How I Built Agents to Find Business Opportunities
I showed how AI can help a sales team move from company research to qualified opportunities and draft sales material, with people reviewing the work before acting on it.
Techies · Singapore · 21 May 2026
Business Opportunity Agents
Finding an opportunity is only the first step. The team also needs to know why a company is a fit and what to say to it. I built a workflow that gathers company evidence, assesses it against the target profile, and carries that research into draft sales material.
At Techies' Roast My Tech Stack, I explained how those stages work together rather than treating research and sales writing as separate tasks. The evidence and qualification stay with the draft so a reviewer can check the reasoning before deciding whether to pursue the opportunity.
From Research to Sales Preparation
Scoped Agents and a Shared Record
CrewAI coordinated research, qualification, and draft preparation as scoped agent roles. Company evidence carried into the assessment of fit, then into draft collateral. Keeping those stages on a shared record let the reviewer inspect why an opportunity was selected and what supported the proposed sales material.
Dashboard, State, and Deployment
Streamlit provided the operational dashboard, while Postgres retained workflow state. Docker packaged the application for Lightsail deployment. The team reviews the agents' prepared output before deciding whether to pursue an opportunity.
Access Controls Around the Workflow
Salted password hashing protected stored credentials, and parameterized SQL kept query inputs separate from SQL instructions. Scoped environment secrets and limits on tool access constrained what each part of the application could reach. I also covered activity logging and private cloud inference as distinct controls over the record of activity and where model processing occurred.