Full Machine Leverage, Full Human Authority
What building an autonomous recruiting system taught us about judgment, and why the system was the easy part.
July 24, 2026 | Optimal Tech Partners Insights
The thesis
We run a recruiting practice where machines do nearly all of the work and humans hold all of the authority. Every profile discovered, scored, evaluated, and drafted against is machine labor. Every message that leaves the building is released by a human who read it, as a structural fact: the system is physically incapable of sending anything a human hasn't approved.
Six doctrines, each paid for with a specific failure
Deterministic detection, model writing: code decides when something should be said, models decide how to say it. Specs mandate the failures: every quality review examines the instructions, not just the outputs. Processes lie, artifacts don't: monitor conversions between stages, not heartbeats. Silence never becomes a confident negative: a legacy filter once silently vetoed 1,400 candidates the reasoning layer had already passed. Judgment gets case law, not sliders: disagreements become written rulings. Context absent at decision time doesn't exist.
What the data taught us
Reply rates track role scarcity: niche roles pull 15 to 22 percent replies, saturated role types 4 to 9 percent. Validated employment history beats self-reported skills. The pool is over 105,000 profiles, and it grows with every search we run, compounding at the client-relationship level. Deliverability is content engineering.
What we refuse to automate
Nothing earns the right to send without a track record a human has reviewed; today 100 percent of outbound crosses a human queue, read for real. We are working toward autonomy deliberately: a message class earns send authority only at a demonstrated error rate of 0.1 percent or better, granted narrowly, watched continuously, and revocable. Autonomy is a performance record, never a default. Ambiguity goes to a human, and role definitions are human-owned inputs the machine reads and never writes.
About this paper
This is the operating doctrine behind Optimal Tech Partners' AI recruiting practice: an autonomous, agentic sourcing system built for talent engineering, paired with full human authority over every message sent. It is how a technical recruiting firm runs machine leverage without surrendering judgment.
Published by Optimal Tech Partners. Read more insights at https://optimal.tech/blog