Perspective
2008: "It's complete gibberish." Larry Ellison, CEO of Oracle, on cloud computing. Oracle now sells $34B a year of it.
2001: "Linux is a cancer." Steve Ballmer, CEO of Microsoft, on open source software. By 2015, Microsoft joined the Linux Foundation.
1995: "The mouse will not replace my pencil." My first customer, ever.
That one I remember firsthand. My very first software implementation: an online accounting system, built in RPG on an AS/400. Twenty floppy disks, loaded in exact order, no small feat back then. The install was a success. The customer did not think so. They did not want to touch a mouse, but that was just the surface. The real issue was giving up "paper and pencil," something known and trusted, for something that was not. It took two weeks, not of technical talk, but of trust building, until the day they saw the first inventory record logged on a screen. That was the real success.
Then came the internet, and typing a credit card number into a screen for the first time (remember that panic?). Same fear, different decade.
Now it's AI's turn to be the scary mouse.
Same fear, new decade, way more compute.
Putting it all in perspective, one thing is clear: every change starts with fear, moves through doubt, and only later becomes obvious and accepted. But after twenty years of leading enterprise transformations, I have learned something else: every previous wave gave people an escape route. AI might not. And that makes trust, not technology, the thing that determines whether your transformation actually moves.
Which phase are you in now?
Phase 1: Fear
The fear that a tool "does not need me" is real with AI. But it is not new. Bank tellers and ATMs. Elevator operators and the push button. Punchcard programmers and the keyboard. Assistants who printed out calendars (how many of you remember that?). Those tasks went away, sometimes whole jobs with them.
So the difference with AI is not the direction of the fear. It is the visibility of it.
You could always see where the old threats stopped. The teller could see exactly what the ATM did, and just as importantly, what it did not. You could point at the task being automated and step to the side of it. When one skill became obsolete, the next one was within arm's reach.
AI does not have edges you can see. It is not "this machine does task X." It is a widening, hard to define set of cognitive work, and nobody, including the people building it, can tell you where the line is. The old fear was "I cannot use this." The new fear is "I cannot even tell what is left to move into." That is a harder fear to answer, because you cannot dodge a threat when you cannot see its edges.
Here is how the ladder used to work. A teller loses check cashing to the ATM, becomes a relationship manager. An assistant stops printing calendars, becomes the executive's right hand. An engineer stops hand-coding ETL scripts when tools introduce drag-and-drop, moves up to designing pipeline architecture instead. Each time, you moved to the harder thing the tool could not do. And the tool stayed put.
AI does not stay put. You move to the harder task, and a few months later, AI can do that too. The move that always worked, going where the machine cannot, stops working when there may be no "where the machine cannot."
Phase 2: Doubt
Doubt used to resolve on its own, given time. The mouse became just the mouse. The internet settled into something ordinary. The ATM did what it did on day one, and still did exactly that ten years later. Every past wave ended its doubt phase the same way: the tool stopped changing, reality caught up, and the strange thing became boring. Boring is underrated. Boring is how trust gets built. When you stop thinking about something, you start relying on it. That is trust.
AI does not hold still long enough to stop thinking about it. Six months ago it could summarize a document. Now it can run a multi-step analysis across ten documents and draft the recommendation. The moment you form a stable opinion about what it can and cannot do, a new version moves the line.
The doubt is not "can AI do my task?" It is "even if it can, does that actually change anything without changing everything around it?"
This is the phase where the historical pattern breaks down the most. "Give it time and it will feel normal" assumes the target eventually stops moving. So far, it has not. But ready or not, doubt is running out of time. Adoption is accelerating whether the questions have been answered or not.
Phase 3: Acceptance
In every earlier story, acceptance was earned. The internet proved itself. The credit card transaction went through and your money was still there the next morning. You accepted these tools because reality gave you a reason to. Acceptance was the reward for doubt well spent.
With AI, acceptance is arriving before the doubt has done its job. And it is arriving through two doors, neither of them good.
