Questions we hear before every engagement.

Audits are $5K–$15K and take one to two weeks. Workflow Automation builds run $25K–$75K over four to eight weeks. Embedded AI Ops retainers are $10K–$20K per month with a three-month minimum.

Each tier is priced on scope, not hours. You'll know the number before the work starts. If we think your workflow needs a smaller engagement than what you came in asking for, we'll say so.

The person on the sales call is the person in the Slack when the system ships. We don't pyramid — there is no junior team behind the demo who quietly takes over in week two.

That's load-bearing for us. It's why the model doesn't scale the way an agency wants it to, and it's why the work lands.

Then that's the answer. Audits include a "don't build this yet" section when that is the candid read. We've used it three times in our last five audits.

Usually the reason is the data layer. The retrieval store isn't ready, the document quality is too uneven, or the workflow itself has too many implicit exceptions to be worth automating yet. We'll tell you what to fix first and how long that should take, and you can call us back when it's done.

No. Most of our day-to-day contacts are operators — heads of ops, GMs, founders who haven't touched code in years. The work is technical; the conversation doesn't have to be.

That said, you'll need someone on your side who can sponsor the work internally — typically a head of platform, ops lead, or technical co-founder. We don't replace that role. We work alongside it.

We've shipped in healthcare and financial services. The pattern is the same: we add a compliance review step to the audit, scope the build to keep regulated data inside your existing controls, and document the audit trail explicitly.

If your environment requires us to sign a BAA, MSA, or vendor security review before we can start, expect the audit phase to add one to two weeks for legal. We'll flag that in the first call.

Claude-first, model-agnostic in practice. We default to Claude for most workflows because the operating characteristics fit how we like to ship — but we've shipped against GPT, Gemini, and open-weights models when the evals pointed there.

Part of every engagement is a model-swap policy that lets your team change the underlying model when a better one ships, without rebuilding the workflow.

We're an early-stage operation and we cap concurrent engagements deliberately. A brief lets us match new work to capacity that's actually available rather than overcommit and miss the day-60 ship date that defines us.

Practically: most briefs get a response within a business day with either a slot, a referral to someone who can start sooner, or an honest "not yet — here's when."

Day-one operability is the bar. The morning after we leave, your team should be able to extend the system without us in the room. That means a runbook, a tested rollback path, eval ownership transfer, and a model-swap policy documented for the next quarter.

If you want continuous capacity afterwards, the Embedded AI Ops retainer is designed exactly for that — fractional senior time, quarter-by-quarter, no long lock-in.