What is your park’s phone saying when you’re not listening?

All three. SignalPony audits what was actually said on the call, regardless of who or what answered.

Yes, more than ever. An AI answering your phone doesn’t remove the need to check it. It raises the stakes. Even the best voice AI on the market isn’t good enough to run unsupervised. Sierra runs the industry test for real-world voice AI, called τ-voice. The top model today (xAI’s Grok Voice, as of April 2026) only gets it right 67% of the time. That means the best AI money can buy is still wrong on about 1 in 3 calls.

When AI gets something wrong, it doesn’t happen once. It repeats the same mistake, the same way, on every call it handles that day. Mis-qualifying a ready-to-rent lead. Quoting the wrong rent. Mishandling a financing question. None of these show up as an obvious error. The call just doesn’t convert. Multiply that across every call, and it adds up to real money, lost quietly.

It also raises your legal risk. Fair-housing and SAFE-Act problems used to be a training issue for one person. Put an AI on the phone, and any wrong phrasing — steering, a bad answer about kids or disability, financing framed as a loan — gets said confidently. At scale. On the record. Every time it comes up.

Here’s the part your AI vendor won’t tell you. The company that sold you the AI is the wrong one to grade it. They’ll happily send you a dashboard showing leads and bookings going up. But that’s a sales report, not an audit. No vendor is going to hand you a report that says “our AI mishandled 12% of your calls” or “our AI tripped a fair-housing line three times this month.” That’s not dishonesty. It’s just how incentives work.

The only way to know what your AI is actually doing is to listen to what it actually said. That’s what we do. Independently. Every month. On every call.

You could. But you’d have to pull every recording yourself, then upload and run it, every single time you wanted an answer. Do that week after week, and it turns into a part-time job on top of running your park.

ChatGPT also won’t score calls the same way twice. One week it might flag a mishandled lead. The next week, the same mistake might get missed. Without a standard way of scoring, you can’t compare this month to last month, or track whether things are getting better or worse.

Our system is different. It audits every call against the same standards, tracks callers and patterns across months, and keeps the evidence connected to each finding. Ask SignalPony then lets you investigate that structured record in plain English—without pulling recordings, rebuilding context or hoping a one-off prompt notices the same things twice.

ChatGPT wasn’t built for any of that. We were.

We review one representative week of recorded calls and show the missed follow-up, resident issues, risk signals, and evidence we find. If there is nothing meaningful in the calls, we will tell you that too.

Your phone system needs call recording turned on and a way for us to access the recordings. Most VoIP systems already support this.

Your audit is delivered in a private portal. Caller details stay inside your park’s audit, and every finding links back to the call evidence available to your authorized users.

No. SignalPony flags language and situations that deserve review. It does not make legal conclusions or replace counsel.