The Enterprise software loop: ASP, SaaS, and the Dawn of Service as a Software
software ate the world and AI agents will eat the software
Thirty years of enterprise software history really boils down to the same struggle repeating itself:
how do we deliver value without dragging customers through the operational mess behind it?
ASP tried one answer. SaaS tried another.
Now we’re in the third attempt — and this one genuinely changes the equation.
Act I: The ASP Dream That Died (1996–2003)
Application Service Providers had a simple pitch: don’t buy software, just rent it over the internet. No CDs, no on-prem servers, no giant maintenance contracts. Pay monthly, we’ll run it.
I actually started my career at Aztec, one of the better-known ASP players in the late ’90s. If you were around back then, you know the optimism we had :)
But the model collapsed under its own weight.
Every customer got their own server. Broadband wasn’t reliable. Most ASPs were reselling software they didn’t build, which meant they controlled everything except the product.
Then the dot-com crash hit. The idea wasn’t wrong — the timing was.
Act II: SaaS Gets It Right (2004–2020)
The next generation rebuilt from zero, and the key insight was multi-tenancy.
One codebase. Many customers. Clean data separation. Suddenly the economics worked: high margins, simple updates, global distribution.
Salesforce, Slack, HubSpot — the entire SaaS wave grew on the same promise:
subscribe, and we’ll make sure this keeps improving.
But SaaS had one limitation baked into the model. It gave you capability, not outcomes.
You still needed:
a RevOps team to run your CRM
analysts to interpret data
support teams to run the support tools
SaaS removed the infrastructure headache.
It never removed the human labor that operated the software - frankly, it didn’t even care for it (the renewal is all the teams cared about).
Act III: Service as a Software — Agents Take Over (2025–)
This decade flips the model.
ASP: “We’ll host the software.”
SaaS: “We’ll build and maintain the software.”
Service as a Software: “We’ll deliver the outcome. Start to finish.”
A lot of the business logic will move to a new multi-agent tier - Satya Nadella
AI agents don’t wait for human instructions. They plan, decide, and execute workflows independently. The software layer becomes background noise.
The result becomes the product.
And Jensen Huang didn’t bother sugarcoating it:
“The IT department of every company will be the HR department of AI agents.”
What New Businesses Look Like
Y Combinator captured the shift in one line:
“Vertical AI agents could be 10X bigger than SaaS — because companies spend far more on employees than on software.”
That’s the real unlock.
SaaS disrupted software budgets.
AI agents will disrupt labor budgets.
This creates new categories:
Outcome-as-a-Product
Don’t sell contract-review software.
Sell “reviewed contracts in 2 hours.”
Charge for results, not seats.
Vertical AI Agents
Purpose-built agents that run underwriting, AR, performance marketing, etc.
Moat = domain depth + proprietary data.
Agent Orchestration Platforms
The OS for managing agent fleets — handoffs, conflicts, escalations.
Micro-Workforce-as-a-Service
A “10-person team” that is actually 9 agents + 1 human supervisor.
Looks like staffing. Operates like SaaS.
Arthur Mensch from Mistral put a crude number on it:
“More than 50% of current enterprise software could be replaced by AI.”
And honestly, that feels conservative.
The Loop Closes
ASP had the right intention but wrong infrastructure.
SaaS fixed the infrastructure but abandoned the service ambition.
Service as a Software brings both together — with autonomy doing the work humans used to do.
Tl;dr:
ASP moved the software.
SaaS rebuilt the software.
Service-as-a-Software replaces the human labor around the software.
In Marc style: software ate the world and AI agents will eat the software !
2030: This is how the world will look like
Humans won’t operate tools anymore; they supervise agents.
Software moves from tool → colleague.
Business models shift from license → outcome.
Legal, finance, HR, sales, support — anything repeatable is now a blank canvas for AI-native companies.
The dream ASPs had in 1999 is finally real.
It just took 25+ years, cloud maturity, and autonomous intelligence to close the loop.
We started with “software as a service.”
Then “software as a subscription.”
Now we’re inching toward “just do it for me.”
What’s your take?