Episode 15: Do More With Every Call Using Your Upgraded AI Calling Agent

Chris: Welcome back to the Deep Dive. You know that, uh, that sinking feeling you get when you dial a business and hear a robot voice?

Elizabeth: Oh, yeah. It's that classic flowchart trap. Like press one for this, press two to just get stuck in a loop forever.

Chris: Exactly. Well, today we are exploring the May 2026 feature update notes for the Sales Message AI calling agent, and our mission is to understand how automated calling is evolving from those, you know, basic Q&A bots into fully capable frontline team members.

Elizabeth: Right. And to really understand that evolution, we first have to look at how the AI is now actively multitasking, like during a live conversation.

Chris: Yeah. Let's talk about that dynamic mid-call SMS capability. The notes say if you ask the AI for an appointment, it texts you a booking link instantly while you are still on the line.

Elizabeth: Yeah, and then it verbally confirms it too.

Chris: Right. But if it's texting me while we're talking, I mean, how does it know I'm not still speaking? Doesn't that risk stepping all over the caller's audio?

Elizabeth: So that is the jump from linear processing to parallel processing. Older bots had to, um, literally stop listening to execute a task.

Chris: Which causes those awkward pauses.

Elizabeth: Exactly. But this AI maintains dual streams. It continuously analyzes your audio input for intent triggers.

Chris: Like hearing you ask for a calendar link.

Elizabeth: Yeah. And the moment it hears that, it fires off an API call to send that text via a secondary background process, all while its primary listening stream stays totally locked onto your voice.

Chris: So it's kind of like a, a hyper-efficient receptionist who hands you a clipboard with one hand while perfectly updating a Rolodex with the other.

Elizabeth: That's a great analogy, yeah.

Chris: But doesn't receiving a perfectly timed text while conversing with a machine completely rewrite our expectations of automated customer service?

Elizabeth: Oh, absolutely. And it goes further. While it's multitasking, it uses something called merge fields to collect missing contact data.

Chris: How does the actual mechanism work there?

Elizabeth: Well, a merge field is essentially a placeholder. Because the AI is actively scanning your profile in Sales Message and HubSpot during the call, it notices if, say, your email is missing.

Chris: Oh, wow.

Elizabeth: Right. So it naturally steers the conversation to ask for your email, extracts it from your speech, and instantly saves it to the CRM before the call even ends.

Chris: I mean, that is a massive leap for CRM hygiene. Ensuring customer profiles are accurate is usually this manual post-call chore that human agents just constantly forget to do.

Elizabeth: Precisely. It closes the loop between live conversations and data entry. There's zero manual follow-up required.

Chris: Okay. But handling all this real-time stuff naturally raises the question of what happens when a caller's request exceeds the AI's capabilities because people are gonna throw weird nuanced problems at it.

Elizabeth: For sure. And when that happens, it performs automatic human escalation. It seamlessly routes callers to a ring group without those awkward holes or weird dead air.

Chris: Wait, I want to push back on this a bit.

Elizabeth: Okay, sure.

Chris: Isn't an escalation fundamentally a failure state for an AI agent? Like if it has to pass the baton, it hit a wall, right?

Elizabeth: I see why you'd frame it that way, but knowing when to step back is actually an engineered feature, not a failure.

Chris: How so?

Elizabeth: Because of what managers can see afterward. There's a new handled by AI filter in the calls tab, plus separate timelines to see exactly what the AI did versus what the human handled.

Chris: I'm still not seeing how a filter makes a failed call a success though. If a human had to rescue the interaction anyway, why does a separate timeline matter?

Elizabeth: Because without that visibility, the call data is a black box. The filter mathematically separates the bot's actions from the human's.

Chris: Oh, so you can see what the AI successfully gathered.

Elizabeth: Exactly. You see the SMS sent, the email captured, all before the human took over. It gives teams the confidence to trust the AI with more volume, knowing the handoff is seamless.

Chris: So because fewer manual tasks are required, human teams only step in when truly needed for complex problem-solving.

Elizabeth: You got it. It's about letting humans do what they do best.

Chris: But think about where this leads for you as a consumer. If an AI can anticipate your needs mid-sentence, instantly text you the right link, and update your file without ever putting you on hold, I mean, how long until we actually prefer speaking to the bot over a human?

Elizabeth: Yeah, that's a wild thought.

Chris: Something for you to ponder. Thank you so much for listening, and we'll see you next week on the Deep Dive.