The AI vs SaaS story everyone is telling is too simple

There has been a lot of noise this year about AI killing off the big software platforms.

The story this week about Leopold Aschenbrenner and his Situational Awareness fund got me thinking. He's a very young hedge fund manager, formerly of OpenAI, and had an extraordinary run this year, up over 400 percent through June, largely built on a bet that AI infrastructure would boom while legacy enterprise software would get left behind (that's the simplified version anyway).

He was still up around 80 percent for the year even after a rough July, which nobody seems to be mentioning nearly as much as the losses that have been making the news this week. But the software side of that bet, the part betting against companies like Adobe, has not, yet, worked out the way he predicted.

Here's why I think it won't. One of the disadvantages he has of being so young is that he's probably not yet had the misfortune to experience how bureaucratic and change resistant large organisations are.

Large organisations do not swap out an established software platform easily, and anyone who saw what happened with the Queensland Health payroll disaster some years ago knows exactly why. It started as a $6 million project. The final cost was $1.25 billion, and it still didn't manage to pay people accurately. There are around a hundred similarly painful case studies out there if you go looking.

Even when a system is genuinely painful and expensive, ripping it out and replacing it with something new is a project with a lot of risk attached, and the cost of getting that wrong is way higher than whatever you might save on a subscription fee. There is a lot of research confirming that one of the most painful things a large organisation can do is change their operating platform, and the impact on their profitability in the year after they do that is very rarely positive.

The perceived threat to SaaS businesses has been that with AI, if everybody can build their own software, why would they pay a subscription for it?

However.

Even if Salesforce or Adobe is costing you $1,000 per user a month, that is still peanuts next to the cost of a failed platform migration and the disruption it causes to a workforce that has built years of muscle memory around the existing system. So the big platforms are not getting replaced. They are doing the sensible, obvious thing instead, which is folding AI capability directly into the product they already own the relationship for.

But here is where I think the rest of the story actually is, and it is not the one hedge fund stumble everybody is talking about this week.

About eight months ago, I had a conversation with a partner at a venture capital firm here in Australia who told me they had handed back all the capital they had raised for their latest fund, because they had decided investing in SaaS companies had become too uncertain. At the time I filed that away as an interesting but isolated data point.

Looking at it now alongside what just happened with that hedge fund, I do not think these two things contradict each other at all. I think they are describing the same shift from two different vantage points.

The big established SaaS platforms are going to be fine, because switching away from them is too risky and too expensive for the organisations that depend on them, and because those platforms are already positioned to absorb AI as a feature rather than be replaced by it.

But if you are an early-stage B2B SaaS company trying to break into that same market right now, you are in a genuinely difficult position, because the businesses you are trying to sell into can increasingly get something built specifically for them instead of committing to ongoing payments for a new platform from a startup company without knowing what features they will add or remove in the future, or whether those features will even suit their business. Why would anyone take a risk on an early-stage vendor when the alternative is commissioning exactly what they need?

Which brings me to the part of this that I think matters most. I do not think the opportunity for SaaS companies going forward is in small business at all. I think that market has quietly closed. Subscription fees hit small businesses proportionally much harder than they hit large ones, and smaller organisations are exactly the segment that can now get custom software built quickly and affordably instead of paying an ongoing fee for a generic product that was never quite right for them in the first place.

The safe ground in SaaS is at the top of the market, where switching costs protect the incumbents. The battle is not really AI versus SaaS. It is big versus small, on both sides of that equation. Big platforms are safe. Big customers are staying put.

It is the small end of both, the early-stage vendors and the small business subscribers, where the real disruption is actually happening. One of the emerging consequences of this AI era looks like it's going to be "big beats small." I have some thoughts about that which I will write another day.

If you are a small business owner, this is not a reason to panic, but it is another sign that the businesses who move quickly and build their own advantage will separate from the ones who wait around hoping their existing software vendor eventually gets around to helping them. Nobody is coming to fix this for small business. The businesses that understand that early will be the ones still standing when the gap finishes widening.