One AI Layer Behind WhatsApp and Voice
An AI layer that lets Fundly's fintech customers get answers and take actions over WhatsApp and voice — and frees relationship managers from routine conversations to focus on what actually needs a human.
A hardcoded bot that broke the moment customers went off-script
Fundly's engagement team handled a large volume of similar conversations — product questions, status checks, onboarding help, reminders and follow-ups. The first WhatsApp bot was hardcoded, with fixed menus, keyword rules and scripted replies. It broke when customers used their own words, mixed languages, or changed topic, and every new use case meant new code. Voice had no automation at all, so every call needed a person.
Want One AI Layer Behind Your Own WhatsApp and Voice Channels?
We'll scope a pilot around your exact customer conversations — product questions, status checks, onboarding or reminders — with clear success metrics before any production commitment.
An agentic conversation platform, in three parts
Agentic WhatsApp Bot
Refactored from a hardcoded, menu-driven bot into an agent that reads the message, plans the next step, uses tools to fetch or update information, and keeps context.
Voice Agent
Designed and built from scratch — a real-time agent that answers calls, understands speech, responds naturally, and completes the same tasks as the WhatsApp agent.
Shared AI Platform
Agent orchestration (fundly-ai-spine), an MCP data and tool server (fundly_mart_mcp), and the RM-facing AI service (fundly-ai-rm) — one layer behind every channel.
Problems it solves
Rigid conversations
The agent interprets free-form messages and decides the next step, instead of matching fixed keywords.
Slow to extend
New abilities are added as tools, not new hardcoded menu branches.
No voice automation
The voice agent handles calls end to end and hands off when needed.
Scattered data
One MCP layer gives consistent, governed access across systems.
RM time on routine work
Routine queries are automated; RMs get AI support where it matters.
Limited availability
Instant responses on WhatsApp and voice at any hour.
A typical conversation
- 1Customer reaches out on WhatsApp or by phone
- 2Agent understands intent and pulls relevant history
- 3It plans next steps and calls the right tools via MCP
- 4It replies as text or speech, grounded in systems of record
- 5Sensitive or unresolved cases go to an RM with full context
- 6Every conversation is logged so quality can be reviewed
Trust and safety
- Grounded answers from systems of record, not from the model's memory
- Human handover for anything sensitive, high-value, or unresolved
- Logged conversations and tool calls for review
- Access controlled by the MCP data layer
Status: the WhatsApp bot has been refactored into an agentic system, and the voice agent is built from scratch with core conversational capabilities. Use-case coverage, voice languages, usage numbers and roadmap are still being confirmed.
An engineering team that ships agentic systems, not slide decks
AWS Advanced Tier Partner
Certified cloud and AI delivery, so your agentic platform runs on infrastructure built to AWS Well-Architected standards from day one.
Pilot-first engagement
We scope, build and validate against real conversation data before you commit to a production rollout.
Grounded, access-controlled answers
Every answer is grounded in your systems of record through a governed MCP data layer, not the model's memory.
Human handover built in
Sensitive, high-value or unresolved conversations route to your RMs with full context and a logged audit trail.
Trusted by forward-thinking teams
AWS-Certified Experts Delivering Cloud Excellence.
More Fintech AI Solutions
Want One AI Layer Behind
Every Customer Channel?
Tell us where your WhatsApp or voice conversations break down — we'll help you scope an agentic AI platform built on your own data, with humans in the loop where it matters.