I get asked this question more than any other when I'm at industry events. Some version of "should I be buying leads" or "I tried buying leads once and they were garbage." And the answer is the same every time. The leads aren't the problem.
I've been on every side of this. I've sold leads. I've bought leads. I owned a 100-seat domestic call center that ran 500 to 600 transfers a day. I've worked with offshore operations in the Philippines and India. Everything tends to break the exact same way, regardless of where you run it from or what you pay for it. The breakage is operational, not lead-related.
What I want to do here is walk through the playbook that's worked for 20 years, the myths that keep agents from running it, and why the math finally became accessible to every agency this year. If you want broader category context, our overview of conversational AI for insurance covers the landscape this post sits inside.
Three things have always made purchased leads work.
Speed-to-lead. Full follow-up cadence. Consistent messaging.
If you respond instantly, hit every touch in a real cadence (eight calls is a fair baseline), and say roughly the same thing every time, the leads convert at predictable rates. This isn't a secret. Every operator in the lead-buying business knows it. You can read it in any training material from any vendor going back two decades.
The way agencies pulled it off historically was simple. They ran like call centers. Dialer infrastructure, tight scripting, QA, coaching loops, and a centralized operation with strict SLAs. The agencies that did this won. The ones that didn't, lost.
The reason most agencies don't run like call centers is that call centers are operationally brutal. Reps are hard to hire. Training runs three to six months with high attrition. Process consistency breaks down inside a quarter without active management. And scaling means more headcount, more management layers, more cost. The playbook was free. The capability to execute it was the part that cost millions.
That's the actual story of why most agencies struggle with purchased leads. Not lead quality. Capability.
Once you've established that the playbook is accessible to anyone who can run it, the next conversation is always about what people think they know about leads. Three myths come up over and over.
It doesn't. Price indicates competition for the lead.
Higher-priced leads are usually the leads your producers complain about the least. That's the actual pricing model. You're not paying for better data. You're paying for fewer complaints from the people working them.
The right question isn't whether a lead is cheap or expensive. It's whether the lead fits a cost of acquisition that works for your business. A $1 lead, a $10 lead, a $30 live transfer, a $90 live transfer. None of these are automatically better than the others. They're different price points for the same question: does this fit my budget for a new customer.
This one's expensive to learn. The pitch sounds clean. Pay more, get a lead nobody else is calling, watch your close rate climb.
The reality is that consumers shop around no matter what. Whether your lead is exclusive or shared, the customer is calling three more places before lunch. The conversion gap between exclusive and shared is flatter than agents expect. Sometimes there's no gap at all.
If your process is broken, exclusive doesn't fix it. It just means fewer competitors are also calling. You're still losing the lead, but now you paid a premium to lose it. The extra spend goes to the vendor, not to your closing rate.
The leads are not inconsistent. The process is.
I keep one set of charts on hand for this exact conversation. The first chart shows one week of conversion data from a single lead vendor. The line drops from 9% to 4% over five days. If that's all you saw, you'd swear the vendor sold you garbage.
The second chart shows the same vendor over six months. The line is remarkably stable, climbing from 3% to 8% as the process matured. Same source. Same filters. Same lead type. The only thing that changed between week one and month five was that the process started treating every lead the same way every time.
This is the most important thing to understand about buying leads. Same inputs produce the same outcomes over time. The day-to-day experience is variable. The aggregate is predictable. Operators who can hold steady through a bad week make money. The ones who panic and switch vendors don't.
Here's a simple frame. If you're buying a $6 lead and need a 2% conversion to hit a $300 cost of acquisition, your math works the moment two out of every hundred leads close. The other 98 are the cost of getting to the two that fund the business.
Most agents fixate on those 98. They count every "no" and convince themselves the leads are bad. The reason is operational. Every one of those 98 leads costs human effort. A full cadence on 100 leads is roughly 800 dials. By day three, "price of admission" starts feeling like punishment. The producer working lead 47 has stopped caring. The cadence collapses. The two that would have closed don't.
The math has always been right. The friction was always in the execution.
AI didn't invent the playbook. AI makes the playbook accessible.
Speed-to-lead, full cadence, and consistent messaging are the things AI does better than humans at scale. The system works lead 1 and lead 1,000 with the same discipline. It doesn't get tired on Friday. It doesn't skip the day-six follow-up because something else came up. The variable volume problem that breaks every human call center, where Monday's 10 leads get handled differently than Tuesday's 100, just stops being a problem.
