Eric began with workshops rather than development — interviewing executives and operational leaders to understand what decisions they actually needed to make, rather than gathering technical requirements. This grounded every subsequent design choice in real decision-making patterns. The KPIs that mattered — production, write-offs, adjustments, appointments, and encounters — were normalized across all 15 clinic groups and 50+ practices with divergent naming conventions and system schemas. Data aggregation from 50+ clinic systems required significant integration work, with nightly automated refresh replacing weeks of manual report assembly. Wireframing sessions were used to design every report view from the perspective of how leaders actually read and acted on data — not how dashboards typically look. Once the descriptive analytics layer was stable and trusted, operational nudge tools were added: daily patient lists for clinical staff flagging missing x-rays before procedures, eligibility checks at intake to prevent unrecoverable billing write-offs, and performance alerts to site managers. AI was deliberately withheld until the data foundation was reliable — a sequencing choice that prevented the common failure of AI built on top of untrustworthy data.