Play 01
The Diagnostic
Two weeks of frank scoping inside the institution. We map the workflows that actually compound, the leakage points that don't, and the data that already exists but isn't yet trusted.
2 weeksto a written diagnostic
04 · Automation
Eight plays. Compounding leverage.
Automation is not a feature. It is the operating system we install inside an institution. The eight plays below are the moves we run, in the order we run them, against the workflows that compound.
The Eight Plays
The moves we run, in order.
Play 01
The Diagnostic
2 weeks
to a written diagnostic
[ Play 01 ]
The Diagnostic
Two weeks of frank scoping inside the institution. We map the workflows that actually compound, the leakage points that don't, and the data that already exists but isn't yet trusted.
[ Play 02 ]
The First Workflow
We pick one workflow with a measurable outcome — not the most strategic one, the most teachable one. We ship it to production in eight to twelve weeks, with the operator's name on the dashboard.
[ Play 03 ]
The Continuous Audit
Once one workflow is live, we wrap it in a continuous audit layer. Every accrual, every decision, every model output is logged, sampled, and reviewable.
[ Play 04 ]
The Anomaly Layer
Sub-minute anomaly detection on the inflows and outflows that matter. Alerts go to the operators who can actually act on them, not to a generic ops inbox.
[ Play 05 ]
The Classification Engine
Document classification across the long tail of templates the institution actually receives. The classifier learns from corrections; the corrections become training data; the loop closes.
[ Play 06 ]
The Decision Surfaces
Briefings, queues, and decision surfaces for the operators and principals who consume the output. Citation-grounded, never opaque, always traceable to source.
[ Play 07 ]
The Capacity Transfer
We train the institution's own engineers and analysts to operate, extend, and audit the system. The goal is not vendor lock-in — it is institutional capacity.
[ Play 08 ]
The Compounding Review
Quarterly reviews against the outcomes set in the diagnostic. Plays compound — each new workflow benefits from the audit, classification, and decision layers already in place.
Play 02
The First Workflow
We pick one workflow with a measurable outcome — not the most strategic one, the most teachable one. We ship it to production in eight to twelve weeks, with the operator's name on the dashboard.
12 weeksto first production workflow
Play 03
The Continuous Audit
Once one workflow is live, we wrap it in a continuous audit layer. Every accrual, every decision, every model output is logged, sampled, and reviewable.
100%audit coverage on covered flows
Play 04
The Anomaly Layer
Sub-minute anomaly detection on the inflows and outflows that matter. Alerts go to the operators who can actually act on them, not to a generic ops inbox.
Sub mindetection latency
Play 05
The Classification Engine
Document classification across the long tail of templates the institution actually receives. The classifier learns from corrections; the corrections become training data; the loop closes.
95%+classification accuracy
Play 06
The Decision Surfaces
Briefings, queues, and decision surfaces for the operators and principals who consume the output. Citation-grounded, never opaque, always traceable to source.
0ungrounded outputs
Play 07
The Capacity Transfer
We train the institution's own engineers and analysts to operate, extend, and audit the system. The goal is not vendor lock-in — it is institutional capacity.
12+operators trained per engagement
Play 08
The Compounding Review
Quarterly reviews against the outcomes set in the diagnostic. Plays compound — each new workflow benefits from the audit, classification, and decision layers already in place.
4×/yrcompounding reviews
How We Deploy
Plays compound over years.
Year 1
- Diagnostic
- First three plays
- Capacity baseline
Year 2
- Plays 4–6
- Cross-workflow integration
- Throughput review
Year 3
- Plays 7–8
- Institutional capacity transfer
- Compounding review
The ninety day commitment: in the first ninety days of any engagement, we ship one production workflow with a measurable outcome on a dashboard the operator already trusts. No exceptions.