Audit before I build anything.

Before writing a single piece of content, I'd run a real diagnostic. Pull Zendesk contact reason data, audit QA scorecards for calibration issues, map the current escalation path, meet with Horatio partner leads, and assess Rain's Notion documentation as the upstream content source.

Wide QA variance isn't a training problem until the measurement is confirmed consistent. High reopen rates aren't a content gap until I know whether it's a routing issue or a product depth issue. The deliverable at the end of month one is a findings report—not a course.

The highest impact fix first.

With the diagnostic complete, I'd start with the intervention that has the clearest data behind it. When empathy scores correlate with low CSAT, that's where I build first: a three-part sequence covering sentiment recognition, response selection with consequence, and a branching scenario that tests it all together. Fast to build, measurable signal within weeks.

In parallel, I'd begin the onboarding curriculum using Rain's Notion documentation as the content source, prioritizing the highest-volume ticket categories and clearest escalation risk. Everything built here is modular—labeled, categorized, and reusable across the full system.

An LMS that works like the rest of your stack.

Once the foundation is solid, the LMS becomes the delivery layer—and it should behave like one. Learning paths segmented by role, tier, and BPO partner. Certification assessments that are auto-scored and branch based on agent decisions. QA data flowing into agent health score views alongside learning completion data—no toggling between systems, no manual pulls.

The result is a live operational tool that tells managers who needs attention before it becomes a problem, and tells agents exactly where they stand at all times.

Extend the system. Don’t replicate the spreadsheet.

When the foundation is trusted and certification means something, BPO standardization actually works. Every Horatio agent moves through the same content and knowledge checks as a Rain employee. Certification criteria are system-enforced across every vendor. Product or policy changes push a required update module to every agent simultaneously—no separate coordination required.

Done in the right order, this is how a learning system scales without losing coherence or adding overhead every time the team grows.