Automate vs. Backfill: The Two Open People Reqs
Source: core/clients/madison-reed/work/MR-people-hires-automate-vs-backfill-2026-06-26.md
Purpose: Internal confidential analysis that critiques the two open People team hires and recommends automating the processing core instead of backfilling with FTEs (Full-Time Employees). This is why your role exists as an embed instead of a traditional hire.
Sensitivity: This doc contains internal CMPRSSN framing. The client-facing versions are the embed proposals. Do not share this analysis directly with MR.
The situation
Two People team members left: Roxy ($95,700 salary) and Beth ($76,000 salary). Combined freed capacity: $171,700/year. MR's plan was to fill both seats with two new FTEs at roughly the same cost, for a net-zero hard-dollar change.
CMPRSSN's recommendation: don't hire two administrators. Automate the processing core (most of which is already built), backfill the genuine human residue with part-time/contract capacity, and hire one builder (not two processors).
Role A: People & Culture Generalist (Total Rewards, Risk Management & Compliance)
This is the role Kyle is filling in the embed (Benefits & Leave seat).
Verdict: Re-scope and downsize. Do not hire as written.
The JD (Job Description) was updated by Grecia on June 26 to add workers' comp and workplace safety (claims management, OSHA, injury prevention, ergonomic assessments, post-incident investigations). This is NOT processing work; it's field-present, investigation-heavy judgment tied to real outcomes (workers' comp mod factor, which affects insurance premiums).
What gets automated vs. what stays human
| JD scope | Bucket | Notes |
|---|---|---|
| LOA / FMLA / ADA intake, tracking, return-to-work | Automate now | Leave agent, live in Phase 2 |
| Benefits-deduction & billing reconciliation | Automate now | Pay reconciliation agent, live |
| Multi-state minimum wage / premium-pay monitoring | Automate now | Minimum wage monitor, live |
| Legislative monitoring (ACA, COBRA, ERISA, HIPAA) | Automate now | Change-detection agent, low build cost |
| Compliance calendar (postings, filings, annual notices) | Automate now | Cron-style monitor |
| HR/payroll data-integrity audits, compliance dashboards | Automate now | Audit + reporting agents |
| Comp-benchmark data pulls | Automate now | Agent feeds human who sets philosophy |
| Benefits enrollment, open enrollment config | Automate later | Needs write-access + human-in-the-loop + carrier integration |
| Workers' comp claims intake, carrier coordination | Automate later | Partially automatable; legal sensitivity |
| Comp philosophy, pay-equity decisions | Keep human | Judgment, not data pulling |
| Workplace safety strategy, injury prevention, field work | Keep human | Requires physical presence on the retail floor |
| Post-incident investigations, corrective action | Keep human | Employee relations judgment |
| Broker/carrier relationships, handbook policy, ADA interactive process | Keep human | Relationships, exceptions, defensible decisions |
Recommended staffing
Instead of one $76K FTE: a fractional or contract Workers' Comp & Safety specialist (est. 0.4-0.6 FTE) plus Grecia elevated to own comp philosophy, vendor governance, and compliance strategy with agent fleet underneath.
Role B: People Systems & Learning Administrator
This is your role (the embed).
Verdict: Split the role. Do not hire one $72-78K person to do both halves.
The JD spans too wide an altitude band. It describes an administrator ($72-78K, "1-3 years experience," daily Paylocity data entry) and then staples aspirational builder duties onto a "12-18 month horizon." Hired as written, it fills at the lower altitude, and the agentic layer has no owner.
What gets automated vs. what stays human
| JD scope | Bucket | Notes |
|---|---|---|
| HRIS lifecycle processing (hires, job changes, terms, LOA entries) | Automate now | Lifecycle agent fleet |
| Onboarding logistics (access setup, Day-1 materials, I-9 routing) | Automate now | Onboarding orchestrator |
| Compliance-training auto-launch + completion tracking | Automate now | Anti-harassment auto-launch agent |
| LMS enrollment, completion reporting, utilization dashboards | Automate now | Learning-admin work as agents |
| Data-integrity discrepancy flagging | Automate now | Lightweight monitor (already built) |
| Docebo config + content migration from Wisetail | One-time project | Fixed scope, contractor-shaped work |
| "Manual coordination bridge" until Workday | Temporary | Explicitly temporary; contract work |
| Integration build-out at Workday go-live | Later (2027) | Gated on Workday |
| Integration architecture, agent-workflow design, evals, alerting, permissions, fleet builds, internal platform expert | Keep human (builder) | This is the real job. It's a forward-deployed engineer role, not a $72-78K administrator. |
Why the embed works
Instead of hiring an administrator who does the manual work forever:
- Daniel (you) sits in the seat for one quarter, does the manual work, and automates it
- The builder half happens by default because that's what an engineer does
- At the end of October, the manual work runs on agents, and MR decides what (if anything) still needs a permanent seat
The combined math
| Approach | Annual cost | What you get |
|---|---|---|
| Two FTE hires (as planned) | ~$151K/year (ongoing) | Two people doing the work by hand. Net zero savings vs. Roxy + Beth. |
| CMPRSSN embed + fractional | ~$90K one-time (the quarter) + fractional WC specialist | Processing on agents, genuine residue correctly staffed, builder work actually done |
Integrity note
CMPRSSN flagged its own incentive divergence in this doc: the faster the hand-off, the sooner CMPRSSN loses the revenue. The doc explicitly resolves this in MR's favor: the engagement is fixed-term, the hand-off is the deliverable, and the success condition is a function that runs without CMPRSSN.
Risks (the honest case against)
- Live agents may leave a larger residue than expected. The Compliance pod measurement is the test.
- Fractional workers' comp/safety specialist: deskless retail floor creates real liability. The part-time person must have genuine coverage.
- Contractor doing manual HRIS bridge for 12+ months (until Workday) carries knowledge-retention risk.
- If MR can't hire a builder at senior altitude and won't let CMPRSSN operate the layer in the interim, fallback is the administrator hire as written (accepting slower agentic maturation).