see compounding returns from both. Companies that keep them separate will keep funding two roadmaps and getting one disappointing result.
One question
Ask your head of talent this week. The capabilities we’ve been trying to build for the last decade, framing, point of view, influence, are they on the AI roadmap? Or did we file them under “human skills” and start a separate workstream for “AI skills”?
If they’re separate, you’re paying twice for the same capability. The cheapest move you can make right now is to stop.
