How we feel about software right now

It’s a confusing world out there (and “out there” being in our little software/AI bubble, never mind the real world). Software is rapidly becoming a commodity, technical moats evaporate overnight, corporates are vibe-coding their own CRM systems (are they?), new models are being released on a daily basis, multiples compress as everyone expects the end of SaaS and so on, you get the idea.

So where does that leave us? Can entrepreneurs still build software businesses today, and is backing those companies even a sensible investment strategy anymore?

We would emphatically argue yes. In fact, more so today than ever before. But, the shape (and the P&L!) of how a software company looks today is dramatically different to how it looked in the pre-AI era.

We believe that more software than ever will be built going forward.

Why? Well, are you happy with your software? Is your CRM, your email app or your note taking tool feature complete and bug free? We did not think so. As the cost of software supply declines, the demand for software will go up and more software will be built and consumed.

0.3-0.4% of the world’s population work in software development. Contrast that to the 0.6-0.7% of people employed in the global car industry. Now think about the time per day you interact with software vs the time you spend in a car (not that the car has no software).

Arguably this is somewhat of an apples to oranges comparison, but the ratio seems off and factoring in the historical cost of building software we would argue that we have been living in a world where software is supply constrained and there is latent demand for more software.

The cost of software creation has decreased dramatically and will continue to do so. A certain set of customers will use the new and upcoming capabilities to build custom software. At the same time, customers running enterprise software such as Workday or SAP won’t just start vibe-coding like crazy to replace these systems entirely. Instead, customers will start building custom tools and deploying agents, which slowly moves value away from those legacy systems of record. They will still be around, but customers’ willingness to pay for those systems will decrease as those systems’ surface area decreases and workflow value moves elsewhere.

Now, let’s circle back to how one can still build a successful software business today (or invest in it for that matter). The first instinct is obviously to build software that captures the customer value best and hence allows companies to extract most value (i.e. get to high ACVs). The problem with this is that it is so obvious (and remember building software is free), that competition will be intense, driving prices down further. Yes, more software will be deployed and TAMs become larger as software captures more and more of the labour budgets. But we expect customers’ willingness to pay to be somewhat subdued (and we also don’t think customers are willing to transfer labour budgets to software one to one), which leaves most companies in a place where they have more customers but at lower respective ACVs. A more fragmented enterprise software and vendor environment vs the oligopolistic landscape of the past. Agents programmatically buying services from other agents, removing humans and existing relationships from the equation, will further push this dynamic.

So is that bad? We don’t think so. It just means founders have to build differently.

The old SaaS P&L was roughly 25% on COGS, 25% on R&D, 35% on sales and marketing and 15% on G&A. That model assumed high ACVs, long sales cycles and large teams. It does not work anymore. If the average customer is paying you $5-10k a year instead of $50-100k, companies simply cannot afford a full account executive per deal and a large customer success team on top.

Vertical enterprise AI companies (our home turf) that will thrive in this environment are the ones that internalise this from day one. Headcounts need to be radically smaller, 50% smaller or more compared to pre-AI SaaS companies at equivalent revenue stages. Small teams are not a constraint, they are an advantage. They stay agile, they move fast, and they avoid the organisational bloat that kills so many startups before they ever reach scale.

More customers at lower ACVs means more of the customer-facing motion needs to be automated. Sales, onboarding, support. These cannot be mostly human functions anymore when a company is serving hundreds or thousands of accounts with a team of fifteen. The good news is that the same AI capabilities driving down software creation costs and ACVs also make this kind of automation possible.

Less headcount, more automation, lower cost base. The result is a company that can build a genuinely great business even at lower ACVs. High gross margins, fast time to profitability, and real optionality. A company that reaches profitability with twenty people is not a small company. It is a company that can choose what it wants to become, whether that means scaling aggressively into adjacent markets, compounding profitably for years, or both.

This also changes the math on how capital gets deployed. These companies simply do not need $50M in venture capital to find out whether the business works. They need enough to build, get to market and start compounding. If a company raises a few million, grows well and reaches profitability, the founders retain meaningful ownership and control. And for early stage investors with significant ownership, even exits in the hundreds of millions produce fund-defining returns. Naturally, the returns on billion-dollar outcomes are even greater, but the point is simply that right-sized funds no longer need to rely on those extreme outliers to deliver exceptional performance.

We think this is the best time in a very long time to be building software companies. The fear and confusion in the market is understandable, but it is misplaced. The opportunity is enormous. It just looks different than it used to.