Insurance agency automation: where the real ROI hides, and the TCPA changes that protect it

Only 7% of insurers have scaled AI. The ones pulling ahead started with the layer most agencies automate last, and this fall the rules around that layer moved more in one month than they had in years.

I've spent a lot of years walking into agencies, and the automation conversation almost always starts the same way: they bought something, ran a pilot, and quietly set it aside. The tool usually wasn't the problem. What tripped them up was the order they automated in. So before you buy the next thing, here's where I'd start, and why:

Automate the communication work first, the calls, texts, and follow-up. It's where the hours actually go, and your licensed people are spending most of their day on work that doesn't need a license.

It's the most measurable layer you've got. Leads contacted within five minutes convert 21 times more often than leads reached after 30, and most agencies still respond in hours.

Make sure whatever you pick sits on top of your AMS and CRM instead of asking you to rebuild.

Build compliance into the workflow before you scale it. The FCC and the courts both moved in September, and I break down what changed further down.

Start narrow. One workflow or one lead source, measure it for 30 days, then expand. That's how you find out what's real before you bet the office on it.

Where the market is heading

On the captive side, carriers are consolidating their tech from corporate down, so a lot of agents are getting tools handed to them instead of choosing their own. Across the industry the outreach model keeps cycling. It went from leads, to calls, and now most shops run some hybrid in between, hunting for a cost-effective way to reach people. A lot of them solved that with third-party teams outside the US, and that move is getting harder to defend. The FCC has an open proposal looking at onshoring requirements, and the compliance lawyers I follow keep raising the same data problem: once your customer list is sitting in an overseas call center, you have very little recourse if it walks out the door.

The survey data backs up where this nets out. BCG found only 7% of insurers have actually scaled AI, even though the industry tests it more than almost anyone. Two-thirds are stuck in pilot. Meanwhile a Vertafore survey of more than 1,300 agency professionals found almost two-thirds optimistic going into 2026. We're in the fourth quarter now, and that gap between interest and execution hasn't closed much. KPMG's work points at why: most shops don't have the tools or the metrics to measure automation ROI, and the skills gap makes it harder to wire in than the demo suggested. BCG's read that insurers pull back after piloting is the same thing I watch happen at the agency level. Vertafore says the industry is ready. Ready and running are two different things.

What's breaking inside the agency

The pattern that shows up most is the big agency so focused on feeding the machine, more marketing dollars, more volume, chasing the comp plan, that they lose sight of the margin and their cost per acquisition. Growth becomes the only number they watch, and the math quietly stops working.

The instinct when the phones aren't producing is to hire. That's where the human-capital problem starts. I've watched agents buy a second book, build a second office, and realize almost immediately they don't know how to manage it. Now they're putting in 90-plus hour weeks trying to hold it together and retain what they've already got. That's why the staffing math stops working for agencies running real lead volume.

The day-to-day underneath it is your licensed agents doing work that doesn't need a license: calling back no-answers, chasing documents, fielding payment questions, re-keying the same information from one system into the next. Most owners have already tried to fix it by buying tools. A CRM, a dialer, an AMS, a marketing platform stacked on top. Each one handles a slice, but none of them owns the communication work end to end, so the team ends up toggling between dashboards instead of doing the work those dashboards were supposed to take off their plate.

Here's how I'd sort it. Agency work falls into three kinds. Communication: lead contact, follow-up, renewal outreach, service questions. Processing: data entry, documents, compliance checks. Decision work: underwriting, advising, closing. The mistake is automating all three at once, or starting with the wrong one. Communication is the highest-volume kind and the most predictable, which is exactly why it's the first worth handing off. Agencies that automate renewal outreach hold onto 94% of clients against 81% for the ones doing it by hand. That gap lands straight on the book.

What it looks like when it runs

Done right, the system handles the first contact, runs the multi-touch follow-up across text and calls, qualifies the person, and only routes the ready-to-talk conversations to a licensed producer. Your people stop starting the morning on dead dials and start it on real conversations. That's the layer we built Mav to run, with the opt-outs and consent tracking living inside the workflow from the first message. Which brings me to the part of this I'd spend the most time on this quarter.

Compliance: what changed this fall, and what I'd do about it

I've watched the TCPA rewrite the playbook more than once. This fall it happened on several fronts inside a few weeks, and some of it went in the agency's favor. Here's what moved, in plain terms. I'm not a lawyer, so run any of this past yours before you change how you operate. I lean on TCPAWorld and Eric J. Troutman's Deserve to Win podcast to keep up, and both are worth your time.

