by Robyn Bolton | Aug 11, 2026 | AI, Metrics
“AI Adoption is a Myth,” declared the headline of a recent viral blog post.
I do love a dogmatic declaration of a controversial opinion and the article opened with a compliment (“you’re already in the top 1% of AI users”), so I kept reading. Eight minutes later, I couldn’t help thinking, “Oh boy. We’re in trouble.”
AI adoption IS a myth
The author is the CEO of a company that implements AI and builds agents for companies earning $500M or more in revenue, which gives him a unique perspective across a wide variety of organizations. In his experience, he estimates that for the average Claude Cowork, or similar, rollout:
- 5-10% of employees become power users (use Cowork daily, create and use skill files for repeated tasks, install connectors to other programs, etc.)
- 20% use it a couple of times a day and either poorly or for “check the box” purposes
- 70% don’t use it at all
Now 10% adoption may be a huge win in the organic baby food market, but it isn’t going to excite a CFO who just spent $2 million on initial deployment and integration and is looking at annual operating expenses of $1-2 million.
Spending on AI is expected to hit $2.6 trillion by the end of 2026, a 47% increase from 2025. The only way that gravy train keeps moving is if CEOs, CFOs, and CIOs believe that their investments will show ROI quickly and increased profit soon after.
Huge swaths of the economy rely on AI spending continuing to rise so it’s in “everyone’s” interest to show data that justifies the spending, namely AI adoption rates, even if the math is a little squirrely.
After all, “frequent use” mean “a few times a week” as it does in Gallup’s report that 28% of employees are “frequent” AI users. Or that adoption should be measured at the enterprise, rather than employee level, as it is in AI Business Weekly’s “AI Adoption Statistics” report that proclaims “88% of organizations now use AI in at least one business function.” (underline added).
AI adoption is NOT a myth
It’s hard to claim that AI Adoption is a myth when research shows that 60% of US adults used GenAI within 3 years of ChatGPT’s launch and by November 2024, 99% had used at least one AI-enabled product in the past week (though 64% didn’t realize they were using AI).
With such widespread adoption, it’s logical to assume that even if AI usage began outside the home, as GenAI becomes widely available with companies, more and more employees will adopt the new tools available on their employer-issued laptops and smartphones.
AI adoption DOES NOT MATTER!
Harvard Business School professor Ted Levitt, famously said “people don’t buy a quarter-inch drill. They buy a quarter-inch hole.”
Yet, with regards to AI, we’re obsessing over the number of quarter-inch holes drilled. Who cares what percentage of your employees use AI? You didn’t install it so everyone could have access to a chatbot. You installed it to get work done faster, cheaper, and/or more accurately. Measure that!
Yes, AI adoption is easy to measure (and easier to “adjust” as needed), and it’s easy to benchmark where your company falls relative to its peers, and it makes everyone feel great when your adoption rate is higher than your competition.
But it’s all vanity unless you’re getting the real results that justify the investment.
Is a high AI adoption rate just innovation theater?
It can be. High adoption numbers are easy to report and easy to feel good about, especially next to a competitor’s lower number. But adoption alone doesn’t tell you if the work got faster, cheaper, or more accurate. Measure what changed in the work AI touched, not how many employees opened the app.
Why do AI adoption stats vary so wildly between reports?
Because the definitions move to fit the story. Gallup counts 28% as “frequent” users, meaning a few times a week. AI Business Weekly reports 88% adoption by counting at the company level: one business function using AI counts the whole organization as an adopter, regardless of how many employees actually use it. Check how a number was measured before you compare it to yours.
If adoption rate is the wrong number, what should I actually measure?
Not how many people opened the app. Measure whether the work AI touched actually got faster, cheaper, or more accurate. As the saying goes, people don’t buy a quarter-inch drill, they buy a quarter-inch hole. You didn’t install AI so employees could have a chatbot. You installed it to get specific work done better. Track that outcome, not usage.
by Robyn Bolton | Aug 4, 2026 | AI, Leadership
“Try AI. It will make you more efficient,” every executive at least once in the past three months.
And it always makes me laugh.
Yes, AI will make save you time and effort. Eventually. But not on the timeline you, or your boss, is comfortable with.
A story that proves the rule
A few days ago, I spent 4 hours building a single AI tool for one research task.