This is the part most agents miss when they hear "AI for insurance." It's not about replacing the producer with a robot. It's about removing the operational ceiling that kept the playbook out of reach. The agencies that used to lose to the big call-center operators now have the call center too. The fight is even.
The other piece is cost. Running a domestic SDR seat runs $30,000 to $60,000 a year. Offshore runs $15,000 and up, with quality variance you have to manage. AI runs a fraction of that, with consistency you actually can predict. For the first time, the playbook is operationally feasible at any scale. We broke down the margin expansion math in detail in a separate post if you want to see what happens when agencies reinvest the savings.
A real example from one of our agencies. We'll call him Gary.
Day 1: A 1:16 PM text reminding Gary about an auto quote he'd started for his Tahoe. By 2:05 PM, we'd qualified him in full. Vehicles confirmed. Homeownership confirmed in Stockton, California. Five-plus years of insurance history.
Day 1 still: At 2:36 PM, Gary tells us he's in the hospital and won't be out until tomorrow.
This is where most agencies lose the lead. The model wasn't trained to handle "I'm in the hospital." That phrase isn't in any script. What it had was enough context to respond like a person would. The reply was a short acknowledgment, an apology, and a question about timing for the next day. No pressure. No abandonment.
Day 2, Day 3, Day 4: Mav continued the cadence with appropriate spacing. A check-in. A "hope you're doing okay" with a soft offer to set the call. Nothing that read like a chase. This is what automated lead nurture looks like when persistence is operational instead of personal.
Day 4: A final continue-or-stop message. Gary replied yes. Within a minute, we transferred him to a licensed agent. He took the call.
Most agencies would have lost this lead three different times. Once at the hospital message, because the producer felt awkward following up. Once on Day 3, because the producer got busy. Once on Day 4, because no one remembered the thread.
The system didn't lose Gary because the system doesn't have a Tuesday. Every lead gets the full cadence. Every interruption gets handled with context. Every recovery gets a clean transfer.
This is what process execution at scale looks like when humans aren't the bottleneck.
Three agencies tend to get the most out of running purchased leads through an AI call center.
First, the relationship agent who's hit a ceiling. You've never bought leads because you've heard they don't work. Your book is built on referrals, and growth has flattened. The consistency layer is what makes leads finally survivable for an agency that doesn't already run like a call center.
Second, the agency owner who's losing top producers. Your best agents don't want to dial 200 leads a day. The agency down the street is feeding their producers 10 inbound calls a day. Recruiting is one-sided in that fight. Giving your producers consistent, qualified conversations instead of dial lists is how you keep them.
Third, the seasoned operator already buying leads. You know the playbook. You're running it. The constraint is headcount. Plugging AI in behind or ahead of your existing call center lets you 2-5x the volume you can work without 2-5x the team. If you're thinking about the agency math at this scale, our piece on how to build a $10M insurance agency without a call center is a longer read on what's possible.
Not in a way that justifies the premium most of the time. Consumers shop around whether your lead is exclusive or not. The conversion gap between exclusive and shared is flatter than the price gap. If your process is broken, exclusive won't fix it. If your process is sharp, shared works fine.
Higher-priced leads tend to be the leads producers complain about the least, which is why they cost more. Conversion is a process variable, not a price variable. Sharper process closes more leads at any price point.
Yes, if your cost of acquisition math works at that price. Aged leads cost less and convert less. Whether that's a good trade depends on whether your CPA budget fits the math. If it does, the lead is good.
I won't quote a number that applies to every business, because it's a budget question, not a benchmark question. The right CPA is whatever fits inside your lifetime customer value and your operating math. If you're hitting your number, the leads are working, regardless of what anyone else's number looks like.
Instantly. Every lead. Every time. This is "speed-to-lead" and it's the single biggest predictor of whether a lead converts. The advantage of running an AI call center is that it answers in seconds, every time, without exception. Manual outreach can't compete with that on response time. SMS is the channel that actually delivers on speed-to-lead, which is why we built Mav around it.
A full cadence means hitting every scheduled touch on every lead, with no exceptions. Eight calls is a fair baseline. The agencies that work all eight calls on every lead consistently outperform the ones that work three calls on the leads that "feel good." The cadence is the asset. Discipline on the cadence is the differentiator.
Leads work well when they're worked well. That's the whole game. The 20-year playbook is the same playbook. The only thing that changed is who can run it.
If any of this sounds like the agency you're running or the one you're trying to build, the easiest thing is to see the system work on a few of your own leads. Book a demo and we'll show you the math on a sample set.
