Evolution of Tech Resistance: From Cloud to AI Adoption

Twenty years ago, people were skeptical about cloud computing.

A quick note before you start reading my thoughts: I don’t want to join the ranks of those so-called experts who claim to be AI specialists and predict that it’s going to revolutionize the world. No. I’m not necessarily an AI expert. But I’ve been an IT entrepreneur for over 25 years, and I have a clear view of what’s happening. Above all, I’ve seen this before. What I want to talk about here isn’t the technology itself. It’s the human reaction that accompanies it every time—and this time, it’s likely to affect a large portion of service workers, all those who spend their days in front of a computer.

I’ve been in the tech industry for 27 years, and I started out as a programmer, both in practice and through my computer science studies. What interests me about the moment we’re going through isn’t what the tools can do. It’s seeing the same protective reflexes resurface, in exactly the same order.

The movie I’ve seen before

The last major transformation—before AI—in the way programmers work dates back about 15 or 20 years, when we shifted from desktop software development to cloud-based development, which we later called SaaS. I experienced that firsthand. Many companies had to hire new, open-minded programmers who knew the new languages and JavaScript libraries. Meanwhile, their existing teams—the ones who had built their entire reputation on the company’s desktop software—were resistant. They were protecting their jobs. They were the experts in their field, and the company was completely dependent on them.

It took some companies years to make the transition. Yet executives knew it all along: sooner or later, the vast majority of desktop software would disappear, replaced by the cloud. The change was abrupt for many, and slow for others. I’ve even heard business owners say they set up secret teams to develop software in an innovative language because the rest of the team was too resistant to change. Think about it for a moment: companies that have to hide from their own employees in order to innovate.

At K3—the former name of Parkour3—we experienced something similar with the transition from our in-house CMS, K3 Soft, to WordPress. The developers were resistant to this transition. Most of them left to work elsewhere, and a tiny fraction of them quietly transitioned over to the new team. This transition was necessary. Without it, in my opinion, the company would have gone under.

The important point isn’t that the cloud won. It’s that the resistance didn’t change the outcome. It merely determined who stayed and who left.

What Has Changed, and What Hasn’t

Today, we’re in a similar transition. All LLMs are deploying agents and bots: Claude Code, Codex, Grok’s bots, ChatGPT’s agent mode, agent-based browsers, Claude Cowork, and everything else. All of this makes it possible to automate manual tasks. It’s a little slower, yes. But what the world doesn’t understand is that you can do something else in the meantime. And that you can run several of them at the same time.

Andrej Karpathy, co-founder of OpenAI and former head of AI at Tesla, describes exactly that. He told the No Priors podcast that he hadn’t typed a single line of code since December, shifting from writing about 80% of his code by hand to almost none, with the rest delegated to agents. He runs several agents in parallel on a split screen, assigns a task to each one, and reviews the results as they come in. According to him, programming is becoming unrecognizable.

I’m not mentioning this to impress you with technology. I’m mentioning it because the reaction itself is perfectly recognizable. At first, programmers were on the defensive. That was the case here, too. It took a few years before we were truly able to develop with AI. It’s true that it wasn’t perfect, and it still isn’t. But since we’ve had in-house experts in AI development, others have come to realize that it’s better, faster, safer, and even reassuring for their jobs. Simply because with widespread and effective use of these tools, the company remains competitive and innovative.

There’s a reason for the resistance, and it’s not irrational

We need to be fair to those who resist. They aren’t entirely wrong.

Stanford economist Erik Brynjolfsson has given a name to the dip we’re all experiencing: the J-curve of productivity. When a general-purpose technology is introduced into an organization, measured productivity often drops before it rises. He explains that to reap the benefits, you have to change processes, retrain teams, and sometimes rethink the products themselves. In other words, those who say, “We’re moving slower than before,” are often right in the short term. They’re just wrong about what comes next.

Ethan Mollick, a professor at Wharton, adds the human dimension. He observes that many people give up on AI within the first few hours because it unsettles them and the initial responses are poor, whereas you have to push through to reach the point where you understand what the tool does and doesn’t do. He also warns that organizations that evaluate AI using free or outdated models systematically underestimate its capabilities and set their goals too low. I’ve seen exactly that: people test an old model for twenty minutes and conclude that “it doesn’t work.”

Resistance, therefore, stems from two very real factors: a temporary dip in performance and a disappointing initial experience. That’s precisely what makes it dangerous. It’s justified by arguments that hold water in the short term.

This is no longer just a programmer’s issue

And that’s where it becomes an issue for far more people than just the IT department. The phenomenon is happening across all professions that require the use of a computer—from CRM implementation experts to accountants.

In a few years, people will be joining companies with the ability to automate virtually all manual tasks—nearly 100 percent. These people will instead become experts at setting up bots to do the manual work for them. They’ll be able to get much more work done and will transition into the roles of validators and approvers.

This isn’t a prediction on my part. It’s already been laid out in black and white in specialized publications for other professions, by people in those fields. In the 2026 survey by Accounting Today survey of industry leaders, this shift is described as follows: AI will handle auditing, detection, and reconciliation, which means that accountants will verify the results rather than seek the solutions—a role shift that moves them from preparers to reviewers. An EY executive describes the accountant of 2026 as someone who orchestrates technology rather than executing processes, and who applies judgment to exceptions.

