Skip to content
Roshil Shrestha
All projects

Proof of concept

Sales Copilot

Agents that create and look up HubSpot deals, a Neo4j graph of the CRM, and natural-language answers with next-step recommendations.

LLMs & agentsGraph ML & knowledge graphs

Overview

A proof-of-concept sales copilot built on HubSpot's APIs. Autonomous agents handle deal creation and lookups in real time, a Neo4j knowledge graph links the CRM's entities, and an LLM answers questions and recommends next steps from past deals, notes and emails.

The problem

Sales context is scattered across deals, companies, contacts, notes and emails in the CRM. Getting an answer or deciding the next step means clicking through records by hand.

What I did

  • Developed the copilot and explored HubSpot's endpoints for deals, companies, CRM data, pipelines and deal stages.
  • Built agents with CrewAI flows for real-time deal creation and information retrieval.
  • Constructed the Neo4j knowledge graph and the question-answering and recommendation layers.

How it works

  • A knowledge graph of deal stages, deals, companies, notes, contacts and emails for contextual querying.
  • Natural-language questions over CRM data, answered with ChatGoogleGenerativeAI.
  • A recommendation engine that uses historical deals, notes and emails to summarise deals and suggest next steps.
  • Prompt engineering to improve answers and graph-based query resolution.

Outcome

  • Natural-language access to CRM data, with deal summaries and next-step suggestions based on previous sales activity.