We’re selling outcomes now
Almost all software businesses sell tools. Something that a person or team can use to get their work done and achieve an outcome. However, this is changing. Thanks to the proliferation of extremely capable Large Language Models, for the first time ever, it’s possible for software businesses to sell outcomes, taking human labour mostly or entirely out of the loop. This will be a big shift that completely changes what we all build, how we build it, and how we sell it. This evolution will disrupt every SaaS company that does not eventually adapt.
Long ago, the software industry adopted the service (i.e., subscription) business model. We used to sell tools, then we started renting them out. The software itself didn’t change much, though. Now, the software is finally catching up to the service model. Like other service businesses, software businesses can now sell outcomes.
Every problem has potential solutions in the form of tools, a tools-as-a-service, and outcomes-as-a-service.
Let’s use laundry to understand these distinctions better:
- A washing machine is a tool: it helps you to get your work done more efficiently, but as an owner of one, you still need to do the work yourself.
- A laundromat is a tool-as-a-service. You pay to use tools owned and managed by someone else. Yet again, you need to do the work yourself.
- Dry cleaners are an outcome-as-a-service. You pay for someone else to do the work. You’re not paying for a tool, but for the outcome of clean clothes.
In digital, businesses typically either rent software tools for their in-house team, or they pay an agency to do the work for them. For example, you can pay for a design tool like Figma, or you can just pay a design agency to do the work for you.
What’s new in 2025 and beyond is the ability to buy the outcome not from an agency, but directly from your software vendors. This is because Large Language Models make software an order of magnitude more capable.
Your software product today is a tool, and it might already be possible for you or a competitor to build an AI-first product that does the work for your customers without any ongoing effort on their part. If not, it’ll probably be possible soon. Either way, a level of disruption that was not possible a few years ago is now possible in almost every software category1.
Footnotes
This is not to say that, in the terms of Clayton Christensen’s The Innovator’s Dilemma, Large Language Models are more of a disruptive (i.e., better for startups) than sustaining (i.e., better for incumbents) innovation. In terms of SaaS, I think AI is probably more of a sustaining innovation, that will nonetheless be disruptive to many lazy and ineffective incumbents. ↩︎