The Laggards Will Inherit AI

Remember the technology adoption curve? Innovators, early adopters, early majority, late majority, laggards. The industry loves to talk about being on the left side of that curve. First to adopt. First to move. First to market.

Nobody wants to be a laggard. It’s the insult VCs throw at portfolio companies that aren’t moving fast enough. It’s the word startup founders use to dismiss established competitors. “They’re too slow. They’ll never catch up.”

Here’s what nobody’s considering: the laggards might be the best-positioned companies to win with AI. And not despite being behind — because of it.

The incumbents’ curse

Look at the companies that “won” the last technology wave. The ones who digitized early, built sophisticated software platforms, moved to the cloud first. They spent the last decade building impressive, complex, deeply integrated systems.

Those systems are now their biggest liability.

Not because they’re bad systems — because they’re expensive systems. In money, in technical debt, in organizational inertia. When you’ve invested $50M in a platform, you don’t rip it out. When you’ve trained 500 engineers on a specific architecture, you don’t change it overnight. When your entire org is optimized around a particular way of building software, “disruption” is a threat, not an opportunity.

I see this in financial services. Banks that spent billions on their core banking platforms are now trying to bolt AI onto systems that were designed in 2005. It’s like strapping a jet engine to a horse. The technology is there, the integration is a nightmare.

The laggard advantage

Now look at the companies that were “behind.” The ones still running on spreadsheets. The ones whose “digital transformation” consisted of getting a website in 2018. The ones whose idea of automation is a shared Google Sheet with some formulas.

These companies have a surprising set of advantages:

No legacy to protect. When you don’t have a $50M platform, you don’t have to worry about migrating it. You can build AI-native from scratch. Green field. No backward compatibility. No “but how does this integrate with our 12-year-old SOA?” Just clean, modern architecture designed for the current state of the art.

Deep domain expertise without the tech baggage. The laggard insurance company that’s been running on paper and spreadsheets for 30 years? They understand insurance incredibly well. Underwriting, risk assessment, claims processing, regulatory compliance across jurisdictions — it’s all in their people’s heads. They have the domain knowledge. They just never had the technology to operationalize it at scale.

Now AI can operationalize it in weeks.

Real pain points, not invented ones. Companies that were slow to digitize tend to have obvious, measurable, painful inefficiencies. Not “our developers spend 15% too much time on boilerplate.” More like “our underwriting process takes 3 weeks and involves 47 manual steps.” These are the kinds of problems where AI delivers transformative, undeniable results. Not incremental improvement — step-change.

Lower expectations, higher ceiling. When your current technology bar is low, the bar for “impressive” is much lower. An AI system that’s 80% accurate at automating a manual process is a revelation, not a disappointment. You get wins fast, build organizational momentum, and iterate from a position of success instead of defending a position of mediocrity.

What this looks like in practice

I saw this firsthand in LATAM insurance. The companies that had “modernized” early had complex legacy systems that made AI integration painful — months of planning, architectural reviews, compatibility testing. The companies that were still running on paper could go from “we have a problem” to “we have an AI solution in production” in weeks.

Not because the technology was different. Because the obstacles were different.

The modern companies had to navigate around their existing investments. The laggards just… built. Directly to the current best practice. No migration path needed. No phased rollout to avoid breaking the old system. Just a straight line from problem to solution.

The four steps to the AI epiphany

Steve Blank’s “Four Steps to the Epiphany” was about customer discovery. But there’s a parallel for AI adoption that I keep seeing:

Step 1: Awareness. The company realizes AI is real, it’s not hype, and it applies to them. This is where many laggards are right now. The ones who get past this step quickly have a leader — often a new hire or an outsider — who’s seen it work elsewhere.

Step 2: Identification. They look at their operations and identify where AI can have the most impact. Because they haven’t automated everything yet, the opportunities are enormous and obvious. This step takes weeks, not the months it takes in complex organizations where everything is “already kind of automated.”

Step 3: Rapid experimentation. They build something. Because they don’t have legacy constraints, they can build it clean. Because their current state is manual, even a rough AI solution is a massive improvement. They get positive feedback immediately.

Step 4: Operationalization. They scale what works. The organizational resistance is lower because everyone can see the improvement. The “we’ve always done it this way” argument doesn’t hold when “this way” is clearly inferior.

For the early adopters of the previous wave, this process is slower at every step. More stakeholders. More systems to consider. More politics. More to lose.

The counter-argument (and why it’s weaker than you think)

“Okay, but these companies don’t have the technical talent.” True. They don’t have 200 engineers on staff. They don’t have a CTO who worked at Google. They don’t have a machine learning team.

They also don’t need one anymore.

The AI tools available now — the same ones any startup can access — are powerful enough that a small team with deep domain knowledge can build production-grade AI systems. You don’t need to train a model from scratch. You don’t need a PhD. You need to understand the problem well enough to describe it precisely, and the tools handle a shocking amount of the implementation.

The technical moat is shrinking. The domain moat is growing. And laggards have domain in spades.

“But they don’t have the data.” Also less true than it used to be. Many of these companies have decades of operational data — it’s just not in a database. It’s in filing cabinets, spreadsheets, PDFs, and people’s heads. Modern AI is surprisingly good at extracting value from unstructured data. What was previously “we can’t use this” is now “we just haven’t processed it yet.”

The great irony

The companies that invested most heavily in the last technology wave built moats that are now becoming traps. Their competitive advantage — sophisticated, custom-built systems — is the very thing that makes AI adoption harder.

Meanwhile, the companies everyone dismissed as behind are free to build directly on the latest technology, unencumbered by the past. They’re not upgrading. They’re leapfrogging.

This is the most underappreciated dynamic in AI right now. Everyone’s watching the tech giants and the AI-native startups. The real story might be the traditional companies in “boring” industries who are about to become shockingly competitive.

Don’t write off the laggards. They might be late to the last dance, but they’re first in line for the next one. And this time, they’re not bringing baggage.