← Selected Work
Financial Services · USAFinancial ServicesBuild · MVP2024

AI knowledge retrieval for a customer support centre.

An AI assistant that puts the right answer in front of every support agent — plus a chatbot for the rest — built on retrieval over the institution's own knowledge.

Support agent assistant
Client
Public · $3.84B rev · 6k staff
Duration
3 months
Scope
MVP
Stack
OpenAI · Pinecone
Architecture
RAG
Region
USA
The challenge

Support agents at a large financial-services firm were searching fragmented knowledge bases to answer customers — slow, inconsistent, and hard to scale across a busy contact centre.

Our approach

We built AI-assisted knowledge retrieval: a vector index over the firm's knowledge, an agent-assist layer that surfaces answers to the human agent in real time, and a customer-facing chatbot for common questions.

What we built

Customer-support agent assistant — automatically provides the right information to the agent as a conversation unfolds. AI chat-bot — handles common customer queries directly, with retrieval grounded in approved knowledge.

Tech stack
Retrieval
OpenAIPinecone (vector DB)RAG pipeline
Capabilities
Agent-assist surfacingCustomer chatbotKnowledge ingestion
Application
Agent console integration
What we delivered
  • A retrieval-grounded agent assistant
  • A customer-facing AI chatbot
  • A knowledge-ingestion & indexing pipeline (Pinecone)
  • Delivered to MVP in 3 months
Timeline
Weeks 1–2
Knowledge audit & indexing
3 months
Build to MVP — assist + chatbot
At a glance
2
Surfaces: agent-assist + chatbot
3mo
To MVP
1
Grounded knowledge index
Outcomes
Faster
Answers, surfaced in real time
1
Source of approved knowledge
3mo
MVP delivered
Next case →AI claims & underwriting assistants for a global insurer.

Have something
that needs to ship?

Tell us the problem. We'll tell you honestly what makes sense — even if that's not us.

Start a conversation