OpenRouter says its AI sales agent gives a five-person team 600 hours back a month

OpenRouter says its AI sales agent, Rasp, returns about 600 hours a month to the company’s five-person sales team, while handling much of the work around inbound leads, call preparation and CRM updates. The case study offers a detailed look at how the company says it built the agent, and where it still keeps people in charge.

OpenRouter Watch analysis

What happened

Rasp researches and qualifies inbound leads, sends first-touch emails, prepares call briefs, drafts notes from transcripts and fills in most CRM fields. OpenRouter says it sends 93 per cent of first-touch emails without help, and that many leads now receive a response within a minute. The company says preparation, notes and CRM work for a demo fell from about 103 minutes to 44.

OpenRouter reports that Rasp costs about $30 a day, compared with nearly $800 a day for two earlier agents, Ace and Dove. It says roughly $18 of Rasp’s daily cost is inference on GLM 5.2, with the rest attributed to engineering time. The company says its routing system can move routine work to cheaper models while maintaining quality, and estimates that one listed alternative could reduce inference costs to about 50 cents a day.

Why it matters

This is a concrete account of an AI agent being used inside a sales team, rather than a demo doing the rounds in a glass box. OpenRouter says each representative can take about two more calls a day, with inbound triage using about a quarter of one person’s time instead of occupying a whole person.

The larger business figures need more care. OpenRouter reports a 2.6-times increase in close rates and a 34 per cent shorter deal cycle, but says pricing changes and market conditions shifted during the same period. It does not attribute all of those gains to Rasp. The figures are the company’s account of its own deployment, not an independent comparison.

Our read

The most useful part is the division of labour: Rasp drafts, sorts, flags and routes, while people run calls, negotiate and approve sensitive actions. That is a much more grounded picture than “the agent replaces the sales team”, and it gives other teams specifics to examine: measure the time saved, keep an eye on quality, and make human approval explicit where the stakes rise.

OpenRouter’s case study makes a persuasive case for testing this sort of workflow. It does not settle whether the same savings, costs or sales results would travel to another business. Even a very efficient agent cannot turn one company’s before-and-after story into a universal price list.

What to watch

• Whether the reported time savings hold as Rasp takes on more tasks.
• How OpenRouter measures quality when routing routine work to cheaper models.
• Whether other teams can reproduce the savings without the same sales process or engineering support.

Discussion spark: Would you trust an AI agent to send most first-touch sales emails if people still approve sensitive cases, or should every message get a human check?

not affiliated with or endorsed by OpenRouter