Ryan Johnson, Chief Product Officer at CallRail, explains how AI is turning phone calls, chats, and texts into some of the most useful marketing data a business has.
Here’s a number that stopped me. CallRail sees about 250 million phone calls a year come through its platform. Around 50 million of them go unanswered.
That’s 50 million people who wanted to buy something, book something, or ask something, and nobody picked up.
I wasn’t on this episode of Deconstructing Data. Jessie Lee Zach hosted with Adam Fitzgerald, our enterprise sales director, and their guest was Ryan Johnson, Chief Product Officer at CallRail. I went back and watched the whole thing because conversation data is one of those areas most marketers still treat as an afterthought. Ryan made a pretty convincing case that it shouldn’t be.
Ryan’s path is a good one. He started in finance in Michigan, got pulled into SEO and SEM for law firms in the early 2000s, then helped build Vitrue, a social media marketing platform that Oracle acquired. After that he led product teams working on early computer vision and natural language processing before landing at CallRail in Atlanta, where a lot of his old Vitrue and Oracle colleagues work too. CallRail started as call tracking for small businesses and agencies. Today it’s a lead experience platform with conversation intelligence and voice AI built in.
What is conversation intelligence, and why is it suddenly so much better?
Put simply, conversation intelligence means understanding what happens inside a conversation. That could be a phone call, a text, a web chat, or a social message. Was the caller happy or angry? Did they want a service appointment or a new purchase?
CallRail has been working on this since 2016. The problem back then was the transcripts. If speech to text gets the words wrong, every analysis on top of it is wrong too.
“You think of like good in, good out, and now it’s great in and great out.”
— Ryan Johnson, Chief Product Officer, CallRail
Modern speech recognition fixed that. CallRail works with AssemblyAI for transcription, and near-perfect transcripts mean the summaries, intent detection, and coaching insights actually hold up.
The bigger change, according to Ryan, is moving from one conversation to thousands. Take a rep who handles 100 calls a day. Now you can ask what that rep does best, where they struggle, what products they push, and whether they follow the script. You can do that across phone, chat, and social for the same customer. Before, that kind of analysis just wasn’t possible.
Can conversation intelligence fix marketing attribution?
This was the part that got my attention, because attribution is something we talk about constantly on this show.
Ryan gave a simple example. Your neighbor Jenny tells you Joe’s Pool Service did a great job. You go home, Google “Joe’s pool service,” click the paid ad at the top, and call. Your analytics say a Google ad drove that lead.
But on the call, you mention that Jenny sent you. The real source was a referral.
“The true real attribution was a referral from his next door neighbor, not the Google paid ad. And that’s really important to understand, but there’s no way you could do that at scale without something like conversation intelligence.”
— Ryan Johnson, Chief Product Officer, CallRail
CallRail calls this self-reported attribution and has a patent pending on it. Businesses have always trained their staff to ask “how did you hear about us?” Ryan’s point is that it doesn’t scale. Staff forget to ask, or the answer comes up naturally later in the call and nobody writes it down.
He also shared a story I loved. A paving company that only did new asphalt parking lots kept getting calls about resealing and repairs. Whoever answered just said “nope, we don’t do that.” Conversation intelligence flagged the pattern. That’s either a new service line or a sign the marketing is attracting the wrong people. Either way, the business learned something it would never have seen in a dashboard.
Why is sentiment analysis about to get a lot smarter?
Here’s a subtle one. Most sentiment analysis today reads the text. Ryan thinks the next big unlock is reading actual emotion, from tone and context.
“I could be using profanity, but I could be using it in a positive way. Like, yeah, this is a damn good deal. If I say that now in certain tech stacks, it could be picked up as negative sentiment.”
— Ryan Johnson, Chief Product Officer, CallRail
The healthcare example makes it clearer. Someone calling their doctor about a cancer diagnosis is talking about something negative, but the call itself might have gone really well. Text-only sentiment gets that wrong. With better audio analysis and context for the model (“this is a doctor’s office, these topics are normal”), it gets it right.
Ryan also expects speech recognition in languages other than English to catch up. English is excellent, Spanish and French are good, and after that quality drops off fast.
How are voice AI agents changing lead conversion for small businesses?
Back to those 50 million missed calls. First contact matters a lot. If nobody answers, the caller just tries the next listing on Google.
Ryan sees businesses using their own call recordings to train voice AI agents for after hours, overflow, and front office work. The agent already knows how that business talks to its customers because it learned from real conversations.
“At 5:01 every day, if that’s when your business closes, you don’t have to pay a really expensive after hours service that’s hard to train and hard to keep up with.”
— Ryan Johnson, Chief Product Officer, CallRail
The agent can answer questions, book an appointment, or text a follow-up link. And because it knows where the lead came from and what page they were looking at, it can personalize the conversation. A landscaping company shouldn’t pitch lawn maintenance to someone who was just looking at a full backyard redesign.
Ryan was honest that this is still early, even at the enterprise level. But he’s seeing very small businesses adopt it fast.
Where does AI take marketing analytics next?
Ryan’s vision for what’s next is pretty appealing. Fewer reports. More answers.
“I envision this day that I wake up and I can go into CallRail and say, what happened yesterday? Tell me what happened with my business with all these people that called in. Tell me the good things, tell me the bad things.”
— Ryan Johnson, Chief Product Officer, CallRail
Picture the system telling you to spend another $50,000 on Google today because it predicts a specific revenue return, or that your best hours for ads are mid-afternoon. It learns what you care about and you get back to running your business.
He also called himself “agnostic” on where conversations will happen. Phone, SMS, social, maybe smart glasses someday. His bet is simple: conversations are going to happen somewhere, and businesses will need tools to understand them.
For his own team’s stack, Ryan named ChatPRD for product requirements, Gemini and Perplexity for research, AssemblyAI for transcription, and Napkin, a new tool that turns notes into visual diagrams. He also tests Claude and ChatGPT regularly to compare them.
The bigger picture
What I keep coming back to is how much intent data is sitting inside conversations that nobody analyzes. We spend a lot of time on clicks, pixels, and identity graphs. That all matters, and it’s what we do at BDEX. But when someone picks up the phone and says “my neighbor sent me” or “do you guys reseal parking lots,” they’re telling you exactly what you need to know.
The businesses that win will connect that conversation data with everything else they know about a customer, and they’ll answer the phone at 5:01, one way or another.
Connect with Ryan Johnson: Learn more about conversation intelligence and voice AI at CallRail, and connect with Ryan Johnson on LinkedIn.
Want to know more about the people behind your leads? Visit bdex.com and click “Talk to an Expert.”
Watch the full episode: From Conversations to Conversions With AI
This article was adapted from an episode of Deconstructing Data, BDEX’s weekly podcast on data-driven marketing. Tune in live every Thursday at 4:15 PM Eastern on LinkedIn.
About BDEX: For companies that need clean identity data to power their products, BDEX offers unmatched quality and execution. Visit bdex.com and click “Talk to an Expert” to get started.
