Throxy: Vertical AI sales agent for manufacturing and distribution — the underserved B2B industries with high ACVs
Jun 11, 2025 with Pablo
Key Points
- Throxy reaches $1.5 million in annual revenue within six months of entering YC, the largest revenue figure disclosed at that point in Demo Day coverage.
- The AI sales agent targets manufacturing and distribution by qualifying prospects on equipment type, certifications, and budget using vision-enabled agents and proprietary data sources that horizontal SDR tools ignore.
- The four-person team keeps humans in the workflow for tasks where the model underperforms, treating the operation as continuous A/B testing rather than full automation.
Summary
Throxy is a vertical AI sales agent built for companies selling into manufacturing and distribution, industries that horizontal AI SDR tools have largely overlooked.
Generic outreach platforms fail when technical specificity matters. A company selling CNC software has no use in contacting manufacturers without CNC machines. Throxy's qualification layer targets exactly that kind of detail: specific machinery types, relevant certifications, sufficient budget, PE roll-up activity. The platform crawls proprietary data sources and uses vision-enabled agents to screenshot company websites and identify equipment on the floor, since mid-market manufacturing owners are largely absent from LinkedIn.
Traction
The company reached $1.5 million in annual revenue within six months of entering YC, the largest revenue figure disclosed across Demo Day coverage at that point. The round was already closed.
Human-in-the-loop model
With a team of four, Throxy deliberately keeps humans in the workflow for tasks where AI still underperforms. The founder frames this as ongoing A/B testing: automate what the model handles reliably, keep humans on what it does not, and stay accountable to clients on outcomes rather than delegating reputation to the agent.
Founding story
The two co-founders met in high school. One worked as an SDR across three companies doing painstaking manual research—checking hospital staff rosters, verifying certifications, assessing buying power—and saw the replacement coming. The other left an AI role at JP Morgan. The product attempts to automate the most painful parts of a job one of them already did.