From Services to Software

For decades, startups followed a familiar playbook. Build software, sell subscriptions, scale efficiently. Get as close to pure software as possible. Services were viewed as a compromise — a phase you escaped on the way to becoming a real software company.

Investors loved software because software scaled. Services did not. Or so we thought.

Artificial intelligence is forcing us to reconsider that assumption.

Many of the most interesting AI-native companies today are not starting as software companies. They are starting as service businesses with software underneath, and the line between the two is becoming harder to draw.

What changed

AI has dramatically reduced the cost of delivering expertise.

Not just software. Expertise.

This is the shift that changes where founders can start. For most of the software era, building anything meaningful required years of product development before customers got real value. You had to build the abstraction first and hope customers adopted it.

Today, founders can begin by directly solving customer problems. Manually at first. With AI assistance. With humans in the loop. With workflows that combine software, automation, and expertise in ratios that shift over time.

The customer does not care whether the value comes from software or services. The customer cares that the problem gets solved.

That sentence is more radical than it sounds. It overturns most of what venture capital believed about scalable companies for the last twenty years.

What I saw at Goodfin

I lived a version of this transition as a co-founder and CTO at Goodfin.

Goodfin started by serving accredited investors with a model that combined software and high-touch service — manual onboarding, human investment decisions, advisor-led relationships. We did not pretend to be a pure software company. We did not try to be one.

What we built underneath was the part that mattered. Every onboarding conversation became data about what investors actually wanted. Every investment decision became training material for the AI Concierge. Every exception, every customer question, every manual workflow became a candidate for automation. The service was visible. The software was compounding underneath it.

Over time, patterns emerged. The same questions repeated. The same decisions appeared. The same workflows surfaced across customers. We could see which parts of the service could become software and which parts needed to stay human.

The business did not need to choose between services and software. It needed to be precise about which was which, at any given moment, and to keep moving the boundary.

The structural pattern

What I lived at Goodfin is now playing out across dozens of AI-native companies in regulated industries.

Companies start closer to consulting. Closer to operations. Closer to the actual work. They learn before they scale. They earn customer trust through service before they automate through software.

Then something interesting happens. Services become systems. Systems become products. Products become platforms.

The strongest companies do not simply automate tasks. They compound knowledge. Every customer interaction makes the system better. Every workflow improves the product. Every exception strengthens future automation.

This creates a feedback loop that traditional software companies often struggle to achieve, because traditional software companies are built on the assumption that the product exists before the customer engagement. AI-native companies invert that. The customer engagement is the product, and the software is what crystallizes from inside it.

Why this matters for 434

This is one of the theses behind 434 VC.

I am especially interested in founders who are willing to start as a service in order to become a system. Founders who understand that the workflow is the moat, not the model. Founders who treat the early manual phase as a research advantage rather than an embarrassment.

The companies that emerge from this pattern accumulate something unusual. Workflow expertise. Customer trust. Operational knowledge. Proprietary data. And, gradually, software built around real-world work that no horizontal product can replicate.

These companies often look unscalable to traditional venture investors. They appear too services-heavy, too people-intensive, too far from the SaaS template. That perception is the opportunity. The investors who can see past it are the ones who will back the AI-native firms of the next decade before the rest of the market recognizes what they are.

The line is disappearing

The boundary between services and software is becoming harder to draw, and that is the point.

The future will not belong to companies that pick a side. It will belong to companies that understand how to move the boundary, intentionally, as they grow. Companies that start close to the customer. Companies that learn faster than competitors. Companies that gradually transform expertise into systems that work without them.

Some of the most interesting companies of the next decade will be built in the space between services and software.

That space has a name now. AI-native.