Estera

AI Receptionist that Answers Calls & WhatsApp 24/7

Productivity

Customer Success

Artificial Intelligence

Visit WebsiteSee on Product HuntInstagram ⧉LinkedIn ⧉Twitter ⧉Facebook ⧉

Timely for us — we run a healthcare directory where most patient–clinic contact happens over WhatsApp, and the hard part isn't answering, it's knowing when to hand off to a human. How does Estera decide when to escalate? Also curious how you handle regulated verticals where you can't store health details in the transcript.

The real estate use case is interesting — I manage rental properties, and the after-hours call problem is real. Tenants call at 10 PM about a leaking pipe, and you're choosing between answering every call or missing something urgent. How does Estera handle triage? Can it distinguish between "my faucet drips a little" and "water is coming through the ceiling" and route those differently? That escalation logic is where most AI receptionists fall flat.

Setup was genuinely fast, had my agent handling test calls within minutes. The natural-sounding responses caught me off guard, way better than the robotic assistants I've tried before.

It looks good. But I can imagine that Estera will need to handle some edge cases when it won't be able to answer.

Then what? Would it be able to create some kind of ticket, ready to be answered by a real person?

Congrats on the launch @fredyandrei @toma_rares ! I've tested it and it was very simple to setup and use. All the flows and visual look were great! Love it!

Estera the website is cool and easily understandable i like that and would love to test this agent how perfectly it can handle calls and disputes and how much data i have to give him and does it remember things and learns from the mistakes it did?

This is nice. How does it handle callers who specifically demand to speak to a human?

the not letting the agent commit when it's unsure approach is a smart way to frame the risk. curious about the other trigger for handoff though - what if the customer just isn't happy talking to a bot and asks for a real person directly, nothing to do with pricing or a gray area. does Estera hand off immediately on that kind of request, or does it try to keep resolving things itself for a bit first? for hospitality especially that could matter as much as the pricing accuracy does.

I build on the support side so take the bias into account. Calls and WhatsApp look like one inbox and they are not, and the follow-up promise is where that shows.

On a call the conversation has a beginning and an end. On WhatsApp it has neither. The customer replies four hours later in the same thread expecting you to remember, so it is really a record that never closes. That is a different data model to a call transcript, not a variation on one.

The specific thing I would check is the follow-up. WhatsApp only allows free-form messages inside the service window after the customer's last message. Outside it you are on approved templates. So automatic follow-up either lands inside that window, or it goes out as a template and reads like one, which for a dental clinic chasing a booking is the wrong tone.

The other one is handover. Estera answers at 2am, a human picks the thread up at 9am, and the customer has no idea a shift changed. Whatever was promised overnight has to be sitting in front of that human, or they contradict their own receptionist and the customer just sees a business changing its mind.

Neither is a reason not to build it. They are the two places where one feature behaves differently per channel, and both are cheaper to design for now than to retrofit..

Fredy, your answer to Gal about the agent knowing when not to decide is the strongest thing in this thread, so I will push on the vertical list instead.

Clinics do not behave like salons or restaurants. A patient calling to book usually says why, and that reason becomes regulated health data the moment it lands in a transcript, a CRM field or a follow up message. In the US that also rules out the channel you lead with, because Meta does not sign a business associate agreement for WhatsApp, so a clinic running patient conversations there is exposed no matter how careful your agent is.

Is a clinic a separate configuration for you, or the same agent with a different script?

Congrats on the launch, Fredy!

I saw you how you were sharing updates the past few months and I respect the consistency.

In a world where businesses can improve their services with products like Estera, it's a no brainer.

The live pricing and human handoff approach is practical. How does Estera prevent double-bookings when two customers try to claim the same slot at nearly the same time across calls and WhatsApp?

the trust question that'd matter most to me isn't accents or background noise, it's what happens when Estera gets something wrong mid-booking - quotes a price that's since changed, or double-books a slot your calendar already had held. does the business get a heads up to catch that before the customer shows up, or is the first anyone hears of it when the customer's standing at the counter confused? for something handling money and scheduling unsupervised, the failure path matters more than the happy path demo

The under-5-second answer time is the spec I'd zero in on, because with voice the delay before the first word matters more to callers than the raw quality of the response. I build voice AI that calls aging parents daily, and getting reliable pickup plus natural turn-taking over real phone lines turned out to be harder than the model work itself. How are you handling barge-in when a caller talks over the agent mid-sentence? That one detail tends to decide whether a voice agent feels human or not.