Do the agents touch our code autonomously?
No. Every run works in an isolated sandbox and stops at the gates: no change reaches one of your branches without explicit human approval.
Do we need in-house AI expertise to use it?
Not really. The console translates everything into plain language: plan, files, checklist, costs, and the underlying tooling stays our job, not yours. What does matter is describing well what you need: clear, spec-driven tasks make the difference between a run that lands and one you have to send back.
Why can't we just switch it on and use it right away?
Because it wouldn't work. The pipeline must be designed around your case, your processes and your risk thresholds, and corrected after the first real runs. That's why the path starts with consultation and readiness.
How long does onboarding take?
It depends on the outcome of the readiness assessment: well-organised systems shorten the timeline considerably. Scope and milestones are defined together during the consultation.
What happens if a run produces a wrong result?
It stops at the gate. You can request a corrective iteration with a note, test it yourself in the sandbox, or close the run without anything being touched.
Does it only work for software development?
No. The same engine orchestrates other kinds of processes: any flow whose components communicate can become a graph, with or without AI nodes. See the Automation page.
How much does it cost?
It depends on the scope: it's a tailored service, not a list-price licence. After the initial consultation you'll have a clear picture of costs and timelines, before committing.