A website chat agent booked 614 meetings and closed over $1 million

SaaStr, a well-known software events and media company, replaced the slow back-and-forth on its sponsor contact form with an AI agent. The company says the agent booked 614 meetings and sourced more than $1 million in closed revenue. In one month, 71% of closed sponsorship deals came from leads the agent qualified.

What changed

The old process was common. A visitor filled out a form, and it sat in a queue. A person passed it to a salesperson, who replied when they could. Then came emails to find a meeting time. All of that took two to three days on a good week.

Worse, after one SaaStr Annual event, an audit found about 1,000 people who had asked to sponsor and never got a human reply. The new agent, called Amelia AI, is built on a Qualified product named Piper. It answers questions, qualifies the visitor, and books the meeting.

Why it worked

SaaStr's Jason Lemkin argues this is an easy job for an agent. The visitor has already raised a hand. The tasks are clear: answer, qualify, route, book. The downside is small, because the alternative is a form sitting in a queue.

• Coverage: the agent handled 45,000 sessions, where a human chat team might handle 100 to 200.
• Context: it knows past attendance, earlier talks with colleagues, and what a company sponsored before.
• Routing: it sends deals to the rep who closes that type best, not just the next person in line.

Sales reps now get a full summary before the call, so they skip the basic discovery questions. Before the agent, inbound leads made up 29% to 34% of closed deals.

What to do

SaaStr says the agent is only as good as what it is fed. It is the most-trained agent in its setup, with one of the largest knowledge bases. Stale information leads to confident wrong answers.

• Write the agent's guidelines in plain human language. Instructions written by another AI performed worse.
• Give it the same onboarding material a new sales hire would get, including decks, PDFs and competitor notes.
• Connect it to customer records. Without them, it is just a chatbot.

What to keep in mind

These are SaaStr's own figures, and one company's results may not carry over. Not every booked meeting closed. The agent also needed tuning: at one point it favored one salesperson on certain accounts, and the team had to correct the weighting.

On cost, SaaStr says the agent costs less per year than one fully loaded sales development employee. It also says it could not staff this volume with people. It would take three, and they would quit every three months.

A related test used Salesforce's Agentforce on the roughly 1,000 ignored sponsor leads, some silent for six months or more. It drew a 72% open rate and a response rate above 10%, and closed deals.