Some context: I use AI Daily and have built prompts, tools and agents to streamline regular tasks. The idea came from a Whop site that I use regularly, which outlined the specific tools, prompts, and steps required to successfully build an AEO (answer engine optimization) scanning tool.
That’s why I only blocked four hours for the build.
To build the tool, I created new accounts on four different websites, worked in Terminal, Claude Cowork, and Claude Code, and referred to instructions from both the originating website and Claude.
If I had a boss or an IT department, neither would be happy with me.
Why? After all, I built a tool for a task that was too complicated and time-consuming to do previously, will run automatically every week, recommends practical fixes, and measures impact against meaningful business metrics.
I’ve also had bosses that, if I told them I spent 10% of my week on this, would say, “that’s nice but where do we stand on (fill in the blank).”
The cost of efficiency
“Lean is the enemy of learning,” according to MIT professor, Ben Armstrong. That’s as true for AI in the office as it is on the manufacturing floor.
Learning is inherently inefficient. Baked into the experience are user mistakes, unexpected outcomes, and dead ends. But those mistakes, outcomes, and dead ends generate the insights required to increase the odds the next attempt will succeed.
As Thomas Edison said, “I have not failed. I’ve just found 10,000 ways that don’t work.”
When we do ultimately succeed, we almost immediately forget time, money, and effort the learning cost us. But forgetting is what makes harder to keep learning. And it results in nonsense like, “try AI. It will make you more efficient.”
The value of waste
Next time you’re tempted to say, “try AI,” stop and ask yourself these five questions:
- “Why am I saying this?” Do you really believe that AI is a solution to the problem you’re hearing or are you just saying it because the company is pushing AI? If it’s the latter, just provide a solution to the problem. If it’s the former, keep reading.
- “What does ‘efficiency’ look like?” As with all experiments, define success first. Do you expect time to completion to be cut in half? Or for it to take just as long, with fewer people? Or maybe it simply costs less. Without a goal, AI is just a toy.
- “How long am I willing to wait for efficiency?” Efficiency doesn’t magically appear the moment a prompt or tool spits out a response. Are you willing to dedicate an hour a week, even if it takes six months to be efficient. Or is success expected in a day, no matter how many hours?
- “Am I willing to pay the cost of learning?” The costs of learning go beyond trainings and tools. It’s the opportunity cost of time spent on another project, calls to customers, conversations with consumers, engagement in meetings. If you’re not willing to pay all those costs, don’t encourage people to “try AI.”
- “How will I reward and spread this learning?” Your “learning waste” can create value for your organization. But only if you reward the people who did the work and spread the learning beyond your team, enabling people to build on what’s known with less “waste.”
“Try AI” is a shortcut to avoid the hard and inefficient work of learning. But paying the cost is the only way to earn the efficiency.
by Robyn Bolton | Jul 29, 2026 | AI
Peak profit is a good news/bad news sort of thing. It’s good news when other metrics, like revenue, customer satisfaction, and employee engagement, are also at their peak. It’s bad news when it signals that you’re in the endgame of disruption. But how can you tell which situation you’re in?
Most students of Creative Destruction or Disruptive innovation will tell you it’s a “bad news” situation if peak profit is preceded by the launch of a radical new technology. They’re partly right.
But it’s not the technology that should scare you. It’s the new business model it unleashes.
What is a business model?
“Business model,” like so many other words popularized during innovation’s heyday, is a buzzword that everyone uses and no one defines in the same way.
A business model is how an organization creates, captures, and delivers value.
That’s it. It’s that simple.
What’s not simple is how to represent all the types of business models, how they work, and what elements make them work. That’s why the Business Model Canvas became so popular. It made visible an interrelated system of decisions that was once assumed, or worse, unknown.
Changing any one element of a business model typically requires changes to other elements, resulting in a new business model.
is also why it’s so hard for existing companies to change their business models or even copy successful new ones.
As a result, companies don’t change their business models unless they’re forced to. Usually by a new technology.
When does disruption actually occur?
When a new technology bursts into the market, it’s an event. The new tech is suddenly everywhere: on the news, in stores, dominating our conversations. It becomes a “where were you when” moment. Where were you when you first logged on to the internet (Miami University’s computer lab, fall 1995)? Used a smartphone (Natick, 2007)? Prompted ChatGPT (Watertown, winter 2022)?
