The situation
A tech-focused staffing agency serving Series B–D startups. Recruiters were sharp, the brand was strong, the inbound was healthy. But submissions per recruiter had been flat for three quarters and the firm was struggling to grow without adding head-count it couldn't easily afford.
When we audited the recruiter day, the breakdown was telling: 28% sourcing, 22% screening calls, 18% writing submission packages, 14% interview prep with clients, 18% admin. Three of those buckets — sourcing, package writing, and prep — were AI-yieldable.
What we built
- Sourcing copilot — given a client brief, drafts a search profile, queries the firm's database and external sources, and ranks candidates with explainable scores tied to the brief.
- Submission package generator — turns a screening conversation into a client-ready submission package, including the candidate write-up the firm is known for.
- Interview-kit assembler — produces role-specific interview kits with questions tuned to the firm's signal model, plus the client's hiring rubric.
How the work ran
The Intelligence sprint came first — a two-week workshop where we trained the leadership team and three senior recruiters. The goal was for them to be able to evaluate AI output and feed it back to the build team. This step matters more than people expect; without it, the AI's output quality plateaus at average instead of getting toward best.
Build was four weeks. Rollout was the harder part — getting 18 recruiters to change their daily workflow took 60 days, four lunch sessions, and a buddy system where high-adopters paired with skeptics. By day 90, the resistant 20% had moved.
We were ready for the tools to be fast. We weren't ready for the tools to be more careful. The submission packages got tighter, not sloppier. — VP, Talent Operations
Outcomes after one quarter
- Submissions per recruiter rose 2.3×.
- Time to first submit on a new role dropped from 5.4 days to 2.3 days.
- Client satisfaction scores rose modestly (4.2 → 4.6), driven by the tighter submission packages and faster turnaround.
What we'd say to a similar firm
The capacity is real, but you only get it if recruiters trust the output. Spend the Intelligence weeks teaching senior recruiters to evaluate AI work — they become the quality gate, the feedback loop, and the cultural champion. Skip that step and you get tools the team doesn't use.