Project
Application Screening Automation
Co-founder & Product Lead, Agenture
Built screening automation that read every application within minutes of arrival, scored it against defined criteria, and ranked candidates so the team only reviewed the strongest.
- Scoring ran on an explicit rubric with a per-criterion breakdown — hard rules for must-haves, LLM judgement for fit — and borderline cases routed to a human rather than auto-decided.
- Every applicant got a fast, personalised response, removing the silence that damages a business's reputation when applications go unacknowledged.
Overview
When applications arrive in volume — candidates, tenants, programme applicants — the bottleneck is reading and ranking them fast enough to act. Through Agenture I built screening automation that read, scored, and ranked every application within minutes, so the team's time went to the shortlist and no applicant was left waiting.
The problem
High-volume application processing is slow and inconsistent by hand. Applications pile up, the best ones get buried under the average, and response times stretch out — which costs the business the strongest applicants, who don't wait around. Manual scoring also drifts: criteria get applied differently depending on who's reading and how tired they are.
What I built
- Instant intake + scoring — every application was read within minutes of arrival and scored against a defined rubric.
- Explainable, rule-plus-LLM scoring — deterministic rules handled must-haves (eligibility, location, hard requirements); an LLM assessed fuzzy fit; the output was a ranked score with a per-criterion breakdown.
- Tiering + human review of the borderline — strong and weak extremes were auto-routed; anything near a threshold went to a person, so judgement was applied where it mattered.
- Personalised responses — every applicant received a prompt, personalised reply rather than silence.
Impact
- Cut application review from a backlog to a ranked shortlist available within minutes.
- Made screening consistent and explainable rather than dependent on who was reviewing.
- Protected the client's reputation by ensuring every applicant got a fast, human-quality response.