What isn’t an event, but is far more disruptive, is the business model that emerges from the new technology:
While new business models are appearing more quickly after the debut of new technologies, it still takes years for the new business models to “prove” themselves in the marketplace. That’s why it’s so easy to dismiss them even if, like Borders and Blockbuster, you see them emerging.
Why should I worry about it now?
As Hemingway wrote, “How did you go bankrupt? Two ways: Gradually, then suddenly.”
I’m sure executives at Borders and Blockbuster never thought they would be able to speak from experience when quoting that line. I’m also sure there are at least a dozen CEOs right now, reaping the rewards of record revenue and profits, thinking the same thing.
After all, AI and LLMs felt disruptive in 2022 but by 2026, companies have harnessed the technology’s power to cut costs, increase efficiency, and maximize profit.
But in just the last 90 days, people have started to openly discuss the need for new business models:
- SaaS: From seat/feature pricing to outcomes
- Marketing agencies, law firms, and consulting: From charging by the hour to value delivered
- Network operators: From usage-based (volume) to per-connection pricing
How much time do you have?
The disruptive technology is here. The new business models it’s driving are appearing. Peak profits and revenue aren’t far behind. The question you need to ask is how long will it take for “gradually, then suddenly” to hit your business?
Special shoutout to my friends at Jedi On the Fly for their research support via their Jedi Signals intelligence platform
by Robyn Bolton | Jul 21, 2026 | Leadership, Leading Through Uncertainty, Stories & Examples
“Disruption [is] driven by the pursuit of profit. That’s the causal mechanism for these things…”
Clayton Christensen at the 2011 Gartner Symposium ITExpo
When I told a client that peak profit was one of the signs that they were about to be disrupted, his jaw fell open. He didn’t believe me because, like any businessperson, achieving record profit is THE reason to celebrate. His company had just doubled revenue in the past five years and was positioned to double again in the next five. AND they supply mission-critical systems to build out data centers.
Business literally could not be better.
Which is exactly what the CEOs of Big Steel thought in 1968.
It wasn’t too big to fail.
“US Steel posted record profit margins in the years prior to unseating by the minimills; in many ways it was blind to its disruption.”
Clayton Christensen in HBR
Since the mid 1850s, steel was produced in integrated steel mills that performed every function required to produce the material that was building America. The costs to build a mill were high, about $8B in today’s dollars, and, to operate efficiently, mills ran 24/7 to produce at least 2M tons of steel per year.
In 1968, a metallurgist at Nucor invented something called the minimill. It could only perform half of the functions of an integrated mill and produced only rebar, the lowest quality of steel. But the minimill cost only $6M and could be profitable at just 50,000 tons of production.
Christensen called the minimill “not good enough.” He was being nice. The minimill was a joke.
Rebar was a joke, too. At just 4% of the steel market, it had the lowest gross margin of any type of steel. Ceding it to minimills freed up integrated mill capacity to produce more high profit steel. By 1977, Nucor was the leading manufacturer of rebar.
It had also spent 7 years improving the minimill.
The pattern continued:
- 1984: Big steel cedes the angle iron, bars, and rods to Nucor
- 1989: Bethlehem Steel’s market value jumps to $2.4B, from $175M just 3 years earlier
- 1993: Minimills directly compete with integrated mills in all segments of the market.
- 1995: Bethlehem Steel’s primary plant ceases operations
- 2001: Bethlehem Steel files for Chapter 11
- 2003: Minimills production exceeds integrated mills while Bethlehem Steel ceases to exist.
By 2017, only 9 integrated mills were still operating in the US, compared to 111 active minimills. The disruption took 35 years to play out.
You don’t have 35 years
The steel industry isn’t the only example:
| Company |
Time to Disruption |
Peak |
Disruption |
Disruptor |
| Sears |
30 years |
1969: World’s largest retailer |
1999: Removed from Dow Jones Industrial Average |
Walmart, Kmart, Target, Amazon |
| Kodak |
16 years |
1996: Record $16B revenue |
2012: Filed for Chapter 11 bankruptcy protection |
Digital photography |
| Blockbuster |
6 years |
2004: Record revenue: $6B |
2010: Filed for Chapter 11 bankruptcy protection |
Redbox, Netflix |
| Nokia |
7 years |
2007: Record Net Profit $51B, 40% of global handset market |
2014: Handset business sold to Microsoft for $7.2B |
iPhone, Android |
| Intel |
3 years |
2021: Record revenue $79B |
2024: Worst ever stock year as price goes below $18/share |
TSMC, Nvidia |
It’s happening right now. Are you seeing it?
