01 · Forge AI
Forge is a managed AI service for staffing firms. We run sourcing, outreach, screening, BD, and back-office workflows for you — outcome-based pricing, no software for your team to learn.
2–3 min
Recruiter time per candidate
Down from 45–60 minutes today. Sourcing, screen, and submit collapsed into one workflow.
8–12
BD meetings per rep / month
Vs. 2–4 today. Account mapping turns placement history into a prospecting weapon.
<8 hrs
Month-end close
From 3–5 business days. Timesheets, invoicing, payroll, and reconciliation, automated.
02 · Why now
Bill rates are flat. Candidate scarcity is real. Recruiter cost-per-placement keeps climbing. The firms that compound margin from here are the ones that take recruiter admin out of the workflow — not the ones that hire faster.
01
Boolean searching, formatting submissions, copy-pasting reqs across portals. The job is supposed to be talent — it's actually data entry.
02
Every new req re-runs the same search. Past placements, past submissions, past screens — none of it compounds into your next req.
03
Timesheet chasing, invoicing, payroll reconciliation, MSP remittance matching. 3–5 FTEs just to keep the books moving.
03 · Who builds this
Forge isn't a vendor selling staffing firms a tool. It's the AI we wished we had when we were running them. The team has built and sold staffing companies, sat in the recruiter chair, run the BD desk, closed the books at month-end. Every workflow we automate is one we've done by hand.
Founder & CEO
Built and sold Veritaaq to ManpowerGroup at $250M revenue, 2,500 consultants on assignment. One of the largest IT-staffing exits in Canadian history.
Operations
20+ years inside Canadian IT staffing. Day-to-day operator of the platform: sales, delivery, recruiter ops. Knows the workflow because he runs it.
Strategic Finance
Interim Regional CEO at Employment Hero Canada. Ex-VP Finance at Humi and Notch. Owns the back-office automation track end-to-end.
We don't pitch software you have to learn. We run the AI in your firm and you keep operating the way you operate.
04 · The Engine
We run the deployment. You see the outcomes. The platform is the engine underneath — your team never has to learn another piece of software.
01 / Recruiting
VMS polling, sourcing, outreach, AI phone screening, submission, and offer lifecycle — req to placement, end to end. Recruiter review time drops from 45–60 minutes to 2–3 minutes per candidate.
02 / Sales & BD
Account mapping turns your placement history into a prospecting weapon. Lead enrichment, multi-channel outreach, AI personalization. 8–12 meetings per rep per month, not 2–4.
03 / Finance
Timesheet validation, invoicing, payroll reconciliation, MSP remittance matching, month-end close. Compresses close from 3–5 days to under 8 hours.
05 · See it work
Three live diagnostics. Each one runs the same engine our managed service customers use, scoped down so you can try it now without a sales call.
Tier 1 · Free
Paste a resume and a job description. Get the cosine score, the 5-component breakdown, and a Claude-generated rationale in 30 seconds. No signup.
Try the widget →
Tier 2 · Free
One JD plus 10–50 resumes from your own pipeline. Forge ranks them with the production scoring model. Keep the shortlist.
Coming this week
Tier 3 · 30-day trial
OAuth your Bullhorn read-only. Forge embeds your candidates and runs the live sourcing agent on your real reqs for 30 days, free.
Request access →
Or take the 5-minute diagnostic — no integration, no signup
AI Readiness Scorecard →Forge Research · Featured
Most AI for staffing claims are inflated. A few are very real. Here is the honest read on where AI compounds inside a staffing firm in 2026, written for staffing firm owners by an operator who has actually done it. Recruiting, sales, finance, the data layer underneath, what to never let AI touch yet, and the five questions that separate signal from noise when you are buying.
16 min · By Jordan Huck
06 · The model
Resumes and reqs are messy prose. Keyword search misses real matches. Asking a model to read every resume against every new req would be the right answer — except it collapses unit economics at staffing scale. Forge runs a two-stage retriever that gets you LLM-quality matching at non-LLM cost.
01
Voyage voyage-3-large maps every resume and every req into the same coordinate space. Done once per resume, cached forever.
02
Cosine similarity narrows tens of thousands of candidates to the top 100 in milliseconds.
03
The reranker scores those 100 in pairs against the actual job and surfaces the real top 10.
04
Claude Sonnet generates per-candidate rationale only for the final 10, with prompt caching across the run.
This is the same two-stage retriever-plus-reranker pattern Cohere, Voyage, and Anthropic recommend for production. Sub-second retrieval on a 30k+ corpus.
07 · Talk to us
30 minutes. Bring an open req from your firm. We'll embed your candidates and run the live sourcing agent in front of you. No slides.