Why technology leaders are losing the AI conversation to the people who report to them

I keep seeing a version of the same scene. A CEO has a question about AI. It is a real question, the kind that will shape where the company spends the next two years. The CEO does not bring it to the CIO. They bring it to a data leader two levels down, or to a vendor who presented at a conference, or to an AI specialist a board member recommended. The CIO finds out the strategy is forming when a slide shows up that they did not build. By then, the direction is already half-set, and the CIO is being asked to react to it rather than shape it.

I want to be precise about what is happening, because it is easy to misread. The CIO has not been removed from anything. Title intact, budget intact, seat at the table intact. What has changed is quieter. On one of the most consequential technology conversations the company will have this decade, the CIO is being routed around. The work still flows through them eventually. The thinking no longer starts with them.

I have watched this happen to capable people who would have given the CEO a better answer than the person who was asked. That is what makes it worth naming. This is not a competence gap. It is a positioning gap, and positioning gaps close in the wrong direction if you ignore them long enough.

How the routing actually starts

The routing does not begin with a decision to exclude anyone. It begins with a CEO who is anxious about AI and looking for someone who sounds certain. AI is moving fast enough that executives feel the pressure to have a point of view before they have earned one. That pressure usually arrives secondhand, from a board member or a peer on the golf course describing what is working at their company. So the CEO goes looking for someone who will confirm the answer they already want to hear, and they keep going back to whoever gives it to them.

Here is where many technology leaders lose the thread. For years, the safe posture in the CIO seat was measured caution. You raised the risks, you flagged the integration cost, you asked who owns the data and what the compliance exposure looks like. That posture built credibility in an era when the failure mode was moving too fast on technology nobody understood. With AI, the same posture reads as drag. A CEO who is being told by three vendors that “the future is already here” does not want to hear why they should slow down and be cautious. They hear caution as losing the race, and they go find a point of view somewhere else.

The data leaders, vendors and specialists who get the call are not necessarily more capable. They are more available with a confident answer. A vendor’s whole job is to arrive with conviction. A data scientist who has shipped one impressive model carries more apparent authority on AI, in that moment, than a CIO who runs the entire estate but talks about AI the way they talk about every other risk. The CEO is not weighing depth against depth. They are weighing the person who said yes against the person who said it depends.

Once that pattern sets, it compounds. The CEO who got a satisfying answer from the data leader goes back to the data leader. The vendor who shaped the first conversation gets invited into the second. Each loop the CIO is not in makes the next one easier to run without them. The org chart still says the CIO owns technology strategy. The actual conversation has relocated.

What it costs before anyone notices

The cost shows up late, which is exactly why it is dangerous. For a while nothing looks broken. The CIO is still delivering. The AI initiatives are still landing on their plate to execute. The damage is happening upstream, in the room where the bets get made, and the CIO is not in that room.

I have seen what arrives downstream when the strategy was set without the person who has to run it. A model gets championed that the data cannot actually support. A vendor commitment gets made that locks the company into an architecture the CIO would have flagged in the first meeting. An agent gets deployed inside a business unit, with executive blessing, and the CIO inherits accountability for it months later without ever having shaped how it was governed. The recent IBM finding that CIOs are increasingly held accountable for AI they do not fully control is the visible end of this. The invisible front end is the conversation the CIO was routed around, the one where the accountability got created in the first place.

What I find most corrosive is what it does to the CIO’s standing over time. Every initiative the CIO executes but did not shape reinforces a story about what the CIO is for. They become the person who runs the technology other people decided on. That is a fine description of an order taker and a poor description of a strategic leader, and CEOs do not promote, fund, or defend order takers when budgets tighten. The routing-around does not just cost the company a worse AI strategy. It quietly recasts the CIO as the implementer of everyone else’s thinking, and that recasting is hard to reverse once the executive team has internalized it.

What the leaders who stayed in the conversation did

The technology leaders I have watched hold their position on AI did one thing first. They stopped leading with caution and started leading with a point of view. Not a reckless one. A real, defensible position on where AI creates value in their specific business and where it does not, delivered with the same conviction the vendors bring, before the CEO went looking elsewhere for it. They made themselves the person with the clearest answer, which is the role the routing-around was filling with someone else.

That requires giving up a posture that felt safe for a long time. The CIOs who made the shift accepted that on AI, being right and cautious is worth less than being early and directional. They formed a view ahead of being asked. They walked into the CEO’s office with where we should place our AI bets and why, rather than waiting to be handed someone else’s bets to pressure-test. The difference is whether you are the author of the strategy or its editor, and CEOs route around editors.

They also changed how they talk about risk. Instead of presenting risk as the reason to slow down, they folded it into the recommendation. The data is not ready for that use case, so here is the use case where it is ready, and here is what we do in parallel to unlock the first one. That framing keeps the CIO inside the conversation as the person making AI happen responsibly, rather than the person standing outside it explaining why it is hard. Same expertise, opposite effect on whether the CEO keeps coming back.

None of this is about pushing the data leaders and specialists out. The strongest CIOs I know pulled those people closer and brought them into the room under their own framing, so that when the CEO wanted the specialist’s input, it arrived through the CIO rather than around them. They made themselves the orchestrator of the AI conversation instead of one of its casualties.

If you are a technology leader right now, the question worth sitting with is not whether you are good at AI. You probably are. The question is whether the most important AI conversations in your company are still starting with you, or whether you have quietly become the person they get handed to after the thinking is done. That answer is set in rooms you may not be in, and the only way to find out is to ask who your CEO called the last three times AI came up. If the answer is not you, the role is still yours. The conversation has already started leaving.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?