‘Don’t be the dog that barks at every passing car’ Richard Wheldon
How should we separate major underlying shifts in technology from the fads that come and go? How do we identify which trends will be long-lasting and high-impact, and which are over-hyped and will change little?
The characteristics of a general purpose technology are that it has a wide-ranging impact (across multiple industries and use cases, broad economic and social implications), it creates fundamental change in how things are done and not just incremental improvement, the technology itself is continuously improving, and it acts as a catalyst for further innovations (Jovanovic and Rousseau define this as ‘pervasiveness, improvement, and innovation spawning’). AI is a general purpose technology. NFTs did none of these things and so we can reliably define them as a fad.
A technology doesn’t have to be classified as general purpose to instigate transformative change. But these characteristics help us to define which innovations are likely to bring significant, long-lasting impact. The innovations that fundamentally shift the cost/benefit relationship and change consumer behaviour or the way in which a market operates, that get better over time and continuously improve, or those that lead to a cascade of other innovations. These are the real underlying characteristics to pay attention to.
In other words the real impact of transformative technology or innovation comes from reconfiguring systems. Sangeet Paul Chowdray has a wonderful example of this in his new book ‘Reshuffle: Who wins when AI restacks the knowledge economy’. He points out that looking at technological innovation as a story of compounding, exponential growth is only a part of the story. Real change in economic and social systems happens when things start working together in fundamentally new ways, and the humble shipping container is an excellent example of this.
Shipping containers and systemic change
When shipping containers were first introduced it had an immediate effect on the efficiency of shipping and ports but these were only first order effects:
‘The real transformation came later, not through scale, but through scope. Standardized containers and standardized contracts enabled intermodal transport. With that, freight became faster, cheaper, and more reliable. And that reliability broke the logic of vertical integration. Firms could specialize and outsource. And as companies specialized, components improved. Improving components led to greater product innovation through recombination. And the cascade continued. High-knowledge work was separated from high-efficiency production, and trade accelerated. This didn’t happen because of compounding. It happened because of cascading effects – one solved coordination problem unlocking the next layer of opportunity.’
The secret of exponential system change is not compounding acceleration, says Sangeet, but ‘cascading coordination’. Initial breakthroughs may drive cost-savings but the second order effects and cascading innovation remakes entire industries. With each new solved coordination problem, new layers of activity are made possible and the system’s scope expands.
The telephone, and ‘socially interesting’ technologies
Clay Shirky once said that new innovations ‘don’t get socially interesting until they get technologically boring’. New technologies only truly impact society and create profound social change once they become commonplace, reliable, and user-friendly, fading into the background of daily life rather than being seen as novel or complex. The internet was once technologically exciting, but it became ‘socially interesting’ (and transformative) when email, browsing the web, and later social media platforms became so simple and ubiquitous that almost anyone could use them without needing to understand the underlying code or infrastructure. The deeper transformation happens not with first order effects (the direct, immediate function or capability of the technology itself), but second order (the changes that occur because people start using the technology widely and consistently) and third order effects (the often unforeseen societal shifts and behaviours that emerge from the widespread adoption and integration of the technology).
When the telephone was first invented in the 19th Century it was initially marketed by Bell & Watson as a unidirectional broadcast medium rather than a two-way communications device. The idea was that telephone users could listen to concerts and lectures from the comfort of their own homes. Its earliest practical applications were often serving specific, one-way links (a senior police officer to their station, a business owner to their office). It was technologically-interesting, but still not transformative.
It took a number of years for the telephone to evolve into a medium for interpersonal communication and it needed complementary innovations like the telephone exchange and the switchboard to realise its true potential as a multi-user device. Even after the first telephone exchange was established in Connecticut in 1878, the devices were still leased in pairs so that subscribers could set up their own lines to connect one telephone with another. It was only when rural communities began pooling resources to create local exchanges almost 20 years after the telephone was invented, that it became a device that started to appear in people’s homes for personal use, and the second-order effects emerged. And it wasn’t until a decade later, 30 years after the invention, that third-order effects began to appear as it started to become widespread in US cities and reshape social norms and interaction, helped by network effects (the more people that had telephones the more useful the telephone became).
Second and Third order effects of AI
We are, of course, at the ‘technologically interesting’ phase of AI right now, but this is moving far quicker than the telephone, and we are already starting to see some second-order effects. Second-order effects signal that AI is no longer a novelty but a system-level enabler, quietly transforming behaviours, norms, and structures. So what are the signals that we should pay attention to? Here’s my take:
The second-order effects of AI (the integration phase)
- AI shifts from novelty to infrastructure: it fades into the background, embedded in tools, apps, and workflows so seamlessly that users don’t think of it as ‘AI’ anymore.
- Jobs recompose around AI, and don’t just disappear: roles evolve to blend human judgment with AI assistance, an emphasis on skills like critical thinking and creativity alongside AI proficiency.
- New norms of interaction and expectation: people come to expect always-on, intelligent systems across multiple contexts, reshaping what’s seen as responsive, fast, or good enough.
- Human-AI collaboration becomes a core competency: training shifts from tool use to managing complex human-AI workflows involving judgment, prompting, feedback, and ethical reasoning.
- Policy and legal focus moves to practical governance: moving from fundamental questions about AI’s existence to specific issues of governance, accountability, bias mitigation, and data privacy in real-world applications.
Anticipating third-order effects of AI (the systemic phase)
Third-order effects are deep, cross-domain shifts in values, structures, and meaning. They are hard to predict, often visible only in hindsight, but also open to strategic foresight and often stem from the accumulation of smaller shifts. So here’s some practical things that leaders and strategists can do to create that foresight:
- Systemic inquiry: relentlessly focusing on the knock-on impacts of every second-order effect (some call this ‘and then what’ chain thinking) and considering who benefits and who is disadvantaged.
- Watch for cross-sector ripples: looking at where AI in one domain impacts other domains can give you early warning of more systemic shifts.
- Look out for shifts in purpose and values: As AI handles increasing degrees of cognitive labour, cultures may reorient around meaning, creativity, relationships.
- Watch for tipping points and convergences: norms can be reshaped when multiple second-order effects align
- Pay attention to power and equity shifts: AI may accelerate the concentration of power. Strategic foresight requires mapping who is structurally empowered or disenfranchised by these shifts.
AI is already becoming pervasive, rapidly improving, and is spawning a cascade of other innovations. It’s moving so fast (too fast actually) that we really should pay close attention to the emerging second and third order effects, both positive and potentially negative.
A version of this post appeared on my weekly Substack of AI and digital trends, and transformation insights. To join our community of over ten thousand subscribers you can sign up to that here.
To get posts like this delivered straight to your inbox, drop your email into the box below.
Photo by Guillaume Bolduc on Unsplash
Leave a Reply