Why Pure Software Loses in the Agent Era
Buyers no longer want tools, they want outcomes. The companies that bring agents and humans together as one offering will own the next decade of work.
Section 5 of 9 · By Mike Birtwistle
This section will make software investors uncomfortable. It is also the part most clearly true.
In the Headless Operations era, pure software loses.
The gospel that built fifteen years of valuations
In August 2011, Marc Andreessen published “Why Software Is Eating The World” in the Wall Street Journal. The thesis was simple: every industry would be transformed by software, the winners would be software companies, and the losers would be the incumbents who tried to stay in services. For the next fourteen years, that thesis defined how venture capitalists wrote checks, how founders pitched, and how public markets valued companies.
The gospel was clear. Software margins were higher than services margins. Software multiples were higher than services multiples. Software scaled, services did not. A pure SaaS business at $50 million in ARR could trade at five to ten times revenue. A services business of the same size traded at one to three times. The math was so obvious it became dogma.
Founders learned not to do the work. They learned to build the tool that helped someone else do the work. The work was a tax on the multiple.
What actually changed
The buyer did not change. The buyer always wanted an outcome. They wanted closed jobs, not estimating modules. They wanted a five-star reputation, not a review automation feature. They wanted their week back, not another dashboard.
The trade was that outcomes weren’t deliverable by software, so the buyer settled for tools. Most software companies even forgot the outcome and sold features, because features are what their product actually had. The marketing language drifted into specs and integrations and seat counts, and the buyer learned to translate “ten new automation triggers” into “I might get a few hours back this week.”
What changed is not the buyer. What changed is that the outcome is finally deliverable. For the first time in the history of business software, the work itself can be done by something other than a person at a screen. The moment that becomes true, the buyer stops settling for the tool. They go straight to the outcome.
The winning model
The winning model in the Headless Operations era is not software. It is not services. It is software with a workforce inside it, configured by people who actually know the trade.
Three things have to be true at once:
The agents have to operate in context. A generic agent that “books appointments” is worthless. The value of an agent is not what it can do in the abstract. It is what it does when it knows what kind of appointment, what kind of caller, what kind of business, and what the right move is. Context is the moat.
The agents have to be trained on the customer’s specific business. Industry expertise gets you to 70% of the answer. The last 30% lives in the owner’s head: the way they price, the techs they trust with which jobs, the customers who get a discount, the suppliers who deliver same-day. There has to be a process for that knowledge to leave the owner’s head and enter the system. Without it, the agents are smart in general and stupid in the moment that matters.
The customer cannot be the configurator. Most companies selling agents today are selling tools that require the customer to figure out what to do with them. That is the same mistake the SaaS era made, with new technology. Trade business owners do not want to learn agent frameworks. They do not want to write prompts. They do not want to debug. The winning model brings the configuration, the training, the operation, the exception handling, and the result. The customer pays the bill.
What happens to pure software
The smart pure software companies will pivot. They will start bringing the workforce, the domain expertise, and the configuration. The dumb ones will keep insisting the customer figure it out, and they will lose every customer who values their week more than their software stack.
Either way, the category that wins the next decade is the one that does all three things at once. Specialized agents. Customer-specific knowledge. No configuration burden on the buyer.
That is the model. The rest of the manifesto is what gets built on top of it.