IT specialists in a server room use AI to analyze infrastructure diagnostics and performance data together.

Why judgment matters more than coding in the AI era

As software creation becomes accessible, leaders must decide what to build and govern to create value.


In brief
  • AI has lowered barriers to building software, shifting the bottleneck to organizational judgment.
  • Enterprises must balance innovation with governance and cost discipline as token use rises.
  • Winning organizations focus on differentiated output and measurable value over activity.

The Tool Builder Era

When everyone in the enterprise can build software, the question is no longer who codes. It is who decides what gets built, who governs it and who captures the value.

In the last few years, the skill barrier to produce working software has fallen. Anyone with an idea, the ability to clearly communicate their intent through a prompt, an agentic coding assistant and a token budget can now build what used to require an engineering team.

Today, inside many enterprises, software is being built by people who do not have a software background, consuming a massive amount of a new currency: tokens. This has created a fork in the road of the AI transformation.

 

On one side is a technology that can propel innovation regardless of skill level and revolutionize how a business operates. Some leaders have embraced the idea of maximizing the consumption of tokens ("tokenmaxxing") to raise the odds of creating the next billion-dollar idea, and they have put leaderboards on their engineering floors to incentivize consumption.

 

On the other side are the companies running out of their annual AI budgets within a month. That spending spree creates problems well beyond finance. Technology organizations now must separate AI slop from real ideas and then scale solutions that were built with no guardrails and no security standards.

 

This fork forces leaders who want to lead in the AI revolution to ask themselves how to balance five things at once: fostering innovation and embracing the new "AI normal," helping employees evolve and upskill, avoiding AI slop, putting the right guardrails in place to manage risk and build for scale, and telling the difference between genuine innovation and undifferentiated AI output.

The leaders who win the next cycle will be the ones who decide, deliberately, what their organization should build, what it should never build, and what the new ratio of builders to governors should be.

This is a practical guide to finding that balance. It offers tactical suggestions for fostering the right culture while designing the blueprint for the future agentic state of the enterprise.

When building gets cheap, judgment gets expensive

The old constraint was talent. You needed engineers and engineers were scarce, so the hard part of any idea was finding someone who could build it. For decades, that was enough of a strategy. Building was the expensive step, so controlling who got to build was how you controlled what got built.

That constraint is gone. When anyone with a clear intent and a token budget can ship working software, the bottleneck moves up a level to the leadership decisions.

Tokenmaxxing optimizes for how much gets built. Budget panic optimizes for how little. Neither one asks whether the right things are getting built, governed and turned into value.

There is one two-part question laying underneath all five variables: how much do you let your organization build, and how deliberately do you decide what it builds?

Part one: build the culture before you build the platform

Embrace the AI normal, but do not reward consumption

Innovation in this era is a default you set. Every time a new task lands, the first question worth asking is how much of it AI could do. The organizations pulling ahead have made that question automatic. That is the “AI normal,” and it is worth embracing.

The trap is in how you measure it. Activity is easy to celebrate, and tokens are easy to count, so tokens end up on a scoreboard. A consumption leaderboard tells your best people that the goal is to use more, and they will use more, whether or not anything comes of it. You end up funding motion.

The fix is to fund outcomes. Keep the enthusiasm and give it a target. Run experimentation with a real budget and a sandbox where people can build without asking permission first, but judge the results on what they change for a customer or a cost line, not on how much compute they burn. The leaders who get this right make consumption boring and make value what we compete for.

Teach people to build and to judge

When building was scarce, companies hired for the ability to build. Now that everyone can build, the scarce skill is knowing what is worth building and when the machine is wrong. A workforce that can only produce output is a workforce that produces a lot of confident, polished, average work.

People need the fluency to build with these tools, and they need the judgment to govern what they build. A builder who cannot tell a good result from a plausible one is a liability at scale. A reviewer who has never built with the tools cannot govern them credibly.

In practice, this means tiering enablement. Everyone should be able to experiment. A smaller group, certified and accountable, should be able to push work into production. A different group should own the gates that work passes through.

Protect what makes you different

When everyone uses the same models trained on the same data, everyone converges on the same answers. AI optimizes for the average, and the average is where competitive advantage disappears.

The good news is that this is a design problem, and you can design for it. The pattern that works is to put the human first in the sequence, introduce deliberate friction and position AI as something to challenge rather than obey.

Part two: design the blueprint

Put guardrails around spend and scale

The most common challenge is cost, because the cost of an agent is often invisible until too late.

Tokens are only part of the cost. True cost includes orchestration, infrastructure, governance, organizational change, failures and regulatory impact. For more information on the total cost of AI, feel free to read Unlocking agentic value: a new investment discipline for the agentic era.

Guardrails enable scale. Organizations must benchmark costs, implement circuit breakers and assign value metrics to every agent from the start.

Decide the ratio of builders to governors

Every organization has a ratio of builders to governors, whether intentional or not. Setting this deliberately requires clear ownership, a defined "never build" list and governance capacity that matches building activity.

The goal is abundant building paired with strong judgment, not unchecked experimentation or bureaucratic slowdown.

A leader's scorecard

Organizations must track innovation, workforce capability, output quality, cost discipline and differentiation.

What winning looks like

The organizations that succeed will deliberately decide what to build, what to avoid and how to balance builders with governors. They will treat building capacity as abundant and human judgment as the scarce differentiator.

Summary 

AI has made software creation widely accessible, shifting the challenge from building tools to deciding what to build and govern. This article outlines how organizations can balance innovation, manage costs and focus on differentiated value rather than activity.

About this article

Authors

Related articles

The data trap: Evolving from data to knowledge as infrastructure

Improve AI readiness with a data strategy that treats knowledge as infrastructure. Learn three steps to close the enterprise AI gap today.

Unlocking agentic value: a new investment discipline for the agentic era

Agentic AI is changing enterprise costs. Learn the total cost of agentic AI, from infrastructure and governance to change, risk and Agent FinOps.

Why responsible AI has become a growth strategy

Responsible AI designed into systems lets organizations scale faster, navigate fragmented rules and turn risk into trust and growth. Read more.