Case Study — Q3 2025

Designing a multi-agent AI system for customer support

Led the strategy and implementation of an agentic AI workflow for a B2B SaaS platform — replacing a legacy chatbot with four specialized AI agents that autonomously resolve 89% of support queries.

Role

AI Product Strategist

Client

B2B SaaS (Series C)

Duration

12 weeks

Team

5 people

Results after 60 days

75%
Cost Reduction
$12.40 → $3.10 per ticket
89%
Resolution Rate
Up from 62%
<8s
Response Time
Down from 4.2 min
4.6
CSAT Score
Up from 3.4 / 5

Interactive System Architecture

How the multi-agent routing works

Click a customer query to simulate routing

The Problem

A scaling trap hiding in plain sight

The client's users had grown 3× in 18 months, but the support team only grew 40%. Their rule-based chatbot handled just 18% of queries successfully — the rest escalated to human agents. First response time was 4.2 minutes, support cost had risen to $12.40 per ticket, and CSAT dropped from 4.1 to 3.4 in two quarters.

The knowledge base was fragmented across three platforms. The chatbot couldn't learn. The team was drowning. Something had to fundamentally change.

Strategic Thinking

Key product decisions I drove

Execution

12-week delivery timeline

Technology

Stack & tools