Outcomes Over Tools: What Comes After SaaS

What's the better business: building the AI that does the work, or being the company that uses AI to do the work?

In December 2024, I published “In Search of a Moat” and “The Appeal of Services as a Software,” exploring how AI was putting pressure on traditional SaaS and what might replace it. Fifteen months later, that question is at the center of a much bigger conversation. Y Combinator’s Spring 2026 Request for Startups explicitly calls for “AI-Native Agencies” built on this model. The broader venture ecosystem has followed. But I think the interesting story isn’t that VCs agree with each other. It’s that we’re watching a fundamental shift in how companies decide what to keep in-house and what to hand off to someone else.

The Move Toward Managed Services

AI is making it possible to build highly specialized companies that can execute a specific workflow or function better, faster, and cheaper than an internal team. When that happens, it stops making sense to do that function in-house. Companies outsource it. But unlike the old outsourcing model, where you hired a consulting firm with hundreds of people and waited weeks for deliverables, these new providers operate more like software companies internally. AI handles the execution. Humans handle the quality and the relationships. The client gets a managed service. The provider runs on software economics.

This is really just specialization taken to its logical end. When someone can do one thing better and cheaper than you can do it yourself, you let them do it. AI amplifies that dynamic. A company that focuses entirely on contract review, or ad creative, or financial reporting can build AI workflows purpose-built for that function. They develop domain expertise that compounds over time. They get better at the thing while their clients are busy doing a dozen other things.

Meanwhile, the SaaS model that once served these functions is under real pressure. AI has made it so easy to build software that competition floods every category. When everyone can ship a credible product in weeks, the moat has to come from somewhere other than the product itself. For a growing number of companies, that somewhere is the service layer on top. The quality of service, the customer support, the relationship with the client. Those become the differentiators when the underlying technology is increasingly commoditized.

Process-First vs. Outcome-First

One of the more interesting splits I’m seeing is between companies that approach this from a process-first mindset versus an outcome-first mindset.

The process-first companies are trying to automate every step in a workflow end to end. Map out the process, replace each step with AI, minimize human involvement wherever possible. It’s a clean vision on paper. But in practice, it tends to run into resistance. Some of those steps exist because they require real judgment. Others are the parts of the work that people actually enjoy, the client conversations, the strategic thinking, the creative problem-solving. Automating those doesn’t just create a technical challenge. It creates a product that people push back against.

The outcome-first companies look at it differently. They don’t care which steps are automated and which aren’t. They care about delivering the result. The client isn’t buying a process. They’re buying a finished contract, a completed audit, a tested campaign. How much of that is AI and how much is human is the provider’s problem to solve, not the client’s.

This distinction matters because the managed services model is inherently outcome-first. The whole point of outsourcing a function to a specialized provider is that you’re paying for the deliverable, not managing how it gets produced. That gives the provider the freedom to figure out the right balance of AI and human involvement for each domain, and to adjust that balance as the technology improves, without the client ever needing to care.

The companies that seem to be getting this right are the ones that think this way. They use AI aggressively on the execution layer but don’t try to eliminate the human judgment and relationships that their clients actually value. The result is a service that feels premium to the client and scales like software for the provider.

What This Means for Founders

The biggest opportunity isn’t necessarily in the domains where capital is already concentrated. It’s in the ones with comparable market size but far fewer players. Go deep in a single function, build real expertise, and compound trust with customers over time. That’s the moat now. Not the software. The specialization. And when the technology is commoditized, the service and support you wrap around it becomes your competitive advantage.

Building from scratch isn’t the only path either. One of the more interesting approaches I’ve been seeing is the roll-up model: acquiring existing services businesses that already have client relationships, domain expertise, and proprietary data, then using AI to scale what they’re already doing. You fine-tune models on their data, automate the repeatable parts of the workflow, and suddenly the unit economics look completely different. It’s an outcome-first approach by default, because the business you’re acquiring was already delivering outcomes before you added the AI. You’re not replacing what they do. You’re making them dramatically better at it.

Whether you’re building from the ground up or acquiring your way in, the core question is the same: can you compound specialization and trust in a specific domain faster than the market moves? I think the founders who can will build some of the defining companies of this era. Not because they had the best AI, but because they understood that the real moat was never the technology. It was the relationship.