The Great AI Build

You’ve heard it in meetings, at social events, and perhaps (if you are like me) even in your own internal monologue: “Rome wasn’t built in a day.” The implication is comforting — progress takes time, and meaningful outcomes require patience. Results come later. For the record, it wasn’t a Roman who said this. It was the English playwright John Heywood, who also gifted us “many hands make light work.” Beautifully embedded in our daily vernacular, and yet both ideas are beginning to feel out of step with the AI era. Rome may not have been built in a day, but it was built through relentlessness, and constrained by labour and coordination. And “many hands”? That logic is eroding quickly. AI doesn’t need to see more hands for work to feel lighter.

For decades, ideas, innovation, initiatives, etc inside organisations has been governed by a simple mathematical constraint: to do, or build, something new, something else has to give. A familiar exchange plays out repeatedly. Marketers know this well. They propose a new tool or capability to solve an important problem; a finance lead asks about return on investment (ROI) and what existing initiative it might replace. Don’t be mistaken - this is not dysfunction—it is simply discipline. Resources are finite, time is scarce, and so new ideas are often deprioritised not because they lack merit, but because they arrive into already crowded systems. As Teresa Amabile argues, creativity operates within constraints (some good, some bad). Until now, these have been binding.

 What AI changes is not the existence of trade-offs, but their severity. Indeed, the economics of building have shifted materially. Across industries, AI-assisted development is delivering step-changes in both speed, cost and (perhaps) quality. An ear on the street reveals that this can manifest in the orders of:

• 2–3× increases in developer productivity
• 20–40% reductions in enterprise software costs

 Perhaps these are underselling the reality.

But, in either case, these are not incremental efficiencies—they represent a structural collapse in the internal cost associated with building. And when the cost of building collapses, the logic of prioritisation begins to change with it. And that folks, is no bad thing!

The emergence of (real) parallel building

Instead of asking whether to build one capability it must come at the expense of something else, firms can pursue multiple, often divergent, builds in parallel. The question is no longer “Do we build X or continue with Y?” but rather:

• How many Xs can we build alongside Y?
• How quickly can we test them?
• What do we scale—and what do we discard?

In making this a reality, low-code and AI-enabled tools are not replacing development teams so much as extending them. Most firms now use these tools to expand capacity (particularly in the face of developer shortages). Innovation, in this environment, begins to resemble a portfolio of live experiments rather than a linear “one shot and done” pipeline.

There is always a flipside:

However, as you might expect, this acceleration introduces a new and under-appreciated tension. AI dramatically lowers the barrier to creation, but it does not eliminate the need for integration, oversight, and long-term stewardship. While AI-generated outputs can be produced quickly, they often require downstream correction, validation, and refinement.

 This shift is already visible within large technology firms. AI is enabling substantial efficiency gains—reducing reliance on large teams and, in some cases, saving significantly in labour costs. At the same time, organisations are confronting a new internal challenge: when building becomes easy, deciding what to sustain to a recognised level becomes more difficult – this one, I am afraid, is where we need strategy.

Put slightly differently, as systems proliferate, architectures become more complex, and the long-term burden of maintenance grows. The bottleneck moves away from engineering capacity and towards integration, governance, and ownership.

 The strategic implication is clear. Competitive advantage will not accrue simply to those organisations that build the fastest, nor even to those with the best ideas. It will accrue to those that can build widely while governing rigorously - like really rigorously.

 The risk profile has clearly inverted. Firms are no longer primarily at risk of under-innovating; they are increasingly at risk of over-building. Those risks manifest in very practical ways:

• Fragmented systems and duplicated capabilities
• Rising technical and organisational debt
• Hidden maintenance costs that compound over time
• Loss of coherence across customer and operational systems

AI expands what is possible, but it also amplifies the consequences of weak coordination. Who integrates it? Who maintains it? Who decides what is worth keeping?

Rome, as the saying goes, was not built in a day. In today’s environment, it could likely be prototyped in one.

The more pressing question, however, is what happens on day two!?