As I walk into the valley of death

Are you playing offence or defence? And did you choose?

I was leading a re-platforming project at a food delivery company, sitting in a one-on-one with the CEO. It was the kind of project that eats budget, people and patience in roughly equal measure.

Then he called my project the valley of death. Wait, what?

He picked up a marker and drew a curve on the whiteboard that dropped before it climbed. While we replaced the platform, the business couldn’t add features. Every sprint and every dollar went into the new foundation, while the market kept moving and competitors kept shipping.

On the other side, if we made it, everything changed. Features that took months would take days. The business could stop reacting and start attacking.

I’d been treating it as a technology project. He was describing how the company had chosen to compete. And if my project failed, the company was dead.

That conversation stayed with me, because it reframed my own work. The project on my plate was the company’s competitive bet, whether or not anyone had said so out loud.

Every organisation is in some version of that valley right now, and everyone inside it, from the engineer on the platform team to the CTO, is making calls that depend on three questions. Is the organisation playing offence or defence? Did it choose deliberately? Is it still the right call for where the market is heading?

Most people can’t answer all three, whatever their title, and in an AI-accelerated market that gap is getting expensive.

To see why, it helps to look at seven operating models that answer the question differently. Four are household names: Amazon, Microsoft, Toys “R” Us and Apple. The other three, McDonald’s, the lifestyle business and the food court, are closer to what most people reading this actually run.

Seven operating models

Amazon: Working Backwards, Not Forward

Amazon built its operating model around the customer. Working backwards from what the customer needs is, on paper, baked into its meetings, product decisions and every new market it enters.

It works, until it doesn’t.

Working backwards means you don’t attack markets where the customer isn’t asking. Microsoft Office has dominated enterprise productivity for thirty years. It is expensive, bloated and widely disliked. Amazon has the engineering talent, the cloud infrastructure and the distribution to build a serious competitor, and it hasn’t. Office customers aren’t complaining to Amazon. They’re just paying the bill.

Customer obsession is Amazon’s great strength, and also its blind spot. If the customer isn’t asking, the model doesn’t move.

Microsoft: The Eye of Sauron

Microsoft obsesses over competitors rather than customers. Netscape, Java, Apple, Google: each time a threat emerged, Microsoft turned its full attention to it and used its existing dominance to attack. Usually late, always overwhelming.

I think of it as the Eye of Sauron. The focus lands, the threat is dealt with, and the eye moves on.

The problem is the quiet between threats. Office sat largely untouched for years, because nothing threatened it enough to earn the eye’s attention.

Now AI does, and Copilot is the response. When I first published this article in June 2026, I wrote that a well-funded startup with a modern language model could build something better than Office in under a year. That turned out to be conservative.

Four months later, ArtCraft, a small team led by Brandon Thomas, published WordCraft, GridCraft and DeckCraft: free, open-source, clean-room rebuilds of Word, Excel and PowerPoint, written in Rust with Claude listed as a co-author on the commits. The repositories were created on 7 October and had installable releases within the week. WordCraft passed 15,000 GitHub stars in under a fortnight.

They aren’t a replacement yet. By the team’s own roadmaps they’re pre-alpha, and WordCraft is roughly 60% of the way to genuine parity. But they’re free, they open real Office files, and every app ships with an MCP server so AI agents can drive it. Microsoft’s own model created the conditions for its disruption, and the first serious attempt took weeks.

Toys “R” Us: Acquire and Extract

Toys “R” Us didn’t lose to Amazon. It lost to its own balance sheet.

In 2005 a private equity consortium bought the chain in a leveraged buyout and put most of the debt on the company’s own books. From then on, hundreds of millions of dollars a year that could have funded better stores, a real e-commerce platform or sharper pricing went to interest instead. The buyers also sold off the company’s property and leased it back, booking a quick profit and leaving the business with rent it had never paid before.

That is the extraction model. Buy a dominant brand, load it with debt it didn’t choose, take value out through fees and leasebacks, and defer reinvestment until competitors close the gap for good. It isn’t incompetence. The model worked exactly as designed, just not for the business carrying it.

For the buyers, it worked well. The fees were collected and the exit could be timed before the consequences landed. For Toys “R” Us, the money it needed to compete was spoken for the day the deal closed.

Extraction isn’t a strategy. It’s a countdown.

