All articles

Expect Some Turbulence on the Way Up

Brandon Gadoci

Brandon Gadoci

November 15, 2026

AI transformation does not move in a straight line. It is more like a choppy flight. Shortly after takeoff, the captain comes on and says we're going to climb, but it'll be a little bumpy. Nobody panics at that announcement. It is a friendly heads-up, and it makes the bumps easier to sit through when they come. The plane keeps gaining altitude even when a given stretch feels like a drop, a bit like a yo-yo riding up an escalator.

This post is our version of that announcement.

Four stages, expect some turbulence

We describe progress with AI in four stages: access, usage, incorporation, and leverage. Access means AI is approved and ready to use. Usage means people bring pieces of their work to it. Incorporation means AI becomes part of the process that produces the work. Leverage means the work within reach expands beyond what a person could realistically have produced before. We define each one in A Working Vocabulary for AI Progress.

The four stages of getting value from AI: the line climbs from access through usage, incorporation, and leverage, with ups and downs along the way.
The four stages of getting value from AI: the line climbs from access through usage, incorporation, and leverage, with ups and downs along the way.

The stages read like a steady climb, but people don't move through them once, in order, without moving backward. New AI products appear, roles change, and some kinds of work take to AI faster than others, so one person can be at incorporation on a recurring task and still at usage on the next. The stretch between usage and incorporation alone can last months or even years, which we cover in The Gap Between Using AI and Changing How We Work.

What we see in the field

For anyone taking AI transformation seriously, the parts that feel uncomfortable are part of the growth, and it helps to expect them. People worry about their jobs. Some dislike what AI is doing to social norms and to the way we interact. Others have power struggles over who owns what and what to do next. In the executive suite, we see tension between moving fast and being responsible, and disagreement over which AI products to use, how to measure progress, where to focus, and what approach to take.

Leaders ask whether to empower people first and clean up the mess later, or to go slow and deliberate. They ask how honest to be with their people. Some employees ask what happens if they don't want to change how they work, or if they are simply tired and don't want to learn something new. An honest question runs through all of it: how much of this is really an AI problem, and how much is a lack of leadership, organization, access, or tooling, or the state of the data? It is almost never a matter of deciding to use AI and then enjoying a smooth flight.

We have been here before

Peter Drucker later wrote that he coined the term "knowledge worker" in his 1959 book Landmarks of Tomorrow (The Atlantic, 1994). Since then, knowledge work has absorbed the calculator, the personal computer, the internet, and the Web, and none of them made for a smooth ride.

Pocket calculators spread through the 1970s, yet in 1980 the National Council of Teachers of Mathematics was still urging schools to "take full advantage of the power of calculators and computers at all grade levels." In July 1987, years after the IBM PC arrived, the economist Robert Solow wrote, "You can see the computer age everywhere but in the productivity statistics." The gains did come: US output per hour grew 2.8 percent a year from 1996 through 2000, up from 1.6 percent over the previous 25 years, and many economists credit computers, according to a May 2001 report from the Federal Reserve Bank of St. Louis. The Web followed its own swing. Tim Berners-Lee proposed it in 1989, and a decade later the Nasdaq Composite closed at a record 5,048.62 on March 10, 2000, before falling to 1,114.11 on October 9, 2002, a drop of about 78 percent.

Economists have studied the delay. Paul David found that factory electrification took about four decades to show up in manufacturing productivity, partly because few people yet knew how to redesign a factory around electric motors (American Economic Review, 1990). Erik Brynjolfsson, Daniel Rock, and Chad Syverson argue that AI is likely in the same kind of lag, because its "full effects won't be realized until waves of complementary innovations are developed and implemented" (NBER, 2017). These studies describe whole economies, not one company, and they can't tell a leadership team how long its own dip will last. They do suggest that turbulence is normal.

What this means for AI

If you aren't frustrated with your AI transformation journey, you probably haven't started.

We say that with sympathy, not as a warning. If your rollout feels harder than the demos promised, you are not doing it wrong, and you are not alone. We hear the same worries and the same arguments across the companies we work with, and we identify with them. The frustration usually means something real is starting to change.

The four stages help here, because each one has its own kind of turbulence.

At access, the bumps are quiet. The tool is approved, licenses are assigned, and many people still don't open it, or they open it once and go back to the old way. The next step is small: give people a reason tied to their own work, and time to try it.

At usage, the bumps are uneven. A few people get real value and others shrug, and leaders start to ask whether it is worth it. The next step is to find what the people getting value are actually doing and help others try the same thing on their own work.

At incorporation, the turbulence is usually the roughest, because the work itself starts to change. This is where questions about ownership, measurement, and jobs get loud, and where the tension between moving fast and being responsible shows up most. The next step is to pick one process, decide who owns the change, and agree on what better looks like before the debate starts.

At leverage, the bumps come from success. People can produce work that wasn't realistic before, and that raises new questions about roles, quality, and what the team should take on. The next step is to treat those questions as part of the plan, not a sign that something went wrong.

Across all four, expect dips, and measure the work instead of the logins. Counts of licenses assigned and first sign-ins can climb while the work stays the same, so we ask whether the way the work gets done has changed, and whether the work being produced has changed with it. A bad meeting or a stalled pilot doesn't erase that progress. It is turbulence, not a change in altitude.

In our experience, the companies that make the most progress have leadership teams that want to move forward and to empower people, for the company's future and for the future of the people who work there. Those leaders tend to keep ego out of it, and they are willing to experiment and to forgive. They don't wait for the air to be smooth. They take the next step and keep moving.

The technology will keep changing, and the line will keep bouncing. The right people with the right mindset get you through the ups and downs.

Want to learn more?

Explore our full library of resources or get in touch to discuss how we can help your business.