What an AI agent actually costs to run

After handover, a WeBuild agent has two running costs: model usage (the API calls to the AI provider) and tool subscriptions (anything the workflow needs that you don't already pay for). For most single-workflow agents, the total lands around A$50–300 a month. These are market figures for typical API pricing at typical SMB volumes — your spec carries your number, in writing, before the build starts.

Why model costs stay small

An agent working a workflow reads and writes text in bursts — an inbound email, a draft, a CRM update. At SMB volumes (dozens to hundreds of items a day, not millions), that's a utility bill, not a line item that needs a meeting. Costs scale with volume, so the spec states the assumption: if your volume 10×es, your model bill grows with it — a problem you'll be delighted to have.

The subscriptions you already pay

Agents are built into the stack you already run — Gmail, your CRM, your calendar. We don't add platforms for the sake of it; every new subscription in the spec has to justify itself against the job. Often the answer is zero new tools.

The cost nobody itemises: lock-in

Rented automation hides its real price in the exit: leave the platform and the flows stay behind. An owned system inverts that — the build is priced once, the running costs are metered and visible, and walking away from us costs you nothing, because the system is yours. The model provider itself is swappable too; that's the point of owning the system layer. The pricing structure — session, fixed-quote build, optional retainer — is published on the pricing page. And if the number you're really weighing is agent-vs-hire, the full AUD comparison is in the in-house cost guide.