# Same-role AI sourcing pilot

Metix editorial worksheet. Compare two finalists on one approved, genuinely open role. This is not a vendor benchmark. Do not send the same person competing outreach or upload confidential candidate data unnecessarily.

## Before the pilot
- Role / approved brief version:
- Hiring manager / review owner:
- Products, exact SKUs, users, geography, currency and billing terms:
- Same pilot dates and approved contact policy:
- Must-have evidence rubric and permitted equivalent evidence:
- Minimum manager-accepted relevance; maximum recurring operator hours:
- ATS connector scope and approval / pause controls demonstrated:
- Unknowns that block purchase:

## How to complete the CSV
Import ai-sourcing-comparison-worksheet.csv into a spreadsheet. Keep missing numbers blank and name them in unknown_cost_components. Enter 0 only for a confirmed absent cost. Keep distinct candidates distinct from messages and follow-ups. Keep uncertain profiles in pending_evidence rather than silently marking them as rejected.

First-year cost = subscription commitment + setup/migration + usage/add-ons + retained ATS/overlap + recurring annual operator hours × internal hourly cost. Count a bill once; do not add an existing ATS twice. Cash due at signing is a separate question from a yearly cost. Confirm invoices, cancellation and renewal in the quote.

Hypothetical arithmetic only: $2,400 subscription + $600 setup + $1,200 retained ATS + 104 hours/year × $50/hour = $9,400 first-year cost, assuming no other charges. If usage is unknown, $9,400 is a known subtotal, not a completed budget. The figures are invented, not vendor prices, savings, salaries or customer results.

Evidence-complete rate = profiles with evidence for every must-have / unique profiles reviewed. Manager-accepted rate = profiles the manager accepts for contact / unique profiles reviewed. Report pending-evidence count beside both. Qualified-interest rate = unique people who meet your agreed criteria and reply with interest / unique people sent approved outreach. A reply, a booked interview and an attended interview are separate results. If a denominator is 0 or unobserved, rate = unknown.

Separate setup hours from recurring search/review, reply handling and booking. Record a source URL or screen reference next to each price/feature. A polished demo or large profile count is not role-level proof.

## Review decision
- Which agreed thresholds passed, with denominators:
- What remains unverified:
- Which work is still owned by the recruiter:
- Decision: continue / change the brief / stop; owner and review date:

For a Metix self-serve evaluation, create one real role, review the structured brief and candidate evidence, then approve outreach. Your team interviews and decides. Staff-run Delivery is a separate quote. Downloading this worksheet is not role activation.
