Listen closely to how your organization talks about AI and you will hear the problem in its own words. The moment your AI “strategy” lives in a committee instead of inside daily workflows, you are on a well-lit path to expensive, well-documented failure.
AI is not failing because the models are weak or the vendors are lying. AI is failing because companies refuse to treat it as what it actually is: a behavior change problem.
So when I read the recent HBR piece, “How Behavioral Science Can Improve the Return on AI Investments,” my reaction was not surprise. It was relief: finally, research that backs up what I’ve been preaching for far too long.
“Many AI projects fail because leaders treat adoption as a tech purchase instead of a behavioral change problem.”
That single sentence explains why so many AI decks look brilliant while the actual work barely changes.
Most companies still treat AI like a software upgrade. They pick a platform, negotiate pricing, run security reviews, send out an “AI is here” announcement, and then wait for transformation to magically appear in their KPIs.
When nothing meaningful happens, they blame “culture,” talk about “readiness,” and start another round of vendor demos.
The truth is simple and uncomfortable: the work never changed, so the outcomes did not either.
Everyone Labels AI As A Tech Problem
HBR calls out the root disease: technosolutionism. The belief that if you buy a more powerful tool, the organization will somehow become smarter, faster, and more efficient by osmosis.
The reality is brutally human.
People fear losses more than they value equivalent gains. A “better” workflow can feel like a threat. They cling to familiar routines even when those routines are obviously inefficient. They will drop an AI tool after one visible mistake and run back to their slower, less accurate process because it feels safer and more understandable.
None of this is shocking if you sit with teams while they work. It is only shocking if you design AI adoption from a conference room instead of alongside product pods, finance teams, newsrooms, support queues, and RevOps.
HBR’s framework, “Behavioral Human Centered AI,” arrives at a simple truth:
Success with AI has far more to do with how you design, introduce, and manage it around human habits than which model, vendor, or stack you chose.
This is why my own work is built around human led AI workshops instead of a proprietary platform. The technology is already ahead of how most organizations use it. The real constraint is behavior, not horsepower.
AI is not stuck because the tech is not ready.
AI is stuck because behavior change is not being led.
The AI Committee Problem
Let us talk about AI committees, because they are everywhere.
If your leadership team is patting itself on the back for “forming an AI committee,” understand how this lands from the outside. You have created a structure that feels serious while avoiding the hard part.
Strip away the theater and an AI committee is often this:
A group of smart, busy people with partial AI knowledge and almost no visibility into how the organization is already using AI in the wild.
They meet. They review frameworks. They score vendors. They debate governance language. They polish a slide titled “AI Vision.” Somewhere in the deck there is a bullet for “behavioral change,” but no one really owns it, so it gets vague language and dies quietly.
Meanwhile, the real AI story is happening far away from the committee.
A project manager is summarizing status notes with AI.
A customer success manager is drafting renewal emails.
A marketer is brainstorming copy.
An analyst is asking AI to explain a gnarly metric.
A product manager is reworking PRDs with AI as a first pass editor.
People are already bending AI around their daily work. Quietly. Without permission. Without measurement. Without any semblance of leadership guidance.
In most knowledge work environments, +90% of employees already use AI in some way. Yet leadership usually cannot answer three basic questions:
How are people actually using it, at the workflow level?
Where is it genuinely making work better, faster, or clearer?
What impact is it having on time, quality, customer outcomes, and revenue?
Since they do not know, AI committees default to the only things they can see from the conference room: tools, policies, vendors.
Important, yes. Transformative, no.
Every AI committee meeting that ignores real workflows is time spent talking about AI, not time spent improving work with AI.
HBR notes that many organizations involve end users only after the system is built or bought. At that point, “adoption” becomes a training invite and a launch email. People log in once, shrug, and go back to their old habits.
The issue is not “resistance to change” in some abstract cultural sense.
The issue is that no one redesigned the work.
Start With The Work, Not The Whiteboard
HBR argues that behavioral thinking must be embedded across design, adoption, and management. Translate that into plain language and the requirement is obvious:
"You do not start with “What AI platform should we buy?”
