01— AI · Insurtech · Conversational commerce
AI Sales Co-Pilot
WhatsApp-native conversational commerce for Indonesian insurance agents
Period2025 — 2026
RoleLead Product Strategy & Design
MarketIndonesia
StatusLaunched · Three production flows
- Team
- Product, engineering, and insurance operations partners
- My scope
- Research direction, product definition, conversational UX, system model, and production prototypes
- Constraints
- Remote access to Indonesian agents, Meta messaging rules, and high-stakes insurance steps
Overview
Agents sell on private WhatsApp; the product had to meet them there. I led research, product definition, and design for a WhatsApp AI Co-Pilot that amplifies — not replaces — their existing workflow.
The problem
Quoting lived in the Ignite app; trust lived in private chat. AI only inside the app added another switch when momentum was already fragile.
My approach
I defined one agent-initiated journey and a phased release — hand-off, quote, close, and signals — then translated it into conversational flows, a four-layer system model, and production prototypes with the delivery team.
Contribution
Research · Product · Design
Value to agents
An always-on sales co-pilot, embedded where agents already sell.
Agents run the full quote-to-close inside WhatsApp — the app their customers already trust — backed by a 24/7 AI assistant. No switching to Ignite, no leaving the chat.
- 24/7 assistant
- Native to WhatsApp
- Zero context-switch
Three roles, one thread from decision to delivery
I owned the WhatsApp Business Account (WABA) experience design, starting from the real working contexts of agents and sales teams before moving into product definition, core flow design, and delivery.
01
Research lead
Built the evidence base remotely through Meta documentation, benchmark flows, a live WABA environment, and the Indonesian agent workflow.
02
Product strategist
Defined the agent-led sales journey and four-phase MVP, clarifying where AI, WABA Flow, and human judgment each belong.
03
Experience designer
Translated the product logic into conversation scripts, WhatsApp Flows, and production prototypes used to deliver the launched experience.
Research
End-to-end WhatsApp sales
WhatsApp is not usable day-to-day on the mainland, so I could not shadow Indonesian agents in situ. I combined Meta documentation, benchmark services, a live WABA environment, and the existing agent workflow to test one question: which parts of quote-to-issue benefit from conversation, and which require structure?
The answer was a hybrid model. Free text supports flexible requests; WhatsApp Flows handle precise data capture; explicit agent confirmation protects high-stakes steps. The diagrams below show the evidence and interaction split behind that decision.
Product
Scope & phases
The product principle was empowerment without displacement: WABA provides tools, while the agent remains the visible owner of the customer relationship. That ruled out a fully autonomous bot and shaped the hand-off points in the flow below.
I divided delivery into four phases so the team could launch useful agent tasks first, then extend toward quoting, assisted closing, and engagement signals. The three product demos show the launched performance, marketing-asset, and quoting flows.
Launched flows · product demos
System design
Four-layer architecture
I translated the multi-channel entry points, AI orchestration, and insurance capabilities into four layers the team could reason about as one coherent agent experience.
Outcome
The WhatsApp AI Co-Pilot launched with production flows for performance inquiry, marketing-asset generation, and conversational quoting. The work turned a broad AI opportunity into an agent-led product model, a quote-to-issue journey, and a phased system the team can extend without displacing the agent-customer relationship.
Next case
02 · Ignite - Insurance Sales Platform
Reimagining an end-to-end insurance platform for Southeast Asia