Apple: The Platform is the Moat

Apple’s model is the platform. Hardware, software, services and the App Store are tied together so tightly that leaving is painful, and the manufacturing advantage behind it is extraordinary.

But a moat protects what already exists. It doesn’t produce the next iPhone. Vision Pro shows the pattern: superb execution, an unclear customer problem, and sales that fell well short. The model is built to defend the platform, not to generate the disruptive offence that created it in the first place.

Apple after Jobs is a different company running the same plays on accumulated momentum. So far, that has been enough.

McDonald’s: The Franchise

McDonald’s doesn’t really sell burgers. It sells a system specified so precisely that judgement becomes unnecessary. Fryer timing, assembly sequence, real estate, supplier contracts: all of it is designed to produce the same result whether the operator is a genius or a journeyman.

That is the point. A franchise optimises for replication, not innovation, and it can scale to tens of thousands of locations because it has engineered out the need for local judgement. The system is the product.

The blind spot is the mirror image of the strength. A model built to remove judgement can’t easily ask for it back when the market shifts. McDonald’s has spent years bolting on delivery, digital ordering and menu localisation. Each was a genuine innovation, and each fought an operating model designed for sameness.

There is a second failure mode, and it is closer to home. My local bubble tea shop is part of a franchise chain. The moment it slips (the wrong ratio, slow service, a flat batch of tea) it is a fast path to one-star reviews. The customer was never loyal to that shop. They were loyal to the standard, and once the standard breaks, the system works against you.

The Lifestyle Business

Not every business is trying to get bigger. The lifestyle business is deliberately built around a ceiling: enough revenue to fund the life the owner wants, and no appetite for the debt, hiring or risk that scale demands.

That can be an excellent business. A consultancy with three partners and a full book of work, a regional manufacturer happily supplying its existing customers, a single well-run restaurant. The model optimises for control, not expansion.

The blind spot appears when the owner’s ambition and the business’s stated direction quietly drift apart. Owners keep using the language of growth and expansion while every actual decision protects the ceiling. The operating model is honest even when the leader isn’t. Recognising which business you’re actually running, rather than the one described in the board pack, is the diagnostic moment.

The Food Court

Another Thai restaurant opens two doors down from the one that’s already there. Same menu, same prices, same hours. Neither has a reason to win beyond being slightly more convenient on a given night.

That’s the food court model: same product, same market, no differentiation, competing on proximity, price or whoever shouts loudest this week. It is rarely chosen. It is what happens when a business enters a crowded market without deciding what it does differently, and never goes back to ask.

The blind spot is durability. Leaders in a food court business often believe the sameness is temporary, something they’ll fix once things settle down. The operating model locks it in. Without a deliberate point of difference in pricing, product or distribution, every decision defaults to matching the competitor next door. There is no version of this model that earns a premium. It only ever earns a margin.

The blind spot built into every model

Every one of these models creates a blind spot, and it comes from the design rather than from bad luck. What you optimise for decides what you can see.

  • Amazon optimises for the customer, so it misses markets where customers aren’t asking yet.
  • Microsoft optimises for the competitor, so it misses slow erosion in markets nothing is threatening.
  • The owners of Toys “R” Us optimised for their own return, so nobody was looking at what that did to the business.
  • Apple optimises for the platform, so it misses problems that don’t fit inside it.
  • McDonald’s optimises for sameness, so it struggles to ask for judgement when the market shifts.
  • The lifestyle business optimises for the owner’s ceiling, so it misses the gap between what the owner says and what they do.
  • The food court optimises for nothing in particular, so it never escapes the comparison with the shop next door.

The model that makes you successful is the same model that limits you. You can’t optimise for everything at once. Focus is a choice, and every choice has a cost.

Having a blind spot is normal. Every leader and every organisation has one. The danger is not knowing where yours is, because your competitors probably do.

Are You Playing Offence or Defence?

Most organisations default to defence. They protect existing revenue, existing customers and existing market position. They respond to threats rather than create them, optimise for not losing rather than for winning, and call it strategy.

Defence feels responsible, and in a stable market it can be enough. Markets aren’t stable right now. In an unstable market, defence is a slow loss: you watch your position erode one quarter at a time and call it discipline.

Offence looks different. It means finding the unloved dominant player in your market where the incumbent customers keep paying because switching feels too hard, and asking whether now is the moment to go after it. It means crossing the valley of death on purpose, because you’ve decided the other side is worth the pain.

