Margin of Safety #7: The Next Big Thing in AI – What Comes After Automation?

AI’s biggest opportunity isn’t just automating human tasks but creating entirely new, AI-native experiences that unlock possibilities previously thought impossible.

The Next Big Thing in AI: What Comes After Automation?

A set of strong trends in Gen AI is the direct automation of classic human tasks and the infusion of AI into existing workflows. While these approaches are creating tremendous value, there is potential for even greater value in solutions that use AI to offer previously infeasible capabilities or experiences. AI in existing tasks and workflows disrupts industries, but it’s novel capabilities and experiences that have the potential to unlock full new markets. If this wave follows the same pattern, then there are huge AI-enabled markets that have yet to emerge.

One example of a disruptive experience was Instagram during the rise of mobile. When Instagram was founded, in late 2010, a top of the line mobile device was the iPhone 4. This was the first device to feature a “retina” display in which individual pixels are small enough to avoid being individually visible, but it was still low resolution – the screen was 3.5 inches with a maximum resolution of 960×640. That’s roughly 1/5^th the pixels in a modern iPhone. Similarly, the camera was great for its time but much lower quality than modern devices. Viewed on a large screen, photos from the device look unremarkable. See below for just a couple of examples of the Ferry Building and Union Square in SF[1].

As a result of these constraints, it may have seemed that the mobile camera wasn’t yet ready to displace traditional cameras. But the trick Instagram used was to lean in on an entirely mobile experience. By not only using phone-captured photos but also sharing them on the device (and providing artistic filters that could partially correct for poorer underlying image quality), the small screen and weaker camera effectively canceled each other out. The result was an experience that was highly portable and deeply immersive by the standards of mobile devices.

Had Instagram instead tried to adapt existing experiences – collecting photos with a device, reviewing and editing them on desktop – it would likely have failed; the inferior capabilities of mobile cameras would have been apparent on large desktop displays. But by designing a fully native experience versus adapting classical ones, it was able to turn the disadvantages of the platform into advantages and create a new, persistent experience – as well as helping spawn a huge influencer and direct to consumer advertising economy from which Meta now benefits.

Looking at Gen AI though this lens, the natural question is what new categories will it spawn? To some extent, Andrej Karpathy’s much reposted comments on vibe coding begin to look like this, in that they represent a wholesale shift away from classic strategies for software creation. We can also see hints in AI for drug discovery or medicine, where previously infeasible numbers of candidates may now become possible to evaluate. But we have yet to see the AI-equivalent of true mobile-first experiences across much of the ecosystem. What will these bets be?

At a high level, we expect a defining feature of AI-native experiences to be workflow reinvention instead of workflow automation. Rather than automating existing workflows, AI-native experiences will enable wholesale new capabilities and lead people to spend time on novel activities rather than increasing efficiency of existing ones. Perhaps one such novel activity could be preemptive, low cost re-architecture of IT services to dramatically improve the security risk profile, versus the careful patching and management of existing risk surface. But this is just one possibility of many. We believe some of the most revolutionary ideas will be the most creative and the least expected.