Why Captive Agents Struggle With Third-Party AI Texting Vendors

Texting is the highest-response channel a captive agency has, and it is the one most likely to get an agency in trouble. The gap between those two facts is where third-party AI texting vendors sell — and where captive agencies keep getting stuck.

Headline. The problem is almost never the AI. It is that a texting tool bolted onto the side of an agency creates a second, invisible system of record for consent — and consent is the one piece of state that must never be wrong. Add A2P 10DLC registration that many vendors quietly push back onto you, plus carrier brand rules that generic vendors have never heard of, and you get campaigns that are rejected, filtered, or compliant-on-paper-only.

We have been through the registration process ourselves for our own product's numbers. Everything below is what that cost us in time, not a summary of someone else's blog post.

Problem 1: A2P 10DLC Is Real, and It Is Your Problem

Since February 1, 2025, US mobile carriers block unregistered application-to-person traffic from 10-digit numbers outright. Not throttle — block. Registration runs through The Campaign Registry in two steps: a Brand (your legal entity, EIN, address) and a Campaign (what kind of messages, what the opt-in looks like, sample copy). Maintenance typically runs $300 to $500 a year.

The delivery difference is not marginal. Properly registered traffic generally lands in the 95 to 99 percent range. Unregistered or badly registered traffic lands far below that — and, worse, it fails silently. Nobody bounces. The producer sees "sent," the client never got it, and the renewal call that was supposed to follow the text never happens.

Two things go wrong for captive agencies specifically:

Registering under the wrong entity. You register under your agency's legal entity and EIN — the one that would appear on your own tax return — not the carrier's. Vendors unfamiliar with the captive channel sometimes assume the carrier is the brand. That gets rejected, and the rejection loop costs weeks.
Vendors who "support" registration without doing it. "We support 10DLC" frequently means "there is a form in our settings page." Ask to see your approved Brand ID and Campaign ID. If a vendor cannot show you both, you are not registered.

Problem 2: You Have a Brand You Do Not Own

This is the constraint generic vendors have simply never encountered, because it does not exist outside the captive channel. A captive agent markets under the carrier's brand, and the carrier has advertising and brand-use rules that govern outbound messaging.

An AI texting tool that generates message copy on its own is, structurally, a tool that generates unreviewed brand communications. That is fine for a plumber. For a captive insurance agency it is a compliance question, and it is the reason "AI writes and sends your texts automatically" should make you slow down rather than speed up.

The workable version is AI-drafted, human-approved: the model writes the message, a person reads it, the system sends it. That preserves the time savings — writing is the slow part — and keeps a licensed human between the model and the client.

Problem 3: Consent Lives Somewhere Nobody Looks

This is the one that actually causes harm, and it is architectural rather than legal.

A client opts out by replying STOP to a renewal reminder. That opt-out is recorded — in the texting vendor. Three weeks later a producer pulls a cross-sell list out of the CRM, which knows nothing about it, and sends a round-out campaign. The client gets messaged after opting out.

Nobody did anything wrong at any individual step. The failure is that consent state and client state live in two systems, and the one people work in daily is not the one holding the consent. Every "integration" reduces the window but does not close it, because integration is a copy with a delay, and opt-out is a fact that must be true instantly everywhere.

The structural fix is the boring one: the messaging system and the client record are the same system. That is why AgencyIQ has the phone and messaging inside the CRM rather than integrated with one — the same argument we made about voice in the JustCall teardown.

Problem 4: Quiet Hours and TCPA Are Not the Producer's Job to Remember

The TCPA carries statutory damages per violation and is actively litigated. Consent, quiet hours, and honoring opt-outs all matter, and they matter per message.

Expecting a busy producer to check the recipient's time zone before sending is a control that will fail. The system should refuse to send outside quiet hours and queue the message for the next window instead. If your texting tool will happily fire a message at 9:40 p.m. local because someone clicked send, that is a defect, not a feature.

None of this is legal advice — get your own on your specific program. It is an architecture argument: compliance rules that depend on a human remembering under time pressure are not implemented.

The Five Questions to Ask Any Texting Vendor

"Show me my approved Brand ID and Campaign ID." Not the settings page. The approvals.
"Where is consent stored, and can I see it on the client record with a source and timestamp?"
"When someone replies STOP, what else in my stack knows within one second?"
"Does the system enforce quiet hours, or does it trust the sender?"
"Does AI send, or does AI draft and a person send?" For a captive agency, the second answer is the right one.

A vendor that answers all five cleanly is worth talking to regardless of whether it is us. A vendor that gets defensive about the first one is the reason this post exists.

Our Position

Most of the "AI texting for insurance" category is a messaging app with a language model attached, sold on response-rate statistics that are true and beside the point. The response rate was never the hard part. The hard part is that texting is the one channel where a system-design mistake becomes a legal exposure, and a captive agency carries brand obligations on top of that.

We would rather an agency send fewer, slower, human-approved messages from a system that cannot text an opted-out client than send more from one that can. If that costs some response rate, it is the cheapest insurance in the stack.

If you want a straight read on a texting vendor you are evaluating — including one that is not ours — book a 30-minute call and bring their answers to the five questions.

Quick Answers

Why do captive agents struggle with third-party AI texting vendors?

Four reasons stack on top of each other. First, A2P 10DLC registration: US mobile carriers have blocked unregistered business texting traffic since February 1, 2025, and registration requires a verified brand and an approved campaign — many small vendors push that work back onto the agency. Second, brand identity: a captive agent registers under their own agency's legal entity, not the carrier's, and vendors often get this wrong. Third, carrier advertising and brand rules govern what a captive agency may say in outbound messages. Fourth, most texting vendors are not connected to the agency's CRM, so consent and opt-out state lives in a system nobody checks before the next send.

What is A2P 10DLC and does an insurance agency need it?

A2P 10DLC (Application-to-Person, 10-Digit Long Code) is the US carrier framework for business texting from standard 10-digit numbers. Since February 1, 2025 US carriers block unregistered business traffic outright. Registration runs through The Campaign Registry: the business registers a Brand, then registers a Campaign describing the message types and opt-in method, and typically runs $300 to $500 per year to maintain. Registered senders generally see roughly 95 to 99 percent delivery; unregistered or poorly registered traffic lands far lower. Yes, an agency texting clients needs it.

Can a captive agent text clients under the carrier's brand?

Register under your own agency's legal entity — the entity with your EIN — not the carrier's. You are the sender and the party responsible for consent. Whether your message copy may use the carrier's brand name is a separate question governed by your agency agreement and the carrier's advertising rules, and it should be confirmed with your field leadership. The registration identity question and the ad-copy question get conflated constantly, and getting the first one wrong is what causes rejected campaigns.

What are the penalties for non-compliant insurance texting?

There are two separate exposures. Carrier-level: unregistered or violating traffic gets filtered or blocked, and messaging-industry penalties for sending without proper registration or consent are commonly cited in the $500 to $1,500 per-message range. Legal: the TCPA carries statutory damages per violation and is actively litigated, and consent, quiet hours, and honoring opt-outs all matter. This is not a place to rely on a vendor's assurance — get advice on your own program.

What should a captive agency look for in an AI texting tool?

Five things: (1) the vendor registers 10DLC under your agency entity and shows you the approved Brand and Campaign, not just a promise; (2) consent is captured and stored on the client record, with the source and timestamp; (3) opt-out is honored automatically and permanently across every message type; (4) quiet hours are enforced by the system, not by the producer remembering; (5) the messages live in the same client timeline as calls and emails, so the next person to touch the household sees what was sent. If texting lives in a separate app from your CRM, you will eventually text someone who opted out.