I’m Andy, a 19-year-old founder building CSflow at Crest Systems
Hey everyone, I’m Andy Salvo, cofounder of Crest Systems.
My cofounder Jameson is 20. He goes to NC State. I go to Penn State. We both study business. We’re building CSflow, an AI phone agent for trades and home-service businesses. It answers calls 24/7, qualifies the caller, books appointments, sends confirmations, and returns the recording, transcript, and a clear record of everything it did.
The idea came from a simple problem: when a contractor misses a call, that customer often calls the next company. We think voice AI can solve that, but only if business owners can actually trust it. That is why we built a live “glass box” demo where anyone can call CSflow and watch the interaction happen in real time.
I focus on product, design, customer discovery, and go-to-market. Jameson leads engineering and systems implementation. We have built the product lean and are now working toward our first paid deployments.
I joined Product Hunt to learn from people who have launched and sold real products, meet other builders, and get honest feedback as we bring CSflow to market.
Glad to be here — especially interested in meeting people working in voice AI, vertical SaaS, home services, and early-stage distribution.
Replies
welcome. also in voice AI, different domain (health checkins, on-device) but same trust problem. the glass-box demo is a smart call - people don't trust a voice agent until they can watch it not screw up. does the transcript/recording get shown to the caller too, or just the business owner reviewing after the fact?
@sabber_ahamed thanks! and honestly you said it better than our landing page does. nobody trusts a voice agent until they can watch it not screw up, so we tried to make watching it the product instead of a buried log.
to your question: right now the transcript + recording + the step-by-step of what the agent actually did is the business owner's audit trail — they can see exactly what was said and done on every call, after the fact.
the caller-facing side is the upfront disclosure (it tells them it's AI and recorded, and to just say so or hang up if they're not okay with it), plus in the public demo you can literally watch your own call fill in live in the glass box. we don't hand the caller a post-call transcript today — they just get the appointment confirmation text.
your domain flips it though: with health check-ins the caller is the person who has to trust and review, so caller-facing transparency probably matters way more there than in a booking flow.
how are you thinking about the on-device angle for that — is keeping it off the cloud the trust play?
Welcome, Andy. I run a care home alongside two other businesses, and missed calls are a genuinely scary problem in that world, a missed call can mean a family goes without care coordination for a day, not just a lost sale. The trust point you raised is the real bottleneck: our staff wouldn't let an AI near client calls unless they could see exactly what it said and did afterward, which sounds like exactly what your glass box demo is solving for.
Curious how you're thinking about the edge cases, an upset caller, a genuine emergency, something outside the qualifying script. That's usually where trust in AI agents actually gets tested, not in the smooth calls.