AI's Impact on Software: Disruption or Expansion?

On the end of Software vs the 10xing of Software TAM

Sep 23, 2024

This is a weekly newsletter about the business of the technology industry. To receive Tanay’s Newsletter in your inbox, subscribe here for free:

Hi friends,

Today, I’ll be discussing a topic of quite a divergence in viewpoints in the tech community: AI's impact on software. We're seeing two main perspectives emerge – one suggesting AI will disrupt software and we’re seeing the end of software, and another proposing AI will drive massive growth in software by expanding TAM’s significantly. Let's examine both sides.

The Disruption Perspective: Is Traditional Software on Its Way Out?

One view of the world is that as LLMs are getting better at coding and software development is getting democratized, the cost of software trends 0, and instead everyone just spins up personalized user interfaces using LLMs for the tasks at hand. An an interim solution to that is that internal tools built with LLMs custom to the unique aspects about a business can be built and maintained at a fraction of the cost of paying software vendors or that there will be no pricing power as millions of products flood the market.

Chris Paik in his piece “The End of Software” likened into the explosion of UGC in media, where he expects a Cambrian explosion in software that can be created at zero cost, meaning that software as a business in that software businesses getting to 100s of millions in revenue may break down.

One early example of the change to software, which has made a lot of noise but in my view has been a bit overblown is Klarna deciding to rip out Salesforce and Workday for what was implied from media coverage was for in-house, AI-built internal tools. Salesforce and Workday are considered two of the stickiest pieces of software and core systems of record, so if they aren’t safe, then who is right?

My understanding is that at least the Workday switch was to Deel and not to an AI-built internal tool, but I guess the broader view on the point stands.

Chamath has also created an incubator that nods to this trend — a version of your favourite SaaS application that’s 80% as good and costs 10% as much given the declining cost of software creation.

The Growth Perspective: AI Agents as Market Expanders

The optimistic view on software is that AI isn't just altering how software functions—it's potentially expanding the software market dramatically. We're not merely looking at incremental improvements; we're potentially witnessing a revolution in software capabilities and business applications with AI agents.

In many cases, finally, software, with the help of AI agents, can truly compete not just for software budgets but for true spend on labour. This means a whole new set of addressable market up for grabs, which is very conservatively 10-20x the size of the software market. Even if software only automates 20-40% of the work and only captures 10-20% of that, that could still mean a doubling of the aggregate software TAM in many markets.

There are already real examples we’re seeing of this, including from Klarna itself — their AI support agent can now do the job of 700 employees. Sure, they built it themselves, but will every company do that? And what’s a piece of software like that worth relative to what Zendesk was worth when it was a tool used by those 700 employees?

This optimistic view on software is one where software, through AI can truly help businesses not just to as a tool for workflows, but to truly do the work completely, substituting for labor in many cases.

If you don’t yet receive Tanay's newsletter in your email inbox, please join the 9,000+ subscribers who do:

My Take: A New Frontier for Software

First of all, I do think all bets are off if we get to AGI in a way that it can be adopted widely soon. But that aside, my view is that this is probably one of the biggest shifts we’ve seen in software, and while competition will be intense, I don’t foresee a world where enterprises are using millions of custom built software tools and forego software vendors completely. Instead AI represents a promising opportunity for many software vendors, new and old.

Ultimately, AI is still software: It's crucial to remember that AI agents are just a new iteration / kind of software and will be made accessible to most end users through software. They're expanding software capabilities, not replacing software entirely. As Satya Nadella noted on Microsoft’s last earnings call:
“So to me, look, at the end of the day, GenAI is just software”
New opportunities are emerging and TAMs will increase: The “do the work” nature of the new generation of software we’re already starting to see is going to help the software market. One, it means that many new opportunities are opened up that are maybe slightly orthogonal or outside the purview of traditional incumbents, that don’t necessarily involve competing with systems of records on day one. Instead, these products can sit on top of disparate tools and do the work of people, and increase the value that software provides. As I’ve written about in the past, we’re already starting to see AI copilots, AI colleagues or coworkers and AI-native services. While the former may compete with incumbents, the latter two are newer opportunities that allow for companies to go after markets where TAMs were small or unaddressable by typica software.
Internal tools will increase, but not dominate: While more companies will develop their own tools, particularly as no-code becomes even easier, the idea of creating custom solutions for every single need isn't practical, and maintenance costs aren’t quite zero. There are also benefits to learned interfaces and shared interfaces. We’ve actually seen a lot of enterprises take the “build” approach with AI initially, with many companies rushing to build internal ChatGPT’s and various chatbots, but in chatting with CIOs, they’re more than open to buying pre-built solutions that can solve the problems they have as well.
Pricing models will evolve: As I’ve written about before, pricing models for software will evolve. We'll probably see a shift away from per-seat pricing towards models based on usage and work done by the software and over time perhaps even actual outcomes delivered, to better capture the value that the new age of software may deliver.

The Bottom Line

AI is undoubtedly reshaping the software landscape significantly. However, it's not a simple case of AI versus traditional software—rather, it's about AI expanding software capabilities. AI does mean the cost of software development declines and more companies opt for internal tools in some cases, more competition, and a higher bar for software that is sold. But it does also mean a new frontier of what is possible with software, and with that the opportunity to go after previously unaddressable markets with software or markets that are 5-10x bigger.