Service-as-software is AI that does the work instead of helping someone do it. You stop buying a tool your team has to run, and start buying the finished result: the policy checked, the invoice posted, the renewal file ready. The shift matters because it moves AI out of the software budget and into the much larger budget businesses spend on people and outsourcing.
Where the term comes from
Venture firm Foundation Capital popularized the idea in 2024. Its argument is that software is no longer just a tool for organizing work; it is becoming the worker. It sizes the opportunity at $4.6 trillion over five years, as AI takes on work currently paid for from in-house salaries and outsourced services.
The comparison it uses makes the point. Salesforce earns about $35 billion a year. Businesses spend about $1.1 trillion a year globally on sales and marketing salaries. Software that does the work competes for the second number, not the first.
What changes for the buyer
With traditional software:
- You buy licenses or seats
- Your team learns the tool and does the work
- You carry the risk that it does not deliver
With service-as-software:
- You buy completed outcomes
- The provider runs the work, including the exceptions
- The provider carries delivery risk, and is paid when results are verified
That last point is the real change. When the provider is paid for results, it has every reason to make the work reliable, and none to sell you more seats.
Why investors are buying service firms
A second version of the idea is playing out in private markets. Instead of building AI and selling it, investors are buying traditional service businesses and running them with AI. The playbook is to buy firms running at thin margins, automate a large share of repetitive tasks, and compound through more acquisitions. Capital committed to this strategy now exceeds $3 billion across the major players, according to one 2026 playbook, in accounting, legal, call centers, IT services and more.
What to check before you buy it
Service-as-software is easy to claim and harder to deliver. Four questions separate the real thing from a tool with a new label:
- What exactly is the outcome? It should be written down, in business terms, with evidence required.
- Who handles the exceptions? Real services have people on the edge cases. Tools send them back to you.
- Can you see the evidence? Every billed outcome should link to a record of how it was done.
- Who carries the running cost? If you pay for model usage on top, you are still buying software.
How we apply it
Agentic MSP is service-as-software for back-office work. Agents do the routine volume, our exception desk handles what the agents are unsure of, your team keeps licensed decisions, and you pay for verified outcomes. Our guide to agentic MSPs covers the model in more detail.