Q11Leadership

Q11. Tell me about an AI system you built and the value it created

Two AI systems I architected end to end: ResolveOps, an AI billing-support agent with cited investigations and human-approved refunds, and the Actionable HQ commitment engine with WhatsApp follow-ups for 10,000+ users.

Rajendra Kumar · · 2 min read

Q11. Tell me about an AI system you built and the value it created

For this question I tell two short stories: one about AI that must be safe with money, and one about AI that changes user behaviour at scale. Each takes about a minute.

Two loops. ResolveOps: billing request to LangGraph agent to Stripe and policy hybrid RAG, producing a cited proposal that a manager approves before it executes and is verified. Actionable HQ: coaching chat to LLM commitment extraction to stored commitments, a scheduler, a WhatsApp nudge and user replies back in the thread.
Architecture · Both systems close a loop: one with a human approval, one with a follow-up where users already are.

Actionable HQ: turning conversations into follow-through

Actionable HQ is a behaviour-change SaaS with over 10,000 registered users. People had good coaching conversations, but follow-through dropped because commitments got lost in the chat.

I built the OpenAI-powered commitment engine that turns goals in the conversation into structured action plans with milestones, along with threaded discussions and WhatsApp Business API workflows. Follow-ups go out on WhatsApp, where people actually reply, and the replies come back into the thread.

ResolveOps: AI that is careful with money

ResolveOps is an AI operations platform for payments and billing teams. A generic chatbot that might hand out unauthorised refunds is a liability, so I designed it differently:

  • A LangGraph agent plans permission-checked lookups across the customer's invoices and payments in Stripe.
  • It checks the company's refund policy with hybrid search on pgvector and writes up its reasoning with citations.
  • It queues the refund or cancellation for a manager to approve with one click, bound to the exact parameters.
  • After execution it reads the record back to confirm, and every step is in a redacted audit trail.

Key takeaways

  1. 01Lead with the business problem, then the architecture.
  2. 02Show one safety story and one scale story.
  3. 03Keep each story to about a minute.
Case studyResolveOpsActionable HQLangGraphWhatsApp APIInterview prep