You start with “What work is still way harder than it should be?”
If you cannot answer that second question clearly, you are not ready for the first.
Designing for reality means sitting with people while they work. Not asking them what they do in a survey, but watching:
How they write, plan, analyze, respond, and document.
Where they copy and paste the same material.
Where they jump between three tools to get one job done.
Where the thinking is high value and where it is just formatting.
You identify concrete jobs: summarizing meetings, drafting briefs, triaging tickets, preparing reports, planning campaigns, documenting decisions.
Then you design simple workflows where AI takes the first pass and humans keep judgment, context, and final say.
If your AI design process lives mostly in PowerPoint and executive workshops, instead of inside Docs, Sheets, CRMs, ticketing systems, Figma files, slide decks, and chat threads, your adoption problem is already baked in.
Adoption, done properly, is then about trust, effort, and control.
People need to see that AI will support their judgment rather than erase it. They need to see mistakes handled in broad daylight, so they know how to critique and correct. They need to experience clear wins on their own work, not hypothetical use cases crafted by a vendor.
This is why our workshops are built on live material. A marketer brings an actual campaign. An ops leader brings last month’s awful report. A manager brings a real performance review draft. A support lead brings real tickets.
We use AI to rebuild those assets in the room.
People see, in real time, that they saved meaningful time, improved clarity, and kept their own voice. That experience hits very differently than another “future of work” keynote.
Once someone has felt a painful piece of their job get 30 to 40 percent faster and better in front of them, they do not need another adoption memo. They need an excuse not to use it.
Management: This Is Change, Not A Product Launch
The final piece is leadership.
Executives love to talk about AI strategy in the abstract, but too often treat the rollout like a product launch: big announcement, some training, a few dashboards, and then on to the next initiative.
Real leadership here looks different.
You need workflow metrics that track how long key tasks take before and after AI. You need performance metrics that show changes in capacity, cycle time, quality, and revenue-linked outcomes. You need behavior metrics that reveal who is actually using AI day to day, where confidence is rising, and where it is stuck.
If you are not looking at all three, you are not leading AI. You are hoping it works.
And if the story you tell the board or your leadership team is mostly vendor names, architecture diagrams, and compliance updates, do not be surprised when no one trusts that AI is actually moving the business.
My BIG Bet? Human-Led Enablement.
This is why the HBR article felt less like a breakthrough and more like a confirmation.
I am not interested in selling another piece of AI software. The world does not need more logos on the vendor slide. It needs leaders willing to do the unglamorous, human work of changing how teams actually operate.
So we build environments where that can happen.
Teams bring real workflows, documents, interactions, and projects into the room. We use AI to redesign those motions in place. We leave behind simple playbooks and custom GPTs that sit on top of whatever stack the company already owns. And we measure the effects in hours, output, and outcomes, not just “logins.”
The tech matters. The behavior you lead matters more.
The gap between those two is where AI ROI quietly disappears.
If you do not change how people work, think, and collaborate around AI, you will end up exactly where so many companies are headed: contracts signed, a few flashy success stories, and a whole lot of unrealized potential.
The Real AI Opportunity
If you lead a team or a company, here is the uncomfortable truth.
Your biggest AI opportunity is not the next vendor pitch. It is the people already on your payroll and the workflows they run every single day.
Use AI yourself on the work already on your desk. Ask your people how they are actually using it. Pick a handful of critical workflows and redesign them together. Invest in training that lives inside real documents, real customers, real projects. Measure the change.
And if you still think an AI committee is the centerpiece of your strategy, read that HBR line again, slowly:
“Many AI projects fail because leaders treat adoption as a tech purchase instead of a behavioral change problem.”
At this point, the limiting factor is not insight. The limiting factor is leadership.
If this article resonates with you and you’re unsure how to get started, get in touch with PerformanceLabs. tony@performancelabs.ai
Couldn't agree more, this insight is spot on. Given the pervasiv nature of technosolutionism, what strategies do you find most effective in shifting organizational mindsets towards behavioral adoption?