How Software Service Companies Can Find Their Moat In An AI World

One new launch from Anthropic, and billions of dollars get wiped out from the global IT stock index. For decades, software development was a magic to business leaders. Now, AI has pulled back the curtain.

In my experience, every major software industry disruption follows a similar psychological pattern for IT leaders:

Stage 1: Denial—”AI can’t do complex work.”

Stage 2: Bargaining—”We’ll use AI as a tool.”

Stage 3: Panic—Stock crashes, layoffs

Stage 4: Depression—”We’re obsolete!”

Stage 5: This is the evolution we’ll explore here.

By my estimation, most parts of the software service ecosystem are currently somewhere between Stage 3 and Stage 4.

I strongly believe that at this point, technology is not the biggest variable. Human reaction is. Some companies will remain stuck somewhere around Stage 3 and 4, but a few companies will step into Stage 5 and redesign themselves.

The Infinite Monkey Moment Of AI

As the old adage goes, if you put a monkey in front of a typewriter and let it hit keys forever, the monkey would eventually type the complete works of Shakespeare.

For a moment, consider AI as that monkey. By utilizing enough tokens, retries and iterations, it will likely eventually produce something that works. AI is able to do this not because it “understands” the problem like a human but because it can try and fail and try again much more quickly and cost-efficiently than humans doing the same.

For years, or rather decades, effort was the product, hours were the currency, and headcount was the growth strategy. But now that code has become cheap, the value is shifting elsewhere. And that elsewhere is defining the right problem, the right architecture and the right trade-offs.

So, in essence, the monkey is not replacing Shakespeare. It’s just challenging the cost of Shakespeare making mistakes. Shakespeare still holds the crown when it comes to the depth of his thoughts!

Why Effort-Based Software Development Is Dying

In the 1800s, candlemakers had a very strong belief that they were in a wax business. Thus, they competed with factors like wax scent and quality.

Then electricity arrived.

Edison didn’t build a better candle. He made the candle irrelevant. People did not stop needing light; they just stopped needing light by buying candles. In other words, the need for light did not vanish, but the need for candles did. The mechanism changed entirely. That moment made candlemakers realize they weren’t in the wax business—they were in the lighting business. But for most, it was too late.

Similarly, many software companies keep believing they are in coding business. But clients never wanted code. Clients want speed, efficiency, growth, stability and clarity, delivered through engineering excellence.

So, AI is not a better candle. It is a new form of electricity entering the room. That’s why clients are moving from buying efforts to buying outcomes.

This doesn’t mean legacy systems will vanish overnight, however. Enterprise transformation is measured in years, not quarters. Regulated industries move cautiously. Complex architectures require careful modernization.

What AI Can’t Commoditize (Yet)

Throughout my pre-AI and post-AI industry exposure, I’ve noticed one persistent truth: Businesses don’t suffer from a lack of answers. They suffer from poorly framed questions, unclear priorities and systems that don’t work together. So, in an AI-driven software world where answers are free, a company’s ability to ask the right questions has become priceless.

The next thing AI can’t commoditize is the cathedral builders. Enterprise software is about more than just features; it’s about reliability, security, compliance and long-term stability. A trustworthy system cannot be generated in a prompt. Trust still needs an architect. A good software engineering team—one made of people who knows it’s never been about using the tool or writing the code, but rather about seeing the right destination for the business and guiding the system toward it—will be difficult for AI to replace.

There is also a key problem with AI that is still unanswered: AI moves one bottleneck to another. For example, writing code manually was a bottleneck until 2020. Then AI made code generation almost instantaneous, and the bottleneck moved to code review and testing. Now, with AI-driven review agents, the bottleneck is shifting again—toward Ops and toward system reliability, stability, maintainability and security.

This makes me believe strongly that if software engineering were a 3D space, AI is stretching the X and Y axes. But the Z-axis—the depth of planning and execution—still depends on software engineers.

Three Species Of Survivors In The New World Of Software Development

After the Industrial Revolution, the ecosystem didn’t shrink. It was reorganized. Similarly, I believe that software services won’t disappear, but rather evolve into different species.

Conclusion: Engineering Depth Over Coding Speed

Coding was never the product; engineering is. Coding stretches the surface; engineering goes deeper, where actual business problems lie.

It’s time to collect all your wisdom gathered in the non-AI era, combine it with the velocity possible with AI, and adapt deliberately. I believe those who pull out ahead will be the IT leaders who recognize a simple truth: AI will not kill software services. It will punish companies that continue selling effort instead of outcomes.

As Shakespeare once wrote, “Our doubts are traitors, and make us lose the good we oft might win, by fearing to attempt.” So, let’s attempt something bold: Replacing coding with engineering. Let’s replace <hello world> with <solve the world>!