We’ve Seen This Movie Before

The Chief AI Officer is about to relive the Chief Innovation Officer’s hardest decade, in fast-forward. If you lived the first one, you can already see the second one coming. Someone decided the innovation team, or you specifically, should own or maybe even co-own the AI agenda, and the pilots, the pressure, and the board’s expectations all came with it. Let’s dig a little deeper, because there are good lessons, and remedies, in how the first one played out.

First the lessons. In innovation, the responsibility for the outcome was given to the innovation leader or team, but because the work was based in reducing uncertainty and shaping business-impacting future scenarios, the authority to make the hard decisions did not explicitly come with the responsibility for outcomes. You were on the hook for results, but could not stop the project everyone had fallen in love with, or move the budget, or tell a senior sponsor that their pet pilot was not working. At the same time, leaders who weren’t familiar with innovation, didn’t want to make the changes necessary for implementing the innovations, felt too politically exposed or for any number of reasons also didn’t make the decisions.

There is a name for what was happening, and it reaches well beyond innovation. It is a pattern named Responsibility Erosion: accountability which leads to impact, erodes whenever work moves, scales, or continues, unless someone actively maintains it. The mandate moved to a new role, and the authority to make the consequential calls wasn’t thoroughly addressed. So, we ended up with a mis-match that led to a whole variety of failures including Innovation Theater.

Now watch the same pattern with AI, but at the speed of AI. In a single year, the share of large organizations with a Chief AI Officer went from about 25% to roughly 75%. The role is appearing at speed and with the same instability we remember: most of these leaders do not have a settled reporting line, and the question is already being asked, if this person leaves in eighteen months with nothing shipped, who is accountable? At the same time, organizations are decentralizing decisions and spreading accountability around as AI moves through the business, which sounds healthy right up until you are the one holding the mandate with no decision clearly your own. We have seen this movie. We know how it ends if nothing change.

And it is costing real money right now, while the decisions sit unowned. The studies are consistent and show the impact: MIT found 95% of generative AI pilots delivered no measurable return; BCG found 74% never scaled past the pilot; S&P found 42% of companies abandoned most of their AI projects in a single year, up from 17% the year before. Look underneath the numbers and the pattern is the familiar one: the pilot never got a real continue-or-stop call from a named person, so it ran on until the money or the patience ran out. The innovation team usually paid the price.

Here is what it looks like up close. New York City put out a chatbot to help small business owners, and it began confidently telling them to break the law, advising landlords they could refuse Section 8 vouchers and employers they could pocket workers’ tips. The failure was public and well documented. No one was defending the bot. It stayed live anyway, for more than two years, until a new administration finally took it down, because no single person owned the decision to remove it.

The chatbot stayed up by default, day after day tying up funds and resources. The teams did not lack will or skill, the mandate and the authority to decide came apart, and structurally, nothing was built to hold them together, so the hard decisions and business judgement, never had a chance.

The good news is that a structural problem can be built differently. That is where the remedies come in, getting impactful decisions back into the hands of the people who are accountable for them, with clear authority and supporting structure. We will work through how in the next piece. For now, one question worth taking to your team:

of all the AI work you are now responsible for, which decisions are actually yours to make, and which are you accountable for without being allowed to call?