From 0.5% to 2.3% conversion. Ten weeks.

The setup.

In 2024, a Y Combinator-backed career-services company was placing roughly half a percent of applicants into client interviews. It had a product, a team, and capital. None of those were the problem.

The company was placing approximately 0.5% of applicants into client interviews. The product worked when it worked, but only about 1 in 200 applicants reached the placement stage at all.

The leadership had assumed the conversion problem was a product problem, and had been iterating on the product without moving the metric. The hypothesis was that the next product iteration would unlock conversion. It would not.

The conversion problem was an operating-model problem. The brief was explicit: the issue was not a lack of human effort — it was a fragmented system causing massive inefficiency. The engagement was as Lead Systems Architect — to re-engineer the operations stack, not to advise on it.

What the operating-model audit revealed.

The audit walked the actual workflow from applicant arrival through interview placement, including the exceptions and the unwritten rules. The delta showed up in four specific layers.

Input quality was unstratified. Applicants who were not a fit for the program were accepted at the same priority as fit applicants. Coaches spent time on conversion paths that could not convert.

Decision protocols were unwritten. When to escalate, refer out, or decline was made on instinct. The same applicant profile got different decisions from different coaches.

Automation handoffs were inverted. Routine work that should have been systematized was being done by humans; judgment work that needed human attention was being templated.

Incentive alignment was weak. Coaches were compensated primarily on hours worked. Conversion to interview was not in the compensation structure. The system paid coaches to spend time, not to convert.

None of these were product flaws. All of them were operating-model flaws.

The intervention.

Ten weeks. Four operating-model layers rebuilt, with AI-driven workflow automation (Zapier + ChatGPT) stripping out the manual drag so human judgment was applied where it actually mattered.

Input quality. Applicants were stratified by program-fit criteria rather than worked at uniform priority, so coaching capacity stopped being spent on conversion paths that could not convert.

Decision protocols. The highest-leverage coach decisions — intake, mid-coaching strategy shift, escalation, decline — were codified into decision frameworks instead of left to per-coach instinct, so the same applicant profile got the same decision.

Automation handoffs. Scheduling, status communication, and initial screening were moved off humans and onto AI-driven workflow automation, eliminating 28+ hours of manual work per month and freeing coach time for the judgment work that needed humans.

Incentive alignment. Coach compensation was restructured so conversion to interview was the primary variable rather than hours worked — the performance and compensation system was rebuilt so the incentive matched the outcome.

By the end of the ten weeks, application-to-interview conversion had moved from 0.5% to 2.3% and held there. No new product. No new team. No new capital. Weekly operational cost fell 21%.

The outcome.

4.6× lift in the company's single most important outcome metric. Achieved through operating-model redesign and AI-driven workflow automation, not through technology investment or team expansion.

The conversion lift was the headline number. The verifiable downstream effects: 28+ hours of manual reporting per month eliminated, weekly operational cost down 21%, and a rebuilt performance and compensation system — the company's unit economics improved enough to support meaningful pricing and growth decisions over the following period.

How this maps to the Mezura journey.

The 2024 engagement is the prototype of the Mezura journey. Today the same work is delivered as a productized Chapter 1 diagnostic plus a scoped Chapter 2 rebuild — faster, because the analytical work is AI-augmented and the scope is productized; the same in substance, because the judgment work stays human.

In today's terms: a Chapter 1 (Operational Friction Diagnostic — the audit and pathology mapping that revealed the four broken layers) followed by a Chapter 2 (AI-Driven Operating Rebuild — the rebuild of input quality, decision protocols, automation handoffs, incentive alignment). What is different in 2026 is compression and packaging, not substance. The diagnostic is a fixed-scope, $25,000, 10-business-day product. The senior judgment work — audit interpretation, pathology mapping, intervention sequencing, decision-framework design — stays entirely with the senior operator. The 4.6× conversion lift specifically is the kind of metric that, in a Chapter 3 partnership, would be a contingent outcome — paid only on verified, durable improvement.

See a redacted one-page sample of today's diagnostic deliverable →