Engagement 00 — The entry point
AI Value Sprint
You have a portfolio of AI initiatives and a funding gate coming. The models are not the first thing to compare.
Ten working days. One business domain, five to eight initiatives. Each one leaves with a verdict: stop, fix, or scale, with the evidence behind it. This is the smallest thing you can buy from this practice, and it is deliberately shaped as a decision, not a study.
Leaders who must decide which AI initiatives get funded next.
Chief data and AI officers, CIOs, COOs and transformation executives in banks, insurers and other regulated enterprises, with several AI initiatives in flight and a funding, scale or production decision inside the next ninety days. Business-unit executives who carry the outcome those initiatives are supposed to move.
Less useful if you are looking for an AI readiness assessment. This engagement does not score maturity. It changes a funding decision.
Detection improved. Response didn't.
Most enterprises now detect faster than they respond. Pilots impress in the demo and stall before the P&L, not because the models fail, but because the surrounding operation cannot commit and absorb inside the window the signal allows. Industry analysts now expect a large share of agentic AI projects to be cancelled for unclear business value. Those failures are predictable, and measurable, before the money is spent.
A portfolio decision, evidenced.
Stop / fix / scale verdicts
One verdict per initiative, each with its evidence trail. The verdict set includes "AI makes this worse", and it is used when the measurement says so.
One measured signal-response chain
The distance from signal to committed response, measured end to end from your systems' own timestamps, against the window the signal allows.
Binding-constraint diagnosis
Where the delay actually binds: data, decision authority, or the operation's capacity to absorb the change. Named, with evidence.
Ninety-day sequence and funding recommendation
What to do first, what to stop paying for, and a recommendation your CFO can interrogate line by line. Optional instrumentation plan for the chain that matters most.
The diagnostic is allowed to say no.
A recommendation engine that cannot refuse is a sales instrument. This is a measurement instrument. One of its five verdicts is that AI makes the situation worse, because accelerating detection while commitment and absorption stay fixed grows the backlog faster. The sprint is delivered with the SRD Value Diagnostic, and the method's structure is published openly at soaringwings.ai; you can read exactly how the verdicts are reached before you spend anything.
Ten working days, fixed.
Days 1 to 3: frame
One domain, the candidate initiatives, and the signal families they claim to serve. Access to the timestamps, not to everything.
Days 4 to 7: measure
One signal-response chain instrumented from system evidence. Initiative claims tested against the measured constraint.
Days 8 to 10: decide
Verdicts, sequence, and the executive session where the funding recommendation is put on the table and defended.
Fixed fee: R120,000. Ten days. One decision your next funding gate depends on.
Nothing has to. The sprint is complete in itself. Where a verdict says scale, the natural next step is an instrumented pilot: the selected intervention run against a permanent holdout, with outcomes measured from your systems rather than the vendor's dashboard. That is a separate decision, made after you have seen the sprint's evidence.