We’ve evaluated such tools running well into triple digits in the past 12 months alone, and believe that the vast majority of those are destined for the graveyard.

We’re in the most exponential cycle of innovation, and therefore value creation, the world has ever seen. Not only has AI allowed for tremendous productivity increases, the rate of change is unprecedented. It is both the most exciting, and amongst the hardest, times to be a venture investor. 

The years 2023 and 2024 saw a mind-boggling rise in AI productivity tools. Vibe coding became real with Lovable, lawyers harnessed Harvey, doctors slashed admin with the likes of Abridge and even the common office worker became far smarter with note taking assistants like Granola. They all deliver as advertised: they search, they summarize, they automate, they save time and capture very useful context in the process. They play across both the first and second phases of the AI development cycle. 

The list of productivity tools, both horizontal and vertical, runs into the many hundreds today. When we are on the precipice of discovering new drugs using in-silico AI modeling, AI Notetaker #25 is not only not needed, it is unlikely to survive as a standalone business. 

Who survives 

Over 50 years ago, Charlie Munger convinced his best friend, Warren Buffett, to ditch the proverbial cheap cigar butts for buying durable, high-quality businesses, centered around their economic moat. Ironically, today, these moats are the weakest they have ever been, specifically in AI-native businesses. 

The pace of innovation that AI has brought about is unprecedented, as is the economic return. Yet the longevity of this economic return is the most unclear it has ever been. Our analysis estimates that around $1TR in net new AI ecosystem revenue was added since the launch of ChatGPT in November 2022 – an unprecedented rate. Meanwhile, the quality of that revenue is amongst the riskiest it’s ever been. AI models are under existential threat from open-source; incumbent chip manufacturers from new entrants; applications from the models themselves, and the weakest of those applications are the plain-jane productivity tools. 

AI applications collectively are today pushing an estimated $150-200BN in ARR, according to Northzone analysis. By far the largest and most mature vertical within this is AI Coding – 20-30% of these revenues – which has amongst the most sophisticated class of AI application products. They too evolved from a basic productivity tool i.e. the Github co-pilot, arguably the first real vertical AI application. From there, it went to a system of action – a Cursor, a Claude Code, a Codex and eventually a Cognition – capable of doing hours’ worth of human work independently. And now full-blown autonomous systems of work (Blitzy, Factory, etc.) that can ingest hundreds of millions of lines of code, understand objectives, and independently ideate, create, and deliver solutions over weeks of autonomous work. In fact, very early signs of recursive superintelligence are already appearing, 

The evolution of the coding vertical is unlikely to be unique. Most, if not all, verticals will follow a similar trajectory. AI doctors and lawyers will deliver autonomous value superior to any single human being. They might come from companies that don’t exist today, or perhaps some of the best aforementioned productivity tools will use their head start, i.e. proprietary data sets and embedded workflow, to evolve into these. 

Northzone’s investments in companies like Tandem Health are already showing this evolution from productivity tool to a true system of action. Others, like XBOW or Blitzy, are true autonomous systems of work, from day one.

So, a few will survive (and thrive) – the rest will perish. 

Where the world is headed

This doesn’t mean we stop funding productivity tools altogether. It does mean that we only focus on those that are creating meaningful new value for the world. 

If the last 24 months of AI were defined by efficiency and productivity increases, the next 12 will be defined by innovation. We’ll likely see a lot more investment behind AI for science – fueling the discovery of new drugs and materials. We’ll see the world become safer for the vast majority of the population (despite the feeling of the converse) through autonomous AI for Defense. Physical AI might be larger than all of Digital AI put together, and will have a lasting impact on human behavior like no other.

By definition, innovation is almost impossible to predict precisely, so perhaps the most meaningful to come is beyond those listed here. At Northzone, we spent almost two years examining what a truly autonomous system of work would look like. And for more than a year, we sat on this (then-) contrarian thesis, not actively deploying capital, even as productivity tools drew vast sums of it. The technology just didn’t exist.

But since the beginning of 2026, we have actively led rounds in excess of several hundreds of millions of dollars, as a convergence of vast foundational intelligence, deep reasoning, and early recursive learning loops saw the arrival of these systems, capable of acting autonomously over long horizons, without being told what to do next. A tool that requires human supervision simply cannot compete with a product that completes months of work in a weekend.

Crudely defined, AGI is the ability of AI to navigate ambiguity, form hypotheses, test them, hit dead ends, iterate to find a solution, execute and deliver value, all without any human intervention. That is the next frontier, and the new standard for investment.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.