# AI Benchmark Sprint Proposal Template

## Objective

Benchmark one production-relevant AI workflow before deployment or vendor commitment.

## Sprint Scope

- Workflow:
- Buyer decision:
- Candidate systems:
- Baseline process:
- Success metric:
- Failure cost:

## Milestones

| Time | Milestone | Output |
| --- | --- | --- |
| Day 0 | Discovery | Confirm workflow, success metric, risk boundary, data packet, and decision deadline. |
| Day 1 | Task packet | Write prompts, expected outputs, scoring rubric, holdout policy, and artifact schema. |
| Days 2-4 | Runs and review | Run candidate systems, capture traces, review failures, and calculate cost/latency envelope. |
| Day 5 | Decision memo | Deliver ranked recommendation, risk register, next tests, fallback plan, and production-readiness note. |

## Deliverables

- Task packet with expected output and evidence requirements.
- Model/provider run table with cost, latency, and reviewer notes.
- Trace artifacts and replay/inspection path.
- Risk register and deployment recommendation.
- Follow-up plan for real pilot, routing, or system redesign.

## Owner Routing

- Sanjay Prasad: Benchmark design and technical delivery
- Saujas: Sales engineering and client solutions