The first door is blind trust. Adopt it everywhere, automate everything, move fast. Do not ask what it got wrong because it sounds so confident it must be right. This is not acceptance. This is skipping the hard part and calling it progress. I have watched teams produce insights as polished dashboards built on incomplete data in 30 minutes, take them as fact, and make costly decisions because of it. Past acceptance meant "I tried it and it works." This version means "I trusted it and never verified." Those are not the same thing, even though they feel identical at the moment.
The second door is refusal. Reject it entirely because the fears are real, the unknowns are too big, the downstream effects, on jobs, on infrastructure, on inequality, on how we organize our lives, are too uncertain to bet on. This is not acceptance either. This is the "pencil and paper" response, and we already know how that story ends.
The honest answer is that there is no world tomorrow without AI. That part is settled. What is not settled is whether we walk through a third door: accepting it without surrendering our critical thinking. Using it while still asking what it got wrong. Adopting it while demanding it earn trust the way every tool before it had to.
That third door is harder. It does not have a shortcut. And it starts the same way it started for me, thirty years ago, in a city on the edge of the Pacific: not with a feature demo or a fear of falling behind, but with someone willing to sit down and build trust, one step at a time.
How a Mouse Taught Me What No Framework Ever Could
It was 1996. Vladivostok, Russia. A remote city on the edge of the Pacific.
My own country, Serbia, was under economic sanctions at the time, but our boss, the owner of an IBM Professional Services shop in Belgrade, had won a contract to build an online accounting system on an AS/400 for a large retail company. They were doing all their inventory and accounting work manually: paper, pencil, ledgers. Our job was to install the software, test it with real data, and train the staff to adopt it.
We spent a month there. (I also learned my first engineering lesson on that trip: always bring backup diskettes. But that is a story for another post.)
The resistance was real. Some people slowly came around, but there was one woman, experienced, sharp, deeply competent at her job, who would literally cry every morning I showed up. She did not want to touch the mouse. She did not want to change anything.
I quickly realized that telling her she had to do it, or escalating to her manager, would get me nowhere. So I stopped talking about the system. I started asking about her life, her family, what mattered to her.
She started crying again. But this time, happy tears. Tears of pride. She told me about her son who had left home for Moscow. How hard it was for him to leave. How uncertain the path looked. But he found his footing, built something for himself, and was now thriving.
That was my moment.
I connected her son's journey to the one she was being asked to take. Leaving behind what is familiar. Feeling uncertain. Trusting the process even when it is uncomfortable. She had already watched someone she loved do exactly this. Slowly but surely, she saw her own change through that lens.
She learned to use the mouse.
So What Does This Have to Do with AI?
The woman in Vladivostok did not need a better demo of the software. She needed someone to understand what the change meant to her personally. The data engineering teams did not need faster pipelines. They needed someone to understand that faster pipelines did not solve the full problem, because the bottleneck just moved elsewhere in the lifecycle, and someone willing to redesign the process so that everyone involved could participate in outcome delivery, not just in execution of their individual task.
Every transformation I have led over twenty years has confirmed this: the technology is never the bottleneck. The bottleneck is whether the people around it trust the change enough to move with it. And trust is not something you can automate, accelerate, or prompt-engineer into existence.
AI can make the technical work faster. It can generate structure, surface patterns, and compress timelines. But it cannot sit across from someone, ask about their son, and hold space while they cry. It cannot read a room of skeptical engineers and know that the real objection has nothing to do with the architecture diagram on the screen. It cannot reorganize a delivery process in a way that makes every person in it feel like they still matter.
Trust is what gets humans to move. It is what got a woman in Vladivostok to pick up a mouse, and what got a room of engineers to rethink how they deliver. AI will not replace that. Not now, not ever.
There is more to making AI stick: rethinking roles, redesigning process, building verification into how work gets done. That is for some other time.
Until then, keep things in perspective. It is the only way to navigate what is ahead without falling into the hype or the fear.