Case studies
Outcomes, not demos.
Two of these are AI employees running real work in production. The third is a deployment we designed and have not built, and it says so on the card and on its own page. No client names. Categories only.
Every number carries its source. Measured comes from the running system. Modeled is calculated from a measured figure at the volumes the client stated. A design target has not happened at all.
HealthTech
HealthTech company cuts a 90-minute meal plan to 15 minutes of review and opens 2 new revenue lines with AI blood test analysis, nutrition planning, and CRM
A 3-coach health company was capped by manual meal plans, blood test analysis, and CRM admin. Three integrated AI agents cut a 90-minute meal plan to 10 to 15 minutes of review and turned blood test reports and meal plans into two new billable service lines. The 9.1 hours a day is modeled from that saving at the client's volume.
Legal Advisory
Legal advisory firm sources 77 qualified leads in 12 days and delivers call packs in 5 to 10 minutes with an AI sales intelligence agent
A firm closing $25K-75K engagements was bottlenecked on manual lead research. One AI agent now scans federal filings every 5 minutes and delivers complete call packs to Slack within 10 minutes. It produced 77 qualified leads in its first 12 days; the 3 to 5x lead multiplier is modeled from the firm's own numbers.
Food Manufacturing
Proposed design, not builtProposed design: an AI security camera monitoring agent targeting 75% lower guard cost and 100% facility coverage at a multinational food manufacturer
A food production facility runs 20 patrolling guards with blind spots between beats. This proposed design would connect the facility's existing 50 to 100 cameras to an AI platform that watches every feed continuously across five detection capabilities, targeting 75% fewer guard positions with 100% coverage. It has not been built.
