Diagnosis engagements usually start within 2 weeks of the first call. Build engagements depend on scope — most start within 3–4 weeks of the diagnosis sign-off. If you have a hard launch date, tell us on the first call.

Probably not. We work with one-person operators, agencies, and enterprise teams. The Diagnose tier exists specifically so smaller businesses can start without committing to a full build. If the work is small enough, we'll tell you honestly — sometimes the answer is a workflow tool, not a custom system.

Per project. We're model-agnostic and pick what fits the task: frontier models for reasoning and tool use, open weights when latency, cost or data control matter, vendor APIs when speed-to-ship matters more than control. Orchestration: typed function calls, custom agent loops, and proven frameworks where they earn their keep.

Your data stays your data. We default to deployments in your cloud, NDAs on every engagement, and we'll match your security posture (SOC 2 vendor stack, regional residency, on-prem if required). Model providers chosen with data policies you can read in writing.

We'll tell you. The Diagnose tier exists partly so we can say “this is a workflow problem, not an AI problem” without wasting your money on a build. We'd rather lose an engagement than ship a bad system.

Yes, with the obvious caveats. Modern voice models are good enough for inbound qualification, callback booking, and structured outbound. We won't build something designed to deceive — disclosure, fallback to a human, and clean handoff are part of every voice build we do.

A senior engineer on call, monthly performance reviews, a shared dashboard for evals and SLA, and quarterly roadmap sessions. You can step off Operate any month — no auto-renew traps. Handover documentation is written from day one, not at the exit.