From preparer to reviewer. From executor to orchestrator. It’s exactly the same idea as for developers, just with different vocabulary.

If this is happening among accountants, believe me, it’s going to happen elsewhere

And one of the roles on my list is that of CRM integrators. Not because it’s our field, but because the platforms themselves are moving in exactly that direction. CRMs are quietly becoming agent-based platforms.

Salesforce reports that Agentforce has reached 18,500 customers and over three billion workflows per month, while HubSpot’s Breeze agents are integrated into all hubs, including the free CRM. Gartner predicts that 40% of enterprise applications will incorporate agents by the end of 2026. The most telling sign, in my opinion, is the pricing: Starting in April 2026, HubSpot will charge for certain agents based on results—$0.50 per conversation actually resolved and $1 per qualified lead. When a vendor starts charging based on outcomes rather than seats, it means they consider the agent to be doing the work.

In a few years, you won’t be doing anything manually in your CRM anymore. Everything will be suggested to you, and all you’ll have to do is approve it. Just like an accountant who verifies the answer instead of figuring it out. One analyst summed up this shift as follows: RevOps are becoming system architects rather than data cleaners.

But all of this requires experts to set it up. A poorly configured agent in a CRM isn’t just a flawed report—it’s an email sent to the wrong customer, a lead disqualified for no reason, or a sequence that fires off on its own. The role isn’t disappearing—it’s moving up a level: modeling data, defining the business context, setting safeguards, and deciding what the agent is authorized to do on its own and what needs to be escalated to a human.

And that’s where the real divide between companies lies. Data and business context will become extremely valuable. As a comparative analysis of major platforms in 2026 put it, a CRM’s native AI is never worth more than the data that feeds it. An agent that knows nothing about your customers, your history, your sales cycles, or your won-lost track record cannot offer you any intelligent suggestions. It will respond with generalities.

Companies that lack structured customer data will therefore be at a significant disadvantage, because they simply won’t be able to automate what others will automate. It’s not a question of budget or licensing: both can be recouped in a single quarter. It’s a question of accumulated assets. A clean, well-documented, and properly modeled customer database isn’t something you can buy after the fact. It’s built up over years. And a company that has spent five years letting its team resist the rigor of data has just lost five years it will never make up.

The reluctance will be widespread

This transition is already underway, and there’s nothing we can do to stop it. We see it internally, within our own company, and in our clients’ projects. This isn’t a projection—it’s reality.

And we’re experiencing the same thing we saw on the developers’ side during previous transformations: some people view the arrival of these tools as a threat. But sooner or later, they’ll have to adapt—otherwise, they’ll miss the boat. Other companies—or even other teams within our own organization—will come along with a new, far more innovative way of doing things.

The difference from previous cycles is the scale. The shift to the cloud took a decade and affected only one line of work—one that was already accustomed to change, at that. This shift will take a few years and affect just about every line of work where there’s a keyboard. Entire professions that have never had to redefine themselves will go through this transformation all at once, and nothing in their culture has prepared them for it.

Meanwhile, employees aren’t waiting for their employers. A study commissioned by IBM reveals that 79% of Canadian office workers use AI tools at work, but only one in four uses enterprise-level solutions. Even more telling for executives: 46% of employees would leave their jobs to join a company that uses AI more effectively. In other words, the organization’s reluctance is already costing it employees.

Jensen Huang, CEO of Nvidia, put it as bluntly as possible. According to him, every job will be affected—and immediately: you won’t lose your job because of AI, but because of someone who uses it. You might think that’s an easy thing to say, coming from the man who sells chips. Still, it describes exactly what I’ve seen happen twice in my career.

So no, I’m not here to tell you that AI is going to revolutionize the world. I’m going to tell you something else—and it’s a lot less glamorous: resistance to change will cost far more than the technology itself. It always has. And it’s never stopped anything, either.

The choice isn’t whether or not to go for it. The choice is whether to be the one who learns the new trade, or the one who’ll be told in two years that another team is already doing it.

Sources

Andrej Karpathy, “No Priors” podcast, reported by Forbes, March 2026

Erik Brynjolfsson, J-curve of productivity, McKinsey Talks Talent interview, July 2026

Ethan Mollick, Stanford Graduate School of Business masterclass

Ethan Mollick, keynote at the AI & The Workforce Summit 2026, Valencia

Accounting Today, AI Thought Leaders Survey 2026, predictions about processes

Accounting Today, how technology will shape accounting trends in 2026

Data from Agentforce, Breeze, and Gartner forecasts, 2026 CRM Agent Comparison Guide

A CRM’s native AI is only as good as its data: a comparison of Salesforce, HubSpot, and Dynamics 365

RevOps as a Systems Architect: Analysis of HubSpot’s Shift Toward Agent-Centricity

IBM Canada Study on “Ghost AI” in the Workplace, September 2025

Jensen Huang, Milken Institute Global Conference, May 2025, as reported by CNBC