“Financial results measure how healthy the business was, not how healthy the business is. Financial results are a particularly bad tool to manage disruption, because moving up-market feels good financially.”
Clayton Christensen and Michael E. Raynor, The Innovator’s Solution
Executives and shareholders may feel good right now because, despite supply chain disruptions and high interest rates, earnings are buoyed by “margin expansion” and “revenue beats.” AI feels like an opportunity, not a threat. And there’s no reason to believe that tomorrow’s results will be worse than today’s numbers.
It’s exactly how the CEOs of Big Steel felt in 1968.
You still have time to find the joke.
by Robyn Bolton | Jul 14, 2026 | AI
You could practically hear the soaring, triumphant anthem playing over a scene of unwashed yet unbowed humans crawling out of hiding as the machines’ ominous hums slowed and the evil tech overlords realized that their reign was ending.
“Ford’s AI Hiccups Lead Carmaker to Rehire ‘Gray Beard” Engineers’ the Bloomberg headline proclaimed.
It was only one company, but the news was received as if winning this battle foretold winning the war.
Sure, other companies, like IBM, Commonwealth Bank of Australia, and Klarna, rehired humans after AI-motivated layoffs. But the Ford decision just hit different because of the honesty that accompanied the announcement: “Mistakenly we thought that by just introducing artificial intelligence and ingesting the design requirements that we had, that that would produce a high-quality product.”
It’s the honesty, more than the action, that signals a new era emerging.
The fever dream takes hold
Since ChatGPT burst onto the scene in November 2022, companies scrambled to “adopt AI” and cram it down employees’ throats in the most striking demonstration of the Red Queen hypothesis since Lewis Carroll penned the words, “Now here, you see, it takes all the running you can do, to keep in the same place.”
The past three years saw a lot of running (and spending) with not a lot of progress. MIT reported earlier this year that 95% of AI pilots fail and even published an article titled, “What leaders still get wrong about AI” listing the following;
- Treating AI as something you do, not a tool to get results
- Starting AI projects without a clear path to value
- Getting stuck in pilots instead of scaling
- Overlooking how AI changes the business itself
- Mistaking productivity gains for value
Two months after publication, there are signs that leaders are starting to get things right.
Signs we’re waking up
After spending $2.5-$3T between 2022 and 2025, companies’ approach to AI isn’t going to change overnight. But there are indications that leaders are learning and adopting new strategies for AI adoption and implementation.
Executives are admitting mistakes.
After boldly committing to AI and promising step-changes in efficiency, innovation, and earnings, executives are moderating their tone and even admitting their mistakes:
- “Artificial intelligence is a fantastic tool, but it’s only as good as the information you use to train it.” Charles Poon, VP Vehicle Hardware Engineering , Ford
- We “did not adequately consider all relevant business considerations…we should have been more thorough in our assessment of the roles required.” CBA announcing its reversal of AI-related job cuts
- “Really, investing in the quality of human support is the way of the future for us,” Sebastian Siemiatkowski, CEO of Klarna, when rehiring 700 customer service agents
Managers are changing who they hire
After years of layoffs, recent analysis indicates that technical jobs aren’t going away. They’re changing what’s required for success.
In a review of 2.85 million job descriptions posted between June 2025 and June 2026, researchers found a dramatic increase in skills related to “judgment, design, and accountability,” and a decrease in skills related to routine work like “boilerplate coding” and manual testing.
Companies are engaging employees
With 70% of large companies monitoring employee AI activity, it’s no surprise that fatigue and anxiety are increasing, trust is plummeting, and employees are resisting.
But in a switch from the authoritarian, top-down, “because I said so” AI implementation model of the past, companies are starting to engage employees as advocates and trainers. Some are going a step further and shifting their approach from “use AI” to “what tools, including AI, do you need to become the professional you aspire to be.”
Slow then fast
Just like waking up from a dream or crawling out of hiding after the apocalypse, the shift from “AI IS EVERYTHING!” to “AI is a tool” will take time. But the process is beginning.
