The AI readiness audit we run before writing any code
A two-week discovery sprint that has killed more projects than it has started — and saved clients a quarter of wasted engineering each time.
Most failed AI projects were unwinnable on day one, and it was knowable on day one. The audit exists to find that out cheaply.
Four questions, in order
1. Does the data exist, and can we reach it?
Not "is it clean" — clean is our job. The question is whether the signal is recorded at all. If the reason a decision is made lives in a person's head and never touches a system, no model can learn it.
2. What does a human doing this today cost, per unit?
You cannot justify an agent without a baseline. Time per ticket, error rate, rework rate. If nobody has measured it, measuring it is the first deliverable.
3. What happens when it is wrong?
Reversible and cheap, or irreversible and expensive? This decides how much guardrail engineering the project carries, and it often moves a use case from "phase one" to "phase three".
4. Who owns it after we leave?
An AI system is not a website. It drifts. Somebody on your side has to own the evaluation set and read the monthly report, or quality erodes quietly.
The output
A ranked backlog with a cost and a confidence attached to every line, and an explicit list of the ideas we recommend not doing. Clients tell us that second list is the more valuable one.
The Reciprocal Solutions
Consulting Team
We design, build and operate AI agents, automations and custom LLM systems — and stay on the hook for how they behave in production.
Got a workflow that should be running itself?
Bring us the process, the constraints and the mess. We will tell you honestly whether AI is the answer — and what it would take to ship it.