Your AI strategy has a trust problem

Her recent newsletter piece on what actually blocks AI adoption inside companies is one of the better-argued takes we've seen this year. The argument isn't about tools. It's about organizational trust, and she makes the case in a way that's worth working through carefully.

The diagnosis Verna is actually making

The newsletter opens with a frame that will feel familiar to anyone who has watched an "AI transformation initiative" get announced, celebrated, and quietly stall. Everyone gets excited about high-impact individual contributors and speed as a competitive moat. Then the approval cycles kick in. The ticket queue grows. The initiative gets a project manager.

Verna's diagnosis is specific: most modern companies are built on a command-and-control model that treats each employee as an assembly-line worker with one defined function. The structure's whole point is to bring order to large organizations with lots of moving parts. Predictability and control are the features, not bugs. The problem is that this structure is fundamentally incompatible with the kind of fast, autonomous, iterative work that AI-native teams need to do.

The tells she names are recognizable:

• Exhausting approval cycles that add days to decisions that should take minutes
• Tight role boundaries that stop people from acting outside their defined function
• Title-based hierarchies that route decisions up regardless of who has the relevant context
• Middle management layers whose primary function is keeping people in line, not accelerating output

Each of these sends the same signal to employees: we don't trust your judgment, so we'll make the decision for you.

Why this matters more now

An employee with a well-configured AI assistant can draft, research, test, and ship in the time it used to take to get calendar alignment for a kickoff meeting. But if that employee needs four approvals to publish a landing page, the AI just means they sit on finished work faster. The bottleneck moves upstream, not downstream.

Verna's formulation is the clearest version of this we've seen: "Everyone's talking about AI agents, but what you really need is employees with agency." The double meaning of "agent" is deliberate. You can deploy autonomous software agents into a system that doesn't trust the humans operating it, and what you get is an automated version of the same slow, risk-averse process you already had.

What "employees with agency" actually requires

Verna doesn't just diagnose the problem; she gestures at what the fix looks like. The structure needs to change in specific ways.

Decision rights need to move closer to the work. The person with the most relevant context should have the authority to act. Title shouldn't determine who decides; proximity to the information should. This is uncomfortable for organizations where title-based authority is a core part of how trust is distributed.

Role boundaries need to expand. The assembly-line model works when the work is genuinely repetitive and the cost of deviation is high. It doesn't work when the best move is for a designer to fix the copy, or for an engineer to run a customer call, or for a marketer to write a basic SQL query. AI tools make cross-functional action easier; rigid roles make it forbidden.

Middle management needs a different job description. If the current function is compliance and oversight, that function gets automated before most other functions do. The managers who survive this shift are the ones already operating as context providers and blockers-of-blockers, not as the people whose approval keeps everything moving.

None of this is simple. Verna acknowledges the original logic of command-and-control: large organizations with lots of moving parts really do need coordination mechanisms. The argument is that the coordination mechanism needs to change, not that coordination itself is the problem.

What operators building inbound pipeline take from this

For the founders, consultants, and solo operators we work with, a lot of this translates directly. The "org structure problem" plays out at smaller scale, but it's still real.

We see it most often as self-imposed approval cycles. Someone writes a sharp LinkedIn comment, then waits to see if a collaborator approves the angle before posting. Someone drafts a newsletter edition and sits on it for two weeks because the framing feels risky. Someone wants to respond to a prospect's post but checks with their business partner first. The speed advantage of being a small, expert operator evaporates in proportion to how much internal permission-seeking they do.

The operators with the most consistent traction, including Verna herself, whose LinkedIn presence and patterns top creators share we've covered before, removed the friction between thinking something and publishing it.

“Everyone's talking about AI agents, but what you really need is employees with agency.”

The part of Verna's argument that gets underweighted

The newsletter is strongest in its diagnosis and lightest on the transition path. She's right that the org structure needs to change, and she's right that "AI transformation initiatives" are a poor vehicle for that change. Where it gets harder is the practical question of how you move a 500-person company from command-and-control to high-agency without chaos.

The honest answer is probably: you don't move it all at once. You find the pockets where high-agency teams already exist, protect them, and let the rest of the org watch what happens. In most companies, those pockets are in product and engineering, where the connection between individual action and measurable outcome is tight enough that trust gets built empirically. The harder cultural work is extending that model into functions where the feedback loops are longer.

The underlying bet

When you give talented individuals the access, the context, and the authority to act, they will outperform any carefully managed process you could design around them. AI makes the talent-to-output ratio more extreme, which means the cost of constraining that talent is also more extreme.

If that bet is right, the companies that win the next five years are not the ones who deployed the best AI tools. They're the ones who redesigned their trust infrastructure before their competitors did.

For operators building their own presence and pipeline, the same principle applies at the level of the individual. The tool is not the constraint. Whether you've given yourself permission to act is.