What does business process automation cost?

Automation is priced by four things: how many systems the process crosses, how many exceptions it has, how much of the work needs judgement rather than a rule, and what a failure costs you. Software licences are rarely the largest line. The engineering that makes a fragile process survive production is.

So no single figure gets published honestly. You can get close by pricing those four drivers yourself, in that order.

How many systems does the process cross?

A process that touches one system is small. A process that moves data between four systems, two of them with no API and one of them a spreadsheet somebody maintains by hand, is most of the cost of the project.

You are paying for integration surface. Workflow length barely moves the price.

How messy are the exceptions?

The happy path is cheap and the exceptions are the work: the document that arrives without an order number, the payment that partially fails, the two records that conflict. Every exception has to be decided, coded and logged.

A process with a dozen undocumented edge cases takes longer to build than one with ten steps.

How much judgement does the work require?

Rules are deterministic code and stay cheap. Interpretation is where an AI agent enters, and it brings its own costs: evaluation, monitoring and per-model usage. Messy documents, ambiguous intent and prioritisation are the usual cases.

Mixing the two badly is the most common way budgets inflate. The line between them is drawn in what an AI agent is.

What does a failure cost?

A missed report is an annoyance. A duplicated invoice, a lost lead or a mis-sent client document is money. The higher the cost of failure, the more the build spends on error handling, idempotency, alerts and human approval steps.

An automation nobody trusts gets switched off after the first incident, and that spend is what keeps it running.

What costs continue after the build?

Ownership moves cost from licences to operations: a server, backups, updates, monitoring and a named person answerable for the system. In most small businesses that line comes in under the per-seat subscription it replaced, and it is not zero. Open source versus SaaS total cost of ownership works through the arithmetic.

How do you get a real number?

Write down the process as it runs today: which systems it crosses, who does it, how often, what the exceptions are and what happens when it goes wrong. Any honest vendor prices from that description. Most "how much does automation cost" conversations skip it and end in a range so wide it is useless.

If a vendor quotes before reading it, you have a guess with a deposit attached. We price a fixed scope per process after a short paid assessment, and that assessment is credited against the build. Tell us about the process and you will hear whether it is worth automating at all.