Portrait of Asbjorn Bjaanes
Asbjorn Bjaanes

Articles ·

The Moat Is Moving: What Is Left to Protect When Code Becomes a Commodity

A reflection on what protects a business when AI makes code cheap to copy

The Moat Is Moving: What Is Left to Protect When Code Becomes a Commodity

There is something your organization has built that you are proud of. A product, a platform, software that took years to get right. Ask what makes it yours, and the answer probably comes quickly: the code, the architecture, the algorithms nobody else figured out.

That was my answer for most of my career. Exploring AI-DLC and spec-driven development has made it less solid. AWS has released its frameworks, Google talks about the elephant and the goldfish, and the cost of turning an idea into working software keeps falling. We are commoditizing the coding part of software development.

I don’t have a finished answer to what that means. This is a reflection, written while I am still working it out.

Intent is the new source code

We have been here before. Compilers commoditized assembly, cloud commoditized infrastructure, and each time value moved one level up. Now it is moving from writing code to knowing what is worth building and why.

In The Spec Is Not the Handover I described the spec as something a team builds together in the room, around the intent product brings. Follow that one step further. If AI can produce the code from a good spec, the code becomes the compiled output, and intent becomes the new source code.

It is a small thought, but it quietly rearranges what we point at when someone asks what protects the business.

The copy

Imagine, for a moment, that you are on another team. You are capable, you have the same AI tools, and you are looking at that product someone else is proud of. Within weeks you have something that looks and behaves the same. It feels like a shortcut, and in a way it is.

Then the market shifts, and you have to decide what to change. Which parts were essential? Which were accidents along the way? Nobody on your team knows, because you never made those choices. You hold a result without a history. Lippman and Rumelt called this causal ambiguity: when it is unclear, sometimes even to the owner, how the pieces produce the result, imitation falls short. A copy answers yesterday’s question. It has no way of reaching tomorrow’s.

What the copy is missing

Picture yourself as a product manager visiting a customer. Nothing dramatic is said, but you notice something nobody asked for: every Monday, their operations lead exports the weekly report to a spreadsheet and rebuilds it by hand. In the next planning session you mention it, half unsure it matters. The team asks why, and the real need surfaces. The customer doesn’t want a better report. They want to see which deliveries are at risk before the week starts. That becomes the intent, and with AI the first version ships within days.

Six months later, a competitor copies the feature screen by screen. They get the dashboard. They don’t get the Monday spreadsheet behind it, or the conversation that turned it into intent, so they can’t tell what to change when the customer’s needs move on.

What you carried out of that meeting was not in any ticket or repository. Polanyi called it tacit knowledge: we know more than we can tell. The engineer who just knows which part of the system will break next carries the same kind of knowing. No competitor can download it.

The advantage now lies in turning that kind of knowing into intent AI can build on. Once captured, though, it becomes explicit, and explicit knowledge spreads: through models, through people changing jobs, through competitors studying what gets shipped. What keeps the advantage alive is that people create new tacit knowledge every time they meet a customer. Tacit knowledge is a flow, not a stock.

Think about what could have happened to the Monday spreadsheet instead. If it waits a week, much of it will be gone; Ebbinghaus showed in the 1880s how quickly memory fades unless it is revisited or used. If it passes through three layers and a handover, it arrives without its nuance. And if speaking up feels risky, it may never be shared at all. Amy Edmondson’s research on psychological safety, later echoed by Google’s Project Aristotle, found that people offer doubts and half-formed insights only when it feels safe to do so.

Two organizations

Step back now, far enough to see two organizations in the same domain, with the same tools and equally talented people. Code does not separate them.

In one, the Monday spreadsheet travels through layers before it shapes anything. In the other, the people who sensed it are in the room when intent is set. Over time, one keeps building on a slightly sharper truth than the other. The difference lies in what they keep doing, not in what they own.

In the Meta Model from NLP, a noun like moat would be called a nominalization, a process frozen into a thing. Seen from here, it thaws back into something moving: sensing, learning, shaping intent, deciding. The moat becomes a verb. What an organization produced yesterday can be copied. What it keeps doing is far harder to reach.

From this distance you also notice what has quietly accumulated in the second organization: a history of decisions and the reasons behind them. Pre-mortems, retrospectives and post-mortems, kept close to where the next decision is made, turn those reasons into guardrails. In The Compass You Never Handed Out intent was the direction; the whys mark the edges of the road. As long as someone keeps questioning them, they stay firm rather than hard.

People move through that organization differently too. Someone new plugs in and draws on the accumulated context within weeks. When someone leaves, much of what they knew is already in the system. And as AI takes on more of the building, the people doing the sensing matter more, not less.

What makes it yours

Go back to the thing you were proud of. It may still be valuable. But perhaps what makes it yours is how your people notice, decide and learn, and the reasons behind the choices that brought you here.

Perhaps ownership is shifting from possession to stewardship. A steward never owned the estate they looked after; their job was to care for it, make it thrive and hand it on in better shape than they found it. That may be the relationship knowledge now asks of us: less about stopping others from getting it, more about keeping it alive and growing.

It also puts strategy back at the centre. When anyone can build anything, choices set organizations apart: who they serve, what they say no to, and why. As I wrote in Build Organizations to Behave Strategically, you don’t have a strategy problem, you have an identity problem.

Shifts like this are hard for a reason. Research on predictive processing, from Karl Friston, Lisa Feldman Barrett and others, describes the brain as predicting from internal models and updating them mainly when reality surprises it. Our maps of what protects a business were drawn in another era, and they change only when we notice where they no longer fit. I am still updating mine, with a few questions that keep coming back:

  • If our code were copied tomorrow, what would we still have?
  • How long does it take an insight from a customer to shape what gets built?
  • Which past decisions could the team still explain the why of today?
  • If the best people left, how much of what they know would stay?
  • What are we continually getting better at doing?

Maybe the moat is no longer something we own. Maybe it is something we keep doing.

References

  • Lippman, S. A. and Rumelt, R. P. (1982). Uncertain imitability. Bell Journal of Economics.
  • Polanyi, M. (1966). The Tacit Dimension.
  • Ebbinghaus, H. (1885). Memory: A Contribution to Experimental Psychology.
  • Edmondson, A. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly.
  • Friston, K. (2010). The free-energy principle. Nature Reviews Neuroscience; Barrett, L. F. (2017). How Emotions Are Made.

Written by Asbjorn Bjaanes. Also published on Substack.

Reach out