Clayton Christensen's The Innovator's Dilemma: When New Technologies Cause Great Firms to Fail (1997) set out to answer an uncomfortable question: why do well-run, well-resourced, customer-focused companies routinely lose to rivals offering what looks, at first, like a worse product? Christensen's answer turns on a distinction between two kinds of innovation. A sustaining innovation makes an existing product better along the dimension existing customers already value — faster, cheaper, more reliable — and incumbents are generally very good at these, because their existing customers are asking for exactly that improvement. A disruptive innovation does something different: it starts by serving a segment the incumbent has overlooked or stopped bothering with — often small, unprofitable, or wanting something simpler and cheaper than the mainstream offer — and it improves from there. By the time the disruptive offer is good enough to interest mainstream customers too, the incumbent is structurally late: its whole organisation was built around serving the customers it already had, not the ones it had written off. This is why "worse" products beat better-resourced incumbents, and why disruption is not a synonym for "big change" or "new technology" — a technology that simply makes the existing offer better for existing customers is sustaining, however impressive it looks.

The theory has not gone unchallenged, and a careful user of it should know why. In a widely read 2014 New Yorker essay, "The Disruption Machine," the historian Jill Lepore re-examined several of Christensen's own case studies — including the disk-drive industry at the centre of his original argument — and argued that the evidence for some of them was weaker, and the cases more selectively chosen, than the theory's popularity suggested. A related and more common criticism is simpler: once "disruptive" became a term every founder wanted attached to their pitch, it was applied to almost any new entrant that unsettled an industry, whether or not it matched Christensen's actual definition. Christensen responded directly to this drift, co-authoring a 2015 Harvard Business Review piece, "What Is Disruptive Innovation?", with Michael Raynor and Rory McDonald, arguing that disruption is a process that plays out over time, not a label for any single disruptive-looking product, and that a company often cited as a textbook disruptor does not actually fit the theory's own criteria once checked against it. The practical lesson is not to treat the theory as settled science, but to use it as a sharper, falsifiable question — which overlooked segment, which improving trajectory — rather than a label applied after the fact to anything that won.

Most of the businesses that get called "disruptive" today share a structural feature that has nothing to do with being new or being technological: they are platforms, not pipelines. Parker, Van Alstyne and Choudary's Platform Revolution (2016) draws this distinction directly. A traditional pipeline business creates value in a line: it designs a product, makes it, and sells it to a customer, and value flows in one direction. A platform business model instead creates value primarily by facilitating exchange between two or more distinct groups — buyers and sellers, drivers and riders, hauliers and shippers — rather than by producing the thing being exchanged itself. The platform's job is to make that exchange easy, trustworthy and liquid, and its advantage compounds in a way a pipeline's does not.

That compounding advantage comes from network effects: the platform becomes more valuable as more people use it. Parker, Van Alstyne and Choudary distinguish two directions this can run. A same-side network effect is when more users on one side make the platform more or less valuable to other users on that same side (more sellers can mean more competition among sellers, for instance). A cross-side network effect is when more users on one side make the platform more valuable to users on the other side — more drivers available means a rider waits less, which is usually the stronger and more deliberately designed effect in a two-sided marketplace. Distinguishing the two matters because they call for different strategies: a platform trying to grow often needs to subsidise or specially court one side to trigger a cross-side effect strong enough to get the whole system moving, which is a different problem from managing competition within one side.

This is also why platform thinking belongs alongside disruption theory rather than inside it. A platform does not have to serve an overlooked segment to be disruptive in Christensen's sense — its advantage comes from a different mechanism, the compounding value of the network itself, which can make an incumbent's pipeline economics look slow and thin by comparison even when the incumbent's product quality is perfectly competitive. Recognising a genuine two-sided opportunity, and not just relabelling an ordinary product listing as a "platform," is therefore a distinct judgement from assessing disruptive potential, even though the two often travel together in practice.

Assessing an emerging technology well means separating two questions that get run together: how mature is the underlying research, and how mature is this technology for this specific business use? Technology readiness is a judgement about the second question, not the first — a capability can be scientifically well-established and still be operationally unready for a particular warehouse, fleet or customer base, because readiness depends on the messiness of the real environment (inconsistent packaging, variable lighting, patchy connectivity, exception cases nobody wrote a rule for) rather than on what a demonstration or a research paper shows under controlled conditions. Judging readiness honestly means asking what happens on the exception, not just the typical case.

