A shift to Model thinking

Over the past two decades, I have witnessed a seismic shift in how project delivery in businesses work.

From the rigid predictability of waterfall, to the adaptive sprint cycles of agile, to the empathy-driven frameworks of human centred design; each evolution has brought us closer to designing work that mirrors the complexity of the world we’re trying to navigate. These methodologies, while powerful, are still too often focused on outputs. Deliverables. Artefacts. Timelines.

What if the next evolution isn’t about another methodology, but a new way of seeing?

Enter multi model thinking.

Popularised by Scott E. Page, The Model Thinker makes the case that we understand the world better not through a single lens, but through many. Instead of asking what are we building and when will it be done, model thinking invites us to ask how does this system behave, and how might we intervene intelligently within it?

“The most effective decision-makers are not those who rely on a single framework, but those who can hold multiple models in tension.”

In a more unpredictable world increasingly shaped by AI, automation, and feedback-rich environments, model thinking might just be the approach we need to future-proof our delivery practices.

Project-based thinking has long been the default for getting things done in business. It thrives in environments with fixed goals, finite timeframes, and a clear scope of work. It’s excellent at driving short-term delivery, ticking boxes, and keeping teams moving forward. But in today’s interconnected, dynamic environments, that’s no longer enough. Projects, by design, segment problems and aim to produce outcomes within a constrained window of time. They rarely encourage us to interrogate the underlying conditions that gave rise to the problem in the first place or how those conditions might evolve tomorrow. And they almost never model how those factors interact.

This makes project thinking inherently backward-facing. We plan based on historical data, current state analysis, and known constraints.

Model thinking breaks that paradigm.

It asks: What are the forces at play? How do they influence one another? What emergent behaviours might arise if we shift even one part of the system?

Instead of pushing tasks through a pipeline, model thinkers observe systems as living, adaptive structures. They are sensitive to change, feedback, incentives, and time. As Donella Meadows in Thinking in Systems stated “Understanding complexity requires not linear thinking, but loops, feedback, and systems awareness.” Where project plans treat uncertainty as a risk to be controlled, model thinking treats the uncertainty as an invitation to learn.

AI supercharges this shift.

Where once our time was spent gathering, cleaning, and analysing historical data, AI now enables us to automate hindsight. This gives us the bandwidth to focus on foresight, to simulate, test, and explore future possibilities. To stress-test our assumptions against models of customer behaviour, network dynamics, and resource constraints.

AI isn’t just a tool for prediction. It becomes an enabler of model-based delivery, where understanding the system is just as important as delivering the feature.

Pairing agility with model thinking leads to a delivery rhythm that is both adaptive and intelligent. It looks something like this:

• Observe: Use diverse models to understand the system, from market dynamics and behavioural trends to resource flows.

• Simulate: Predict what might happen next. Run scenarios. Understand the ripple effects of a single change.

• Intervene: Design experiments, deploy features, shift incentives. Not just to build, but to influence the model.

• Adapt: Use feedback loops to refine the model and guide the next iteration.

“You don’t rise to the level of your goals. You fall to the level of your systems.” James Clear, Atomic Habits

Model thinking doesn’t live in a whiteboard—it lives in your delivery rhythm. Companies like Airbnb, OpenAI, and Notion don’t just think in systems, they’ve operationalised it by building scaled experimentation engines. These are systems that enable rapid hypothesis testing, measurement, and system learning—across product, marketing, design, and ops.

To make model thinking real, delivery teams need to:

• Run more experiments, not just more sprints.

  You can’t predict complex systems. But you can learn your way through them.

• Automate the rigour.

  Use tooling (e.g. Statsig, Amplitude) to ensure trust in experimental design and analysis. Stratified sampling, pre/post-metrics, and detection of novelty effects make your insights credible.

• Capture institutional memory.

  Avoid running the same test twice. Build experiment archives that allow you to stand on the shoulders of past learnings.

• Make experimentation a habit, not a heroic act.

  Democratise testing, empower product owners and project managers, and remove dependency bottlenecks on data scientists or engineers

A true experimentation engine turns model thinking from a philosophy into a scalable, evidence-based operating system.

As promising as model thinking is, it’s not without criticism. One of the most common concerns is that it can become overly abstract or theoretical, especially in environments where speed and execution are prized. If not anchored to tangible action, teams risk stasis. While model thinking sharpens our lens, it does not replace the need for decisive movement. Insight without action results in inertia. It helps teams ask better questions, identify smarter leverage points, and test strategies with greater clarity.

The goal isn’t to model everything perfectly. It’s to think systemically enough to act strategically, supported by live data and feedback-rich loops.

We’ve moved from managing projects to managing complexity. In that transition, our most valuable asset is not speed or scale, but shared understanding. To build better products, we need to build better models. And to build better models, we need to shift our mindset, from producing deliverables to shaping systems. That means embedding experimentation engines, investing in model literacy, and elevating delivery as a strategic craft.

Model thinking is not a trend. It’s a transformation.

It’s time to stop asking what’s the output and start asking where are the models—and how are we evolving them?