We talk a lot about what voice agents say during a call. I think the post-call step is where a lot of the business value is. What I'd want after every call: - a short summary (who called, what they wanted, what was promised) saved to the CRM - a text to the caller confirming anything that was booked or promised - an alert to a person only if action is needed - the call tagged by outcome, so the owner can see a weekly count of booked, answered and needs callback Without that, the agent handles the call, but the business still has to listen to recordings to know what happened. I'm curious how others are handling this, especially the line between "log it" and "alert someone". What does your agent do in the minute after a call ends?

When I build an agent that talks to customers, I start with what it should say when it doesn't know something. Then I write the rest. It sounds backwards, but it forces the important decisions early: - What topics is it allowed to answer? (hours, prices, booking) - What does it do with everything else? ("I'll have someone call you back today. What's the best number?") - How does it avoid making things up? Prices and availability come from a lookup, never from the model's memory. Most embarrassing voice agent moments come from an agent that tried to be helpful outside its lane: a guessed price, an invented opening time, a promise the business can't keep. A short, honest "I don't know, but here's what happens next" builds more trust than a clever answer. What's your agent's fallback line when it can't answer?

Voice agents that work perfectly in a quiet room can fall apart on a real call. Before going live with a client, I'd test with the conditions their customers actually call from: - speakerphone in a car - background noise: a kitchen, a construction site, a TV - strong accents and fast talkers - someone who interrupts halfway through the agent's sentence - long pauses while the caller looks for a reference number Also test spelling: names, email addresses, street names. That's where most transcription errors hurt, and the agent should read them back to confirm. It's an hour of slightly awkward phone calls. It's also the hour that tells you if the agent is ready, or just ready for a demo. What's the real-world condition that broke your voice agent the most?

When evaluating OpenClaw, in a fictional call, a second voice mentions a different date while the caller is answering. The decision I would test: ask the caller to confirm the intended date before changing the appointment request. How would you distinguish the caller from background conversation?

For this OpenClaw example, sample input: A sample caller pauses after the street name before giving the unit number. Confirm the complete address before looking up a route. Which address component does voice intake most often miss?