Mapped this way, the seven models tell an interesting story.

ModelOffence or defence
AmazonDefence, dressed as customer focus
MicrosoftOffence, but only when threatened
Toys “R” UsNeither. The owners harvested value and left before the consequences landed
AppleDefence, wrapped in a premium blanket
McDonald’sDefence, dressed as consistency
Lifestyle businessOpted out of the game, by design
Food courtNeither. No position to defend and no target to attack

None of them is playing proactive offence: looking at an adjacent market and attacking it because the moment is right, rather than because a threat forced their hand.

That gap is sitting open in your market too. The question is whether you see it, and whether you have the model, and the nerve, to go after it.

AI changes the maths

You could always be copied. What’s different now is the speed.

AI has collapsed the cost of building software. A small team with a modern language model and no legacy infrastructure can build in months what used to take years. The things that protected incumbent platforms (switching costs, integration complexity, the sheer effort of matching features) are eroding faster than most leaders realise.

Every market has a Microsoft Office in it. A dominant, expensive, widely disliked product that customers keep using because replacing it felt impossible. For twenty years that feeling was largely correct, because a credible alternative needed Microsoft-sized resources. That calculation has changed, and Office itself is now the test case.

The valley of death is shorter now, and far more organisations have the capability to make the crossing. A year ago your position was protected by complexity. Today it’s protected by inertia, and inertia breaks suddenly.

The organisations that understand this are already moving. They’re mapping their market for unloved dominants, deciding which valley is worth crossing, and building offensive capability before the threat arrives.

The ones that aren’t are sitting in their own valley of death. They just haven’t drawn the chart yet.

Three questions a leader should be able to answer

These aren’t only questions for the executive table. Wherever you sit, your answers shape which work you push for and which risks you name. Most people never ask them, because the model they work inside has never required them to.

1. Which model is your organisation running?

Are you working backwards from the customer like Amazon, or watching competitors for threats like Microsoft? Extracting value like the owners of Toys “R” Us, or defending a platform like Apple? Or is it closer to the franchise, the lifestyle business or the food court?

Most organisations run a hybrid, and usually an unintentional one: some customer focus, some competitive reaction, layered over a cost structure that looks a lot like extraction. Name it honestly before you try to change it.

2. Did anyone choose it?

This is the harder question. Most operating models are inherited. They came from the founder’s instincts, the first big customer or the competitive pressure of an earlier era, and they set into process and culture before anyone asked whether they still fit.

An inherited model isn’t automatically wrong. Running one without questioning it is how organisations sleepwalk into irrelevance.

3. Is it still right for where your market is going?

Look at where your market is heading, rather than where it has been.

If AI is cutting the cost of your competitors’ build, your defensive position is weaker than your balance sheet suggests. If the unloved dominant in your market is suddenly vulnerable, that window won’t stay open for long. And if you’re in the valley right now, the real question is whether the other side justifies the pain.

These are leadership questions, and you don’t need a title to ask them. What changes with altitude is what asking looks like.

For individual contributors, knowing what your work means to the operating model, and saying so when a technical decision quietly changes the bet.

As a leader or manager, choosing what your team works on deliberately, choosing to play offence or defence, not just whatever is next in the backlog.

For the C-suite the risk is much higher. You are putting the operating model on the table where strategy is set, rather than inheriting it into the roadmap.

The Chart on the Whiteboard

I never forgot that conversation. The CEO knew exactly which game he was playing, what the crossing would cost and what was waiting on the other side. What changed for me was seeing that the person leading the project needed to know that too.

That clarity is rare at any level. Most organisations are somewhere in a valley right now, and the people who see it first are often nowhere near the boardroom. The engineer who notices the platform can’t keep up. The manager whose roadmap is all defence. The CTO who inherited a model nobody chose.

Whatever your title, you can draw the chart. The question is whether you pick up the marker before someone else draws it for you.

So, is your organisation playing offence or defence? Did anyone choose that? And what will you do about it from where you sit?


Everything in this article is my personal opinion based on publicly cited references, listed below. This is one practitioner’s view, offered in the spirit of understanding what makes something great, and what it takes to keep it that way.

References

Amazon: Working Backwards

Microsoft: Embrace, Extend, Extinguish

Toys “R” Us: Acquire and Extract

Apple: The Platform is the Moat

McDonald’s: The Franchise

The Lifestyle Business

Food Court