Category explainer

What is Service as Software?

For thirty years, business software has followed one shape: a company builds a tool, your people learn the tool, your people do the work with the tool. Software as a Service changed how that tool was hosted and billed — but the work never moved. It stayed with your team.

Service as Software is the inversion. The software does the work. You don’t licence a tool and staff it; you buy an outcome — invoices reconciled, reports produced, documents processed — and the system that delivers it happens to be an AI agent rather than a person at a desk.

The definition

Service as Software: a service performed by AI agents, delivered and managed by a provider, bought and measured by output — not by seats, licences or usage.

The unit of purchase is the giveaway. SaaS sells you access (per seat, per month). Service as Software sells you work done (the month-end pack delivered, the day’s invoices matched, the inbox triaged) — with accuracy and turnaround agreed up front, like any service contract you’d sign with a firm.

SaaS vs Service as Software

SaaSService as Software
Who does the workYour team, using the toolThe agent
What you pay forSeats and tiersThe service, monthly
What you manageLicences, training, adoptionNothing — you review output
What you measureUsageOutput, accuracy, turnaround
When volume growsMore seatsSame service, agreed price

Why this is possible now

Two things converged. First, AI models became capable of genuine knowledge work — reading a contract, matching an invoice to a purchase order, compiling a report from six systems — not just chatting about it. Second, the engineering around agents matured: supervision, audit trails, exception handling, human checkpoints. The first without the second is a demo. Together, they’re a service you can put a name and a number on.

The second part is where the “managed” matters. An unsupervised agent is an intern with admin rights. A managed agent has every action logged, every exception routed to a human, every month’s output measured against the agreed standard — and an engineering team accountable for it. That’s the difference between buying AI and buying work.

What to put an agent on first

The best first candidates share a profile: high volume, clear rules, mostly judgement-free, soul-destroying for the humans currently doing it. Invoice matching. Report assembly. Document intake. Order-status chasing. Data-quality checks.

Not the work your experts should do — the work that stops them doing it.

What it looks like month to month

A defined service. A simple subscription — ours start at £1,500 a month on annual terms. A monthly report: volumes handled, accuracy, exceptions raised, time saved. A standing review where the service gets extended, tightened or improved. Your data stays in your systems; the agent comes to it.

If that sounds less like software procurement and more like hiring a very fast, very consistent department — that’s exactly the point.

Pick the service that eats your team's week. We'll show you an agent doing it.