On Time, On Budget... but Still Wrong

To start with, a warning: this is not new thinking. The “Iron Triangle” has been under challenge for more than twenty years, particularly since the publication of the Agile Manifesto.

And yet, I continue to see the project management community hold tightly to the idea of “on time, on budget, to quality” as the guiding definition of success in software development.

I struggle to reconcile that position with the reality we now operate in, especially in a world shaped by discriminative and generative AI, where learning cycles are faster, capabilities evolve rapidly, and assumptions decay more quickly than ever before.

So this is not an attempt to introduce a radical new theory. It is an attempt to re-open the discussion in light of how dramatically the environment has changed.

Introduction

For decades, project success has been summarised in three reassuring measures:

On time. On budget. To scope or quality.

This “Iron Triangle” became the shorthand definition of control. If you delivered within those three constraints, you succeeded. If you missed one, you failed.

In modern software development, especially in agile and AI-enabled environments, that definition is increasingly misleading.

You can deliver exactly what you promised, on the date you committed to, within the approved spend, and still fail completely.

Because you delivered the wrong thing.

Where the Triple Constraint Came From

The Iron Triangle emerged from traditional engineering disciplines, construction, infrastructure, manufacturing, where requirements were relatively stable and change was expensive.

In those environments:

• Scope could be defined up front.

• Time and cost could be estimated with reasonable accuracy.

• Discovery was limited once design was complete.

The model made sense.

The problem is that software does not behave like civil engineering.

Even early software thinkers recognised this. In The Mythical Man-Month, Frederick P. Brooks Jr. argued that software projects are fundamentally different from physical construction. His famous observation, “Adding manpower to a late software project makes it later,” illustrates the non-linear nature of software work. Schedule is not a simple variable you can compress at will.

Software is not repetitive execution. It is problem-solving with uncertainty.

Software Is a Discovery Process

Modern software initiatives, particularly data platforms and AI systems, are discovery-driven:

• Legacy complexity emerges during migration.

• Hidden technical debt surfaces under load.

• User behaviour reshapes priorities.

• Regulatory or market changes shift requirements mid-delivery.

• AI capabilities evolve faster than governance cycles.

Scope, in these environments, is not a fixed input. It is an evolving output of learning.

Trying to fix scope, time, and budget simultaneously becomes mathematically unstable. Something must give.

Agile frameworks quietly acknowledged this years ago. In Agile Project Management with Scrum, Ken Schwaber reframed delivery around fixed iterations and stable teams. Time and capacity are constrained. Scope is variable.

The question shifts from:

Can we deliver everything we originally imagined?

To:

What is the most valuable thing we can deliver in this time window?

This is not a minor tweak. It is a structural shift in how success is defined.

Value, Not Compliance

Research into high-performing technology organisations reinforces this shift.

In Accelerate, Nicole Forsgren, Jez Humble, and Gene Kim analysed data from thousands of organisations. They found that elite performers optimise for:

• Deployment frequency

• Lead time for changes

• Change failure rate

• Mean time to recovery

  
  

Notice what is absent: adherence to original project baselines.

Performance correlates with flow and adaptability, not rigid constraint management.

Similarly, the Project Management Institute has evolved its own thinking. In recent standards and thought leadership, PMI emphasises value delivery, benefits realisation, and stakeholder outcomes rather than simple time–cost–scope compliance. The definition of project success is expanding beyond the triangle.

AI Accelerates the Breakdown

Artificial intelligence amplifies the fragility of the old model.

AI systems are probabilistic. Capabilities improve rapidly. Feasibility can change within months, sometimes weeks. What seemed impossible at project initiation may become trivial mid-delivery.

This dynamic aligns more closely with experimentation models such as those described in The Lean Startup by Eric Ries. The Build–Measure–Learn loop assumes that initial assumptions are wrong and must be tested. Learning velocity becomes more important than plan compliance.

In AI-driven environments, locking scope early is often an act of optimism rather than discipline.

Long-range certainty becomes a fiction.

The Real Risk of Clinging to the Triangle

The triple constraint is not wrong. It is simply insufficient.

The real danger is what it incentivises.

Teams optimise for:

• Hitting milestone dates

• Protecting budget baselines

• Defending agreed scope

Instead of optimising for:

• Business impact

• Customer adoption

• Adaptability

• Speed of learning

This is particularly acute in data platform modernisation and AI transformation programmes. Discovery of technical debt often expands scope before it contracts (see The Data Modernisation Pressure Point). New capabilities emerge mid-flight. User needs evolve once they see working prototypes.

If success is defined purely as constraint compliance, teams are discouraged from adapting to reality and instead risk efficiently delivering irrelevance.

What Replaces It?

The emerging model of success in complex technology environments appears to rest on three principles:

1. Capacity is bounded – stable teams and predictable funding.

2. Value is prioritised continuously – scope is reordered based on learning.

3. Adaptability is protected – change is expected, not treated as failure.

Success becomes the sustained ability to deliver meaningful outcomes in the face of uncertainty.

Time and budget still matter. Discipline still matters. Governance still matters.

But they are means, not ends.

A More Honest Definition of Success

In predictable environments, the Iron Triangle remains useful.

In complex adaptive systems like modern software and AI, it is insufficient.

Perhaps the more relevant question is not:

Did we deliver what we planned?

But:

Did we deliver value as reality evolved?

In an era of rapid technological change, competitive advantage belongs not to those who perfectly execute outdated plans, but to those who learn faster than their assumptions decay.

Being on time and on budget is admirable.

Delivering the right thing, even as the world shifts beneath you, is transformational.