The FCC loosened opt-outs, then opened the door to tighter deadlines. In September the FCC adopted an order that walks back the "revoke all" rule that was set to hit January 31, 2027 (TCPAWorld's breakdown). Two changes matter for an agency. You can now designate an exclusive way for people to opt out, like replying STOP or using a specific number or web page, as long as you disclose it clearly on the call or in the text. If you don't designate one, the old standard still applies and any reasonable method counts. And an opt-out from an informational message, like a renewal reminder or a payment notice, can stay limited to that type of message. An opt-out from marketing still ends all marketing from you. The order takes effect 30 days after it's published in the Federal Register, so confirm the date with counsel before you change your opt-out language.

Alongside it, the FCC opened a new proposal that could shrink the 10-business-day window you get to honor an opt-out and require two-way texting, so people can always reply to stop. Most agencies I know that run a few producers, a lead vendor, and a texting tool already struggle to sync opt-outs inside 10 days. If that window gets shorter, anything done by hand breaks first. People also don't always type STOP. Palm Beach Tan paid $2.5 million over texts sent after customers asked to stop in their own words, and one court held that a STOP to a text can cut off calls too.

The Cannon ruling. On September 11, Judge Aileen Cannon in the Southern District of Florida ruled in Anthony v. Brian Marketing Group that cell phones don't count as "residential" lines under the TCPA's Do Not Call rules. Her reasoning: Congress authorized DNC protection for residential subscribers back in 1991, and the FCC went past that authority when it stretched the definition to cover wireless numbers in 2003. About three weeks later, a federal court in Missouri reached the same conclusion. Troutman said on his podcast that if this view spreads nationally, the DNC list is close to dead, since almost nobody has a landline anymore.

My advice: leave your scrub exactly where it is. These are two district court rulings, and most courts still treat cell phones as residential. Finish Line lost that exact argument in a Texas federal court this year, against a plaintiff with no lawyer. Consumer advocates are already pushing Congress to close the gap, and a ruling in Florida does nothing for a call coming out of your office in Ohio. Keep scrubbing against the National DNC Registry and your internal list as if nothing happened.

Texts vs. calls is still a coin flip. The same split runs through texting. In July the Seventh Circuit held that texts aren't "calls" for DNC purposes, and a Missouri court followed. New Jersey courts keep going the other way. If you write business across state lines, plan for the strictest version.

Proving consent matters more than having it. The Troutman team put it simply on the podcast: autodialer cases are down, DNC cases are up. So the fight has moved to one question. Can you prove this person asked to hear from you? A Medicare marketer just lost a summary judgment bid after its Jornaya consent record was thrown out as hearsay. If you buy leads, take the advice Tom Ryan of Bigly Sales gave on that same episode: every lead vendor will tell you their leads are perfect, so trust but verify. Validate the TrustedForm or Jornaya certificate when the lead comes in, before anybody dials, and send back what doesn't check out.

Wrong numbers are where the big checks get written. Concora Credit agreed to pay $9.375 million to settle a wrong-number class action over prerecorded calls. Per the Troutman team, the deal also commits them to spend over $1 million on TCPA compliance across three years, under court supervision. That's a federal judge with a say in how they run outreach. When someone says "wrong number," every future touch to that number should stop right then, automatically.

Your state has its own rules. Pennsylvania's SB 992 takes effect October 18 and stacks state telemarketing rules on top of the federal ones. Florida, Texas, and Oklahoma already have their own. On the carrier side, unregistered business texting has been blocked since February 2025, so if your texting tool isn't registered for A2P 10DLC, your messages may never land. If your producers are licensed in several states, your outreach has to follow the strictest one.

What I'd actually do. The pattern I keep seeing in agencies that got burned: the right policy on paper and a manual process underneath it. A producer forgets to log a STOP. A vendor sends a lead with a stale certificate. Somebody dials off a spreadsheet nobody scrubbed. People make mistakes, so the system has to catch them. That means opt-outs, DNC scrubs, quiet hours, wrong-number flags, and consent checks running automatically on every contact, with a record you can pull up later.

One line from the podcast stuck with me. Breaking down the Finish Line loss, Troutman said winning comes down to one word: credibility. Big law firms are losing TCPA arguments to people with no lawyer because judges stop trusting them. Outreach works the same way. An agency that can show a clean consent record, an opt-out honored the same day, and a scrub log is a much harder target.

Before you sign anything, ask me this

If you're weighing a platform, I'd start with four questions. What share of your team's time goes to work that doesn't require a license. How far outside that five-minute window your real lead response sits. Whether the tool layers onto what you already run or forces a rebuild. And whether it can show you, for any contact, the consent record and the full opt-out history. Answer those honestly and the starting point usually picks itself.

If you're working through that sequencing in your own shop, that's the conversation I like having most. Get in touch and tell me which of the four is hardest for you to answer right now. What I hear back tells me what to get into next time. All said, the agencies that pull ahead going into 2027 are the ones who get clear on which work still needs a person, hand the rest to a system that doesn't forget, and can prove they did it right.