AI automation that earns its place in the business.

A single scoped automation is typically $4,000–$15,000 to build, depending on how many systems it touches and how much human review the output needs. Running cost is model and API usage — usually tens to low hundreds of dollars a month — plus optional monitoring. We give a fixed range in writing after a discovery call, not an open-ended hourly arrangement.

Sometimes. It pays back on processes that run often, follow a mostly predictable pattern, and currently consume real staff hours — document handling, inbox triage, lead qualification, routine lookups. It does not pay back on low-volume work, on tasks where an error is expensive and cannot be reviewed, or where a setting in software you already own would do the same job.

Rules-based automation moves information and triggers actions along a fixed path — it is predictable and cheap where the steps never vary. AI automation handles the parts that resist fixed rules: reading a document that is laid out differently every time, drafting a reply, classifying a message by intent. Most real systems use both, with rules doing the deterministic work and a model doing the judgement, behind a review step.

Not always, but it has to be in the scope. If the information a process needs currently lives in someone's head or in an email thread, the first piece of work is building the record the automation will read. That is legitimate work — it just belongs in the estimate rather than being discovered mid-build.

A focused automation is usually two to six weeks from kickoff to production, including testing against your real cases. Larger programmes are split so one process is live and measured before the next one starts.

Not under the configurations we deploy. We use business-tier API access where the provider's terms exclude training on your inputs, and we document exactly which provider handles which data.