A sales and ops agent over real company data.

Conversational access to Gong and HubSpot data.

The agent answers questions across sales calls, CRM records, contacts, deals, and operational context. Instead of making a team hunt through tools, transcripts, and dashboards, it gives them one place to ask the question and understand the answer.

The company name stays private, but the work is real: architecture, implementation, testing, deployment, and production use.

The hard part is not the chat box. It is the system around it.

Most demos make AI look easy because they skip the uncomfortable parts: messy data, tool permissions, ambiguous questions, unreliable outputs, and the difference between a good answer and an answer a team can trust.

This is where my background matters. I can talk to the business about the workflow and the risk, then translate that into a technical system that can actually run.

Start narrow. Make it useful. Then expand.

• Pick one high-value workflow instead of trying to "AI-enable" everything.
• Connect to the real source systems, not exported spreadsheets that go stale.
• Design for trust: show what data the answer came from and where the agent is uncertain.
• Test with the people who will actually use it, not just with perfect demo prompts.