You Bought the AI. Nobody Changed Their Tuesday.

Much like the introduction of the last trending tool, the licenses went through; there was a kickoff email with a lot of exclamation points and an all-teams meeting. Somebody made a Slack channel.

Now it’s a Tuesday, months later, and we’re all back to doing the same thing we did before, but our emails sound a little flashier, and we were able to have AI help us decide if we should get the single- or double-drawer air fryer thanks to the chatbot tab open in the corner, a little underworked for the cost.

Across organizations everywhere, AI adoption results are underwhelming despite the mass purchase of AI licenses. You’re not alone; you’re in good company. It’s not even divided by department. Gartner’s 2026 CMO Spend Survey found marketing leaders putting 15.3% of their budgets into AI, while only about 30% feel ready to scale it. McKinsey’s latest State of AI report says nearly 9 in 10 companies use AI regularly. Only 37% can point to any impact on earnings.

Is this how we were supposed to use AI?

There are 3 key areas I’ll cover this week on what I have heard from countless conversations, virtual coffees, and a few clients.

Leaders, your team watches what you do

I contracted with a company that invested real money building custom AI tools. The founder endorsed it. That same leader ran the business almost entirely from his phone and an iPad and, aside from curiosity, rarely engaged with any of the corporate tech and tools himself. They were paid for; all of the ‘rollouts’ happened with him in the room, but the final impression was do as I say, not as I do.

The result: The sales team was slow to adopt it. Coincidence? Unlikely.

Behavioral economists call this a descriptive norm. We take our cues from what people actually do, and we watch leaders most closely of all. The memo says AI is important; the actions of the leadership team say it’s optional, guess which one sticks?

A descriptive norm in behavioral economics refers to the perception of what most people in a group actually do, acting as a cognitive shortcut (heuristic) that guides individual choices based on what is popular or common rather than formal rules.

McKinsey’s data backs this up. Companies getting real returns from AI are twice as likely to say their senior leaders demonstrate commitment to it. That kind of commitment looks like using the thing in front of your team, which a signature on the purchase order can’t fake.

You bought an answer tool when you needed a workflow tool

I’ve asked many AI implementation consultants what they see as the biggest headache or challenge in AI adoption with the clients they work with. The answer is more common than not. Companies buy a pile of licenses, hand them out, and nobody knows where AI fits in their actual day. There is no training, no guidance, just a license and some loose guidelines. A shiny tool and no plan for what changes on Monday.

This is why it’s become the replacement for Google. They ask it questions. Not sure how this tool even fits, other than polishing an email, so they go back to the old workflow.

Behavioral science has a name for the space between deciding and doing: the intention-action gap. Buying the license feels like progress, so the decision feels finished. A decision only turns into behavior once the new way of working becomes the default, the path someone falls into without thinking. Most of the time, the old routine remains dominant, not because it’s easy but because it’s already there; it's routine, familiar, reliable. ( also an example of status quo bias)

The companies seeing returns did the unglamorous part, the real work, ironically, to save work later on. Nearly three-quarters of McKinsey’s high performers say they fundamentally redesigned workflows around AI. Among everyone else, it’s about one in four.

AI Accountability pulls the work back

This is the one I find most interesting. I initially read about it in a Harvard Business Review article.

A trust gap.

If you’re accountable for an outcome, you want control over how it gets done. So when the task is handed to AI, yest we are responsible for tbge outcome, there is a gap. Made worse when we see one odd or inaccurate result, we either quietly interrupt AI and take back tasks or never trust the results. Without a clear line between what the person owns and what the machine owns, we assume responsibility, especially when our job is on the line.

Researchers at Wharton call this algorithm aversion. In a 2015 study, people who watched an algorithm make a mistake lost confidence in it faster than in a human who made the same mistake. They went back to the human even when the algorithm had been more accurate.

The hopeful part came in a follow-up study. When people could adjust the algorithm’s output, even slightly, they were far more willing to use it. They performed better and liked the process more. A little control went a long way.

The IKEA effect highlights that we value what we help build. If we contribute to or we are part of the team that shapes how AI handles the work we will trust it more over time, especially if we get options for human-in-the-loop opportunities for verification.

I get it. The first time I let AI organize my computer folders, it felt like driving blindfolded. So I started with one folder instead of the whole drive. With email, I let it sort first, then segment, and only draft much later. Each success earned another step. I got to decide with my comfort level.

What good looks like

What makes adoption stick? The expectation and measurement of AI work ties straight back to the company’s goals and gets measured against them. Another measure is if AI adoption keeps climbing month after month instead of spiking at launch and fading. To manage trust and flexibility, the split between AI and human work stays open, so people can hand over more as trust builds, and existing employees are involved in the AI application(s).

That often surprises people. Ramp and Revelio Labs tracked more than 21,000 US companies and found that heavy AI adopters grew their headcount by about 10% in the two years after adopting, compared with similar firms that hadn’t started yet. Light adopters, the ones dabbling, showed no measurable change.

It’s a correlation, and those companies were already growing faster. Still, the pattern is hard to ignore. Commitment matters more than testing or trying with no plan.

Before you buy another license

When did you last use the tools you’re asking your team to use? How comfortable are you with your own AI use?

Pick one workflow and redesign it with the people who actually do the work, let them set it up, set the guardrails, and decide at levels when trust is earned. The tools are probably fine. What needs work is Tuesday.

So…

Where has AI adoption stalled in your company, and what finally got it moving? And if you’ve ever pulled work back from AI because you were on the hook for the result, what would it take for you to hand it over again?