AI-assisted delivery and operational efficiency
Building a Customer Training Platform in Weeks Instead of Months
Illumitrac Training Hub
- Role
- Lead Product Manager / Hands-on Builder
- Organization
- Alatus Solutions (Illumitrac)
- Timeline
- 2026
- Status
- Launched
Visuals



Shared with approval. Customer and internal business data redacted.
Context
- I independently designed and built a customer-facing Training Hub in roughly 2–4 weeks using AI-assisted development, with no engineering involvement in the initial build. Customer Success now handles routine content updates the same day.
- Training content updates depended on engineering and took at least two sprints.
- That delayed customer-facing content and pulled engineering time away from product priorities.
What was needed
- Working with Customer Success, I mapped which updates were routine and which genuinely needed engineering.
- Most changes were content, not code. The bottleneck was ownership, not complexity.
- Rather than queue a large engineering project, I scoped a version I could build myself and that Customer Success could own.
- An earlier engineering estimate was about six months, but scope and approach differed, so I don't claim a precise speed multiplier.
What I did
- Built personally: design and initial build of the Training Hub using AI-assisted development.
- Owned: scoping, QA, and the content-ownership model.
- Team: Customer Success now manages routine content updates.
- A customer-facing training platform with content Customer Success can update directly.
- An AI assistant that answers customer questions and escalates to a person when needed.
- The assistant is named Toothy. It appears in the Training Hub and as a widget inside the illumitrac provider portal, where it acts as an AI growth coach for membership-club questions.
- The hub includes training videos, scripts, role guides, resources, a success guide and a value calculator.
What changed
Status: Launched
~2–4 wks
Initial build
Built independently, no engineering in initial build
Same day
Routine content updates
Previously at least two sprints
88%
Conversations without human escalation
367 of 417 engaged conversations; not verified resolution
More detail
AI approach & safeguards
- AI-assisted development let me move from requirements to a working product without an engineering queue.
- The assistant offers a path to human escalation. I report 'no escalation' as exactly that, not as confirmed resolution or cost savings.
- Privacy safeguard: the chat shows a legal disclaimer telling users not to enter patient-identifying information.
Tradeoffs
- Scoping: I kept the first version deliberately small so it could ship in weeks rather than wait on a fuller spec.
- QA: building alone meant I owned testing too, so I prioritized the paths customers use most.
- Content ownership: handing updates to Customer Success speeds things up but requires clear guidelines to keep quality consistent.
Lessons
- Some bottlenecks are about who owns the work, not how hard it is.
- Measuring AI assistants honestly matters: escalation rate is a useful signal, not proof of resolution.
Built at Alatus Solutions. Internal architecture, customer data, and confidential business details are intentionally omitted.