People Team Gap Analysis
Source: core/clients/madison-reed/work/MR-people-team-gap-analysis-2026-05-21.md
Purpose: The diagnostic that framed the entire engagement. Prepared for Brad (budget sponsor) and Shlanda (operating sponsor) in May 2026.
The headline
The MR People team has high AI appetite and a tight August 2026 deadline, but lacks legibility: the four critical workflows live in people's heads with no end-to-end documentation. Everything sits at Class 0.5 on CMPRSSN's Workflow Maturity Matrix. The Q3 target is Class 2 (automation) on all four.
Workflow Maturity Matrix
| Level | What it means |
|---|---|
| Class 0 | Fully tacit. The workflow lives in one person's head. |
| Class 0.5 | Where MR is today. Some documentation exists but it's incomplete, inconsistent, or out of date. |
| Class 1 | Legibility. The workflow is documented end-to-end: every step, system, owner, and decision point is captured. |
| Class 2 | Automation. Rules-based steps run on agents; humans handle exceptions and judgment. |
| Class 3 | Infrastructure. The automation is reliable, monitored, and self-healing. |
| Class 4 | Agentification. Agents own the workflow end-to-end; humans govern and handle edge cases. |
The four-pillar diagnosis
CMPRSSN scores organizations across four dimensions. Here's how MR's People team rated:
People
- AI appetite: High. The team wants this.
- AI fluency: Wide variance. Shlanda builds Claude skills; others are at different levels.
- Adoption posture: Protective. Team members worry that automating grunt work means losing a dedicated contractor they could offload to.
- Freed capacity: ~50 hours/month already freed by existing automations, but no plan for where those hours go.
- Sponsor readiness: Strong. Shlanda is the translator between the team and the technology.
Technology
- 9+ systems in onboarding alone with near-zero automation between them
- Greenhouse-to-Paylocity overnight feed expires August 2026 (the single most critical infrastructure deadline)
- Claude is on individual desktops but not embedded in workflows
- Shadow tooling (Shlanda's personal Zapier, manual spreadsheets) absorbs the gaps
- HR inbox has no routing layer (13,000 emails/year manually sorted)
Data
- Paylocity is the canonical employee record after the onboarding-to-HR conversion
- Inter-system feeds are fragile and undocumented (Paylocity-to-Lattice breaks silently)
- HRIS data quality is poor: 4 identity collisions (wrong people merged), 12 unreadable department labels, 9 missing titles, ~50 missing managers
- Ambiguity about whether baseline CSV data is real or sample data
Organization
- Shlanda is the single point of failure. She holds the entire People function in her head: AI transformation, governance, fluency, VP translation, and workflow context.
- Syra owns onboarding but was hired ~3 weeks before the audit with no documented transfer criteria
- Nickole is an AI-Ops candidate but her mandate is undefined
- Mandates are bundled by role, not by workflow (so one person leaving creates gaps across multiple workflows)
Optimization potential
| Scope | Low | Mid | High |
|---|---|---|---|
| Onboarding alone (HCB + HQ) | $37K/yr | $78K/yr | $134K/yr |
| All four critical workflows (extrapolated) | $120K/yr | $250K/yr | $420K/yr |
| August API break (unmitigated cost) | +$9K/yr labor | + compliance exposure |
Prioritized gaps (top 11)
Listed in recommended order of attack:
| # | Gap | Effort | Why it matters |
|---|---|---|---|
| 1 | Greenhouse-to-Paylocity API rebuild | 6-12 weeks (FDE-heavy) | August 2026 hard deadline. Class 3 infrastructure. Blocks everything. |
| 2 | Agent-addressable employee record | Inside #1 | Unblocks every Class 2 automation |
| 3 | Owner-validated Class 1 workflow maps + second-seat coverage | 2-3 weeks each | Closes single-owner risk (if Syra is out, who runs onboarding?) |
| 4 | Plan-selection automation | 2-4 weeks | Closes coordination tax per hire |
| 5 | Drive folder auto-create | 1-2 weeks | Replaces Shlanda's broken personal Zapier |
| 6 | Light diagnoses on offboarding / role change / comp change | 2-3 weeks each | Calibrates Q3 scope |
| 7 | Fluency curriculum + freed-capacity absorption model | 4-8 weeks | Ensures time savings actually compound |
| 8 | HRIS data quality + real-vs-sample CSV resolution | 2-4 weeks | Can't automate on top of bad data |
| 9 | HR inbox routing automation | 4-8 weeks | 13K emails/year; pricing decision pending |
| 10 | AI-Ops role formalization + governance ratification | Inside engagement | Gives Nickole a mandate |
| 11 | Confidential-hire policy-aware routing | 4-8 weeks | Requires #1-4 first |
Open questions from the audit
These were flagged and may or may not be resolved by now:
- Are the four workflows (onboarding, offboarding, role change, comp change) the right four?
- Does the HR inbox bot belong in Season 1?
- What is trailing-12 HCB hiring volume by month? (Largest sensitivity on the optimization math)
- Is the Atlas employee CSV real or sample data?
- Who is the second seat per workflow? (If the primary owner is out, who covers?)
- Where are the ~50 freed hours/month going?
- Is the August API renewal cost a sunk cost or a budget line?
- What is Nickole's formal AI-Ops mandate?
- Which sensitive flows carry the same policy-without-system pattern?
Why this matters for your role
This analysis is why you're embedding instead of MR hiring an administrator. The diagnosis showed that the team's problems aren't solved by adding another person to do manual work. They're solved by making the workflows legible (Class 1) and then automated (Class 2). Your Month 1 (doing the work manually while documenting every step) is the Class 0.5 to Class 1 transition. Your Month 2 (building agents) is Class 1 to Class 2.