Case study · HR & People
They stopped renting their own workforce. ~$2.4M a year came back.
A good manufacturer grew into three plants. The people machine didn’t. Two recruiters carried ~250 hires a year. Roles sat open 44 days. And while they sat, the plants covered the empty seats the expensive way: overtime, temps at a 35 to 45% markup, and a contingency recruiter for anything hard.
Nobody decided to rent the workforce. A machinist search went outside once, then every hard search did, and the invoices stacked up so quietly nobody called it a strategy. By the audit, the agency and temp line had nearly doubled in two years.
The antagonist · Agency and temp spend, 8 quarters
From ~$480K to ~$920K a year in eight quarters, one hard vacancy at a time. Hold this line. It comes back at chapter 06.
A Practical Intelligence AI strategist sat with the two recruiters and two plant managers for four weeks and traced hires from “we need someone” to first shift. Four findings made the file.
A new role opens in email and dies in a template.
The req starts in a hallway or an inbox. The JD gets rebuilt from an old posting, then waits through two approval loops. Three weeks before anyone outside the company could even apply.
Every hard vacancy becomes an agency invoice.
Machinists, maintenance techs, a controller: any role that aged past a month went to a contingency recruiter at ~$18K a placement, while the plants bridged with temps at a 35 to 45% markup. ~$900K a year, renting people the company’s own name could attract.
Week one is sink or swim.
New hires arrived to no badge, no uniform, no buddy, and a paper packet. The early exits quietly restarted the whole search, agency invoice and all.
Nine people answering eleven hundred phones.
PTO, pay stubs, benefits, policy: the same questions all day, by phone and walk-up, for a workforce with no company email. A fifth of them land after hours, when nobody is there to answer.
You don’t fix a people shortage by renting people.
You give the people team more hands.
An alignment layer over the people stack,
and three new teammates on top.
THEIR EXISTING STACK · CONNECTED, NOT REPLACED
A manager asks for a role in Teams. The recruiting agent drafts the JD from the company’s own templates and pay bands, routes it to HR for sign-off, and has it live on Indeed and LinkedIn the same day.
The three dark weeks became an afternoon. The 44-day clock now starts the day the manager asks.
Every one of 212 applicants gets an answer in minutes. The agent screens on availability, eligibility, and certifications, ranks the shortlist with the reason in plain English, and books interviews straight onto the hiring manager’s calendar.
Recruiters stopped digging through applications and started closing the good ones. 4 of 5 shortlisted candidates advance past the manager screen, and the agency calls mostly stopped: agency and temp spend fell ~70%.
The moment an offer is accepted, the onboarding agent runs the machine nobody had time to run: paperwork, background check, badge, uniform, locker, buddy, first-day schedule by text. Then it stays with the hire through the ramp, answering questions and handing off to the front desk’s check-ins.
Nobody walks into a plant unannounced anymore. First-90-day exits fell by a third, and the check-ins now keep machinists and techs from walking in year one: 4 to 6 saves a year, at 50 to 200% of salary to replace.
“How much PTO do I have before the 4th?” Any employee texts the front desk and gets their own answer in seconds, pulled live from their UKG record: balance, pay stub, schedule, benefits. No portal, no password, no waiting for morning. The same number takes a supervisor’s incident report at 6 AM and flags the punch that doesn’t add up.
78% of tickets never reach a human, around the clock, and the rulebook runs on the same thread: incident intake the minute it happens and certs kept current (the comp premium's mod rate started moving) · payroll anomalies caught at the source · at open enrollment it walks each employee to the plan that fits. HR's hours go to the conversations that need one, and sensitive matters route straight to a person, every time.
Smarter every week
Every shortlist the recruiters correct and every answer HR refines teaches all three teammates at once. The intelligence stays theirs.
~$2.4M a year at run-rate, most of it on invoices and payroll records: ~$1.41M in agency fees, temp markups, overtime, comp premium, payroll leakage, and retired tools. The rest is staffed days, kept machinists, deflected tickets, benefits fit, and avoided re-hires, modeled and held conservative.
That’s the rented-workforce line alone. Below it: the overtime, the comp premium, and the payroll leakage, all off invoices and payroll records, then the modeled lines, held conservative. Run it against your own agency bill. The number only goes up.
Cumulative value · first 24 months
Orange: invoice-backed (the agency line, overtime, comp premium, payroll leakage, retired tools), 59% of the number. Periwinkle: staffed days, kept skilled roles, deflected tickets, benefits fit, and avoided re-hires, modeled. It paid for itself inside year one.
The antagonist, ended · agency and temp spend
Chapter 01's line, continued. Two years of climbing, undone in under a year, down to the searches worth paying for.
Req to live posting
from ~3 weeks
Time to fill
Cost per hire
Shortlist quality
Interview no-shows
How we count it. The agency, overtime, comp-premium, and payroll lines come straight off invoices and payroll records. The staffed-day line is held at $300 a day against benchmarks of $450 and up; the retention, ticket, and benefits lines are modeled with their own rates. Not counted: faster ramp to productivity, the HR hire they didn’t make, avoided penalties. A human signed every hiring decision. Year one is discounted for ramp, and we show the full model and expect you to stress-test it with your own agency bill.
Close of file Nº 004
Open a file on your people operation.
The Audit is fixed-scope and fixed-fee. We come to you, sit with your recruiters and your plant managers, and leave you a plan to put AI into production against the number your board already watches.