Is AI Eating SaaS?

Tomasz Tunguz shared a thought-provoking post analyzing Klarna’s recent annncement on their plans to replace 3rd party vertical SaaS tools with internally developed AI and how that could signal a shift in the B2B SaaS orthodoxy that has driven the last 20 years of B2B SaaS development.

Does this signal a fundamental change in the dynamics – the ‘orthodoxy’ - of the B2B SaaS market? Maybe, but there will be some non-trivial hurdles:

1. It’s a big company model (for now). In the MongoDB example, Tomasz estimates the annual cost offset for sales to be $12-15M for a company with 1,250 people in sales. I would venture most companies with yearly budgets of even a few $M would struggle to keep parity. And that’s just 1 SaaS tool vertical among many.
   2. Focus. The fact that lack of focus = failure is a truism. Instead of the leadership having laser focus on the core company mission, ICP & value props, and competition, they’re now going to have to split their focus between building for external vs. internal customers. There is a reason there aren’t many companies left that look like GE in the late 2000’s.
   3. Resource contention. Let’s say AI can handle most of the dev lift. There will still be people that manage & support it, costs to operate it, etc. Like any company with limited resources (all of them), there will constantly be tradeoffs between investing $1 to make the AI tools better vs. bringing more customers through the funnel. 

For those reasons and more, I’m skeptical that this approach will become the de-facto IT strategy for the broader enterprise market. But, at the risk of contradicting myself…

Let’s say instead, Klarna builds the absolute best AI-driven CRM (would we still even call it that?) ever and decides to sell it to the market before Salesforce et al. can effectively respond. What if this ends up looking more like how Twitter & Slack were born vs. just a massive internal IT exercise?

Companies that can harness the best talent in AI could also demolish the entire vertical construct in B2B SaaS and instead build entire AI-driven solutions around JTBD (jobs to be done) frameworks. 

For example, the JTBD for a manufacturer is to ship 10,000 widgets this year. That would entail engaging AI for:

1.  Product design
   2.  Marketing & lead gen
   3.  Sales
   4.  ERP & Supply Chain
   5.  Shipping & Logistics
   6.  Accounting & collections
   7.  Customer Support

Rather than a company buying 7 different SaaS products from 7 vendors, they buy 1 AI bot designed specifically for manufacturing customers that manages the entire product lifecycle. AI seems to offer a feasible path toward these kinds of JTBD bots, in contrast to ‘traditional’ software dev architecture.

Ultimately, I 100% agree with Tomasz that the orthodoxy of enterprise SaaS will fundamentally change because of AI. How long will it take? Who knows, but it will be fun to watch.