Four short steps, and only the first is required. Plain English is the right format: no technical detail, no prompt writing, nothing to configure.
Your description comes back as a spec within one business day: what the agent will do, what it will not, and what it costs. If an agent is the wrong tool for the job, the answer will say so.
Within one business day you get the spec: what the agent will do, where a human stays in the loop, how long it takes, and what it costs. Price is set on a call once the spec is agreed, never quoted blind from a form.
No. Name the systems you work in and the spec will say how the agent reaches them. If something has no integration, that is usually still solvable.
You do. It runs against your accounts and your data, and the output is yours. Keeping it correct as your process changes is part of the product.
It can, and it usually should not. Agents that do one job are the ones that stay reliable. If you have three jobs in mind, describe all three and the answer will say which to start with.