AI BDC Call Confusion: How AutoVox Handles IVR and Receptionist Loops

If someone told you an AI called your dealership, got routed to your receptionist, and spent three minutes pitching a service menu to her — you'd probably cross AI off your vendor list for good.

I get it. That story is real, it happens, and it's embarrassing for everyone pushing AI voice technology right now. I'm not going to pretend it doesn't happen with some systems. What I want to do instead is explain exactly why it happens, what we built to stop it, and what trade-offs still exist that you should know about before you talk to any vendor, including us.

Why AI Agents Get Stuck Talking to Phone Trees Instead of Decision-Makers

Most outbound AI voice agents are built for one thing: talking. They are optimized to deliver a script, handle objections, and book a meeting. What they are not optimized for — at least not by default — is listening to the context of who just picked up the phone before launching into that script.

When your dealership's main line picks up, a lot can happen before a human decision-maker ever speaks. You might have:

• An automated IVR that says "Press 1 for Sales, Press 2 for Service"
• A receptionist or operator who answers with "Thank you for calling Riverside Motors, how can I direct your call?"
• A voicemail system that sounds enough like a human that a poorly tuned AI just starts talking
• A floor rep who answers sales calls but has zero authority to make vendor decisions

A naive AI agent hears audio, detects that someone "answered," and starts the pitch. It doesn't know it's talking to a menu. It doesn't know it's talking to someone whose job is to route calls, not receive them. By the time the receptionist figures out what's going on, your AI has either been transferred to a random extension or disconnected — and you've burned a call credit on nothing.

This is not a hypothetical edge case. It is the most common failure mode in outbound AI calling, and any founder or vendor who tells you otherwise is selling you something.

What "Detecting a Decision-Maker" Actually Looks Like in Practice

The fix sounds simple: before you say anything about why you're calling, figure out who picked up.

In practice, this requires the AI to do a few things in the first three to five seconds of a call that most systems skip entirely.

First, the agent needs to listen for IVR indicators — keypad prompts, hold music cues, automated voice patterns — and either navigate them or terminate the call and retry at a different time or number. Burning a call on an IVR you can't get through helps nobody.

Second, when a human does answer, the agent needs to ask a qualifying question before delivering any content. Not a weak "Can I speak to the GM please?" — that's easy to deflect. Something with a decision-branch built in, like: "Before I go further — am I speaking with the General Manager, or would it be better to call back at a specific time to reach them directly?"

That question does several things at once. It signals that the call has a specific purpose. It respects whoever answered by not assuming they're the right person. It gives the receptionist or floor rep a clean, non-awkward way to redirect. And if the GM does pick up, it establishes immediately that this call is for them specifically — which tends to get more engagement than a cold opener that could be meant for anyone.

This is the exact language our agent uses, and it came directly from listening to real dealer calls where the previous version was failing. We didn't engineer it from theory. We tested it on actual dealership phone trees until it stopped getting stuck.

The Three Scenarios Every Outbound AI Call Has to Handle Correctly

When an outbound AI calls a dealership — or when your AI receives an inbound call that gets internally routed before reaching you — there are really only a few situations it can land in. Here's how each one should be handled:

1. IVR or phone tree answers. The agent should detect the automated response, attempt navigation if the options are clear, and if not, flag the call for a retry at a direct number or a time when a human is more likely to answer directly.

2. Receptionist or operator answers. The agent should not pitch. It should qualify — politely and specifically ask whether it's reached the right person, and if not, ask for the best time or direct number to reach the GM or decision-maker.

3. Decision-maker answers directly. Now the agent can engage. Not before. The qualifying question still happens, but it's fast, and the conversation can move forward with the right person on the line.

4. Voicemail. The agent should leave a short, human-sounding message with a clear callback hook and not attempt to have a conversation with a recording.

5. Floor rep or non-decision-maker picks up a sales line. Similar to the receptionist scenario — qualify first, don't pitch, and either get routed or schedule a callback.

Every one of these scenarios needs a different response. An agent that only handles one of them reliably is going to waste a meaningful percentage of your call volume on dead ends.

Why This Problem Is Worse for Dealerships Than Other Industries

Dealerships have a specific phone infrastructure problem that most other small businesses don't have: you deliberately route incoming calls through multiple layers before they reach anyone with authority.

That's good for customer experience. It means callers get to the right department. But it also means that any outbound AI calling into your store — or any inbound AI routing call that bounces internally — is navigating a system designed to filter and redirect.

Layered on top of that, dealership staff turnover means the person answering the phone this Tuesday might be brand new and have no idea who handles vendor conversations. The receptionist might transfer an AI agent to the wrong department three times before anyone figures out what's happening.

None of this is your fault. But it means the bar for AI call intelligence at a dealership is higher than at, say, a dentist's office with one front desk person. Your phone environment is complex, and an AI that wasn't specifically tuned for that environment is going to struggle.

This is part of why we focus specifically on automotive dealers rather than trying to be a generic AI BDC for every industry. The edge cases in your phone system are different from the edge cases in a law firm's phone system. If you're evaluating any AI voice product, ask them directly: how does your agent handle a dealership IVR, and what happens when it reaches a receptionist instead of a GM? If they give you a vague answer about "intelligent call routing," keep asking until they show you.

For what it's worth, you can see how we've structured the outbound and inbound call logic for dealership sales flows specifically at autovox.vernallc.com/stacks/sales — it's not a sales page, it's the actual flow breakdown.

What the Data Says About Missed Calls and Why Getting Through to the Right Person Matters

Before I close, I want to ground this in something concrete rather than just talking about call logic.

According to Cox Automotive's 2023 Car Buyer Journey Study, the majority of car buyers who contact a dealership expect a response within an hour, and a significant portion will move on to a competitor if they don't get one. The lost-lead problem at dealerships isn't primarily about marketing spend or inventory — it's about response time and follow-through on calls that are already coming in.

That's the inbound side. On the outbound side — when your BDC is calling leads back — the same logic applies in reverse. Every call your team makes that ends in an IVR loop, a wrong transfer, or a confused receptionist interaction is a lead aging out while you're burning time on phone system navigation.

A human BDC rep can cost you $5,000 to $8,000 a month all-in when you factor in salary, benefits, turnover, and training time. That cost is only justified if the reps are actually reaching decision-makers and having productive conversations. If a material percentage of their call time is being eaten by phone tree navigation and receptionist handoffs — and it is, at almost every store I've talked to — you're paying human wages for work that should be automated.

The argument for AI BDC isn't that AI is smarter than your best rep. It isn't. Your best rep, on a good day, is more persuasive and more adaptable than any AI we have access to right now. The argument is that AI can handle the volume, the repetition, the off-hours calls, and yes — the phone tree navigation — at a cost structure that doesn't require you to staff for peak load and then carry that overhead through slow months.

But only if it's built to actually get through to the right person. An AI that talks to your IVR for three minutes and calls it a completed call isn't saving you anything. It's a new kind of waste with a technology label on it.

We built AutoVox specifically because we kept seeing that failure mode in other tools and thought we could do the qualifier logic better. Maybe we're right, maybe you'll poke holes in it.

Don't take my word for it. Call our live AI agent right now at +1 (604) 229-7496 and try to buy a car from it.