A useful way to apply this to robotics and automation is to separate tasks by how structured and repetitive they are, rather than by how impressive the underlying technology sounds. Tasks that are physically repetitive, happen inside a controlled environment, and have a narrow range of expected variation — picking a known item off a shelf, moving a pallet along a fixed route, sorting items by a scannable code — are the categories where current warehouse robotics and automation are most reliably mature today. Tasks that require judgement about an unfamiliar or ambiguous situation, negotiation, or responsibility for an unusual outcome remain far less reliably automatable, and typically still need a person in the loop, either doing the task directly or supervising and handling the exceptions the automated system flags. This structured-versus-judgement distinction is a more durable way to assess a specific use case than any claim about a particular product's current capability, which risks dating quickly as the underlying technology continues to improve.

None of this makes a genuine productivity gain disruptive in Christensen's sense. Automating a structured, repetitive task inside an existing warehouse operation is squarely a sustaining innovation: it makes the existing model more efficient for the existing business and its existing customers, rather than changing who the customer is or what they value. That distinction — real value, but not a change to the model — is worth stating plainly, because the two kinds of innovation call for very different kinds of attention from leadership: a sustaining gain is primarily an operational rollout question, where a disruptive one is a strategic question about who you serve.

Customer-service automation — chatbots, automated booking, self-service portals — raises the same structured-versus-judgement question as warehouse robotics, but with a second dimension layered on top: the customer is present for the automation, and experiences it directly, rather than only experiencing its downstream result. A structured, well-defined enquiry with a small number of likely answers — checking a booking status, requesting a standard document, choosing from a known set of service options — is a strong candidate for automation, because the customer's need and the system's capability are well matched, and a fast automated answer usually beats a slow human one. An open-ended, ambiguous, or high-stakes enquiry — a complaint, an unusual exception, anything where the customer needs to feel heard rather than merely answered — is a weak candidate, because the mismatch between what the system can reliably do and what the customer actually needs produces exactly the frustration automation was meant to avoid. The practical design question is rarely "should we automate this journey" as a whole, but which steps in it sit on which side of that line, and how cleanly the system hands off from automated to human handling at the point the enquiry crosses it.

This lesson's sharper point connects back to disruptive innovation theory rather than sitting apart from it. Christensen's overlooked-segment pattern is not only about start-ups disrupting incumbents from outside; an incumbent can apply the same logic to its own under-served customers. A segment a business currently turns away — because serving it by traditional (human, relationship-based) means is not economically viable at that customer's size or margin — is exactly the kind of overlooked segment the theory describes. If automation can serve that segment profitably where the traditional model could not, the business has not just found an efficiency saving; it has found a route to a customer it previously excluded, which is the hallmark of disruptive, not merely sustaining, potential. Recognising that pattern is what separates a customer-service automation project that is "the same service, cheaper" from one that quietly expands who the business can serve.

Every lesson in this module has been building towards one judgement: does an emerging technology justify adding a feature to the existing business, or changing the business model itself? The two are easy to conflate because they can use the same underlying technology — the warehouse robotics, the chatbot and the platform reassessment earlier in this module all involve a form of automation or networked software — but they ask for entirely different commitments and entirely different kinds of organisational attention. A feature-level change improves the existing offer for the existing customer and can usually be delivered, piloted and reversed within an existing team's authority. A business-model-level change redefines who the customer is, what they are being asked to pay for, or how the organisation captures value from the exchange — platform economics, in particular, tends to demand this kind of change, because facilitating exchange between new groups is a different business from producing and selling a single product.

The judgement that matters is not simply "which is more exciting" or even "which is more disruptive" in the abstract, but whether the scale of commitment a business-model change requires is justified by the evidence in hand. A feature-level pilot is comparatively cheap to test and cheap to abandon if it fails. A business-model-level commitment — opening a platform to a new side of the market, restructuring how revenue is earned, taking on the operational and reputational risk of serving a genuinely new customer segment — is not, and reasonably requires a higher bar of evidence and a different level of organisational sign-off before an organisation commits to it. Recognising that a technology could justify a bigger move is valuable strategic insight; recognising that the business has not yet gathered enough evidence, or secured enough organisational commitment, to make that move responsibly is just as valuable, and is often the more honest answer in practice.