AI Voice Calling for Reactivation Campaigns Runs on Consent Records

TL;DR: AI voice calling for reactivation campaigns breaks on the list, not on the model, because the elapsed time that makes a contact dormant is the same elapsed time that ends your legal basis for calling it with a synthetic voice. Two regulators run two different clocks and they do not agree. The FTC measures an established business relationship at 540 days from a purchase, rental, lease or financial transaction and 90 days from an inquiry or application (16 CFR Part 310, the Telemarketing Sales Rule definitions), while the FCC measures the same relationship at eighteen months and three months (47 CFR 64.1200(f)(5)), so the shorter clock governs the campaign and most reactivation lists begin after both have run out. Even inside the window, the relationship only answers the do-not-call question. It does nothing for the voice. The FCC confirmed in its Declaratory Ruling adopted February 2, 2024 and released February 8, 2024 (FCC 24-17) that the TCPA restrictions on an artificial or prerecorded voice cover current AI technologies that generate human voices, which puts the call under 64.1200(a)(2) for wireless numbers and 64.1200(a)(3) for residential telemarketing, both of which require prior express written consent as defined at 64.1200(f)(9): an agreement bearing a signature, naming the telephone number, and clearly authorizing an automatic telephone dialing system or an artificial or prerecorded voice. A two-year-old box that says yes, contact me does not say any of that. Revocation makes a dormant list worse rather than better, because 64.1200(a)(10) lets a called party revoke by any reasonable method, treats stop, quit, end, revoke, opt out, cancel and unsubscribe as reasonable per se, forbids an exclusive channel, and requires the request honored within a reasonable time not to exceed ten business days. If the call delivers a prerecorded message, the Telemarketing Sales Rule at 16 CFR Part 310 adds mechanics most agent designs skip: fifteen seconds or four rings before disconnecting an unanswered call, and an automated interactive voice or keypress opt-out played within two seconds of the completed greeting, available at any point in the message, that immediately disconnects and adds the number to the seller’s entity-specific do-not-call list. Sort the list by consent provenance instead of last-touch date, point the synthetic voice only at records whose signed consent names both the number and the artificial voice, dial the rest with a person, and track opt-outs and complaints per hundred connects next to booked meetings.

Dormant does not mean dead. A deal stalls because a budget moved, a champion left, or a competing project ate the quarter, and the contact still recognizes the company a year later. That part of the reactivation pitch is true, and it is why every vendor on the first page of this search is selling you an agent.

Here is the part they skip. What defines the list you want to call? How long it has been quiet, and that is the same measurement the federal rules use to decide whether you still have permission to call it at all. Those two facts collide, and the collision happens before the agent says a word.

This article is about that collision. It uses the current text of the federal rules rather than a general reminder to check local requirements. A reminder to check requirements is not a campaign design. Every rule cited below is linked at the bottom. Confirm your own motion with counsel, and treat state law as additive rather than as a substitute, starting with which states require two-party consent to record the call you are about to make.

What AI Voice Calling for Reactivation Campaigns Actually Changes

Strip the demo away and the agent does four things. It places the call, asks a short set of questions the CRM already implies, writes a disposition, and then either books a next step or hands the conversation to a person. So what is it actually replacing? Rep time per attempt, on a list where most attempts go nowhere, which is the one input that has always made dormant-list outreach the first thing a sales leader cuts. That part is real.

What does it not change? Who you are allowed to dial. The agent inherits the list. If the list is wrong, the agent is a faster way to be wrong, and the damage arrives as complaints rather than as a bad quarter.

So name the campaign accurately before you design it. Why does the label matter this much? Because the four things teams lump together under reactivation carry different clocks and different consent histories, and only one of them is a genuinely cold call.

• Dormant lead reactivation revisits an inquiry or an opportunity that stopped moving.
• Customer win-back contacts someone who bought and then lapsed.
• Missed-call recovery returns an inbound call that never reached a person.
• Cold outbound reaches records with no prior inquiry and no prior transaction.

Those are not four flavors of the same campaign. A missed inbound call from this morning and a quote that went cold twenty months ago sit on opposite ends of every rule in this article. Most teams scope the pilot around script quality and voice realism. Then week three arrives and the constraint turns out to be the segment they exported on day one.

The Two Dormancy Clocks Every Reactivation Campaign Runs Against

The national do-not-call registry is the first gate, and the established business relationship is the thing that carries you through it, which is why every reactivation playbook you will read treats a prior purchase as the green light. Both regulators recognize that relationship. Neither one measures it the same way.

The FTC defines an established business relationship in the definitions section of the Telemarketing Sales Rule at 16 CFR Part 310 as a relationship based on the person’s purchase, rental or lease of the seller’s goods or services, or a financial transaction between them, within the 540 days immediately preceding the telemarketing call, or on the person’s inquiry or application regarding a good or service within the 90 days immediately preceding the call.

The FCC defines it at 47 CFR 64.1200(f)(5) as a prior or existing relationship formed by voluntary two-way communication, on the basis of a purchase or transaction within the eighteen months immediately preceding the call, or an inquiry or application within the three months immediately preceding the call.

Read those two numbers again. Eighteen months is roughly 548 days. The FTC gives you 540. So on a customer who last transacted 545 days ago, one regulator still sees a live relationship and the other one has already closed the window, which is a gap of about a week sitting inside a segment most teams define in whole quarters. Which clock governs? Both do, so the shorter one wins.

Two more conditions sit inside the FCC definition, and both of them delete relationships quietly enough that a segmentation query will never notice. A seller-specific do-not-call request terminates the relationship for telemarketing purposes even if the person keeps buying from you (64.1200(f)(5)(i)). The relationship also does not extend to affiliated entities unless the subscriber would reasonably expect it, given the goods or services offered and the identity of the affiliate (64.1200(f)(5)(ii)). Assembled your reactivation list across two brands after an acquisition? Then that second clause is not a footnote. It is the list.

So what does the clock change operationally? Queue priority. Reactivation is a decaying asset with a published expiry date. The right time to call the 90-day inquiry bucket is day 60, not whenever someone decides the pipeline looks thin. Sort the segment by days remaining on the shorter clock. Work the top of it first.

Why an Established Business Relationship Does Not Cover AI Voice Calls

This is the step that sinks most pilots, and it sinks them quietly, because a team that has carefully cleared the do-not-call question reasonably assumes it has cleared the call itself. It has not.

So where does the relationship actually live in the FCC rules? Not in an exemption. It is written into the definition of a telephone solicitation at 64.1200(f)(15), which excludes a call to any person with whom the caller has an established business relationship. Remove the call from that definition and 64.1200(c) never attaches in the first place, which is the subsection barring solicitations to registered numbers and outside the hours of 8 a.m. to 9 p.m. local time at the called party. The Telemarketing Sales Rule reaches the same place by a different road. It lists the relationship as a named exemption to the do-not-call prohibition in 16 CFR Part 310.

Now, what does that reasoning not touch? The voice. The prohibitions on an artificial or prerecorded voice live somewhere else entirely, at 64.1200(a)(2) for the wireless, paging and charged-call numbers described in (a)(1)(i) through (iii), and at 64.1200(a)(3) for residential lines. Neither one carries an established business relationship carve-out. The relationship got you past the registry. It did not buy you a synthetic voice.

And the synthetic voice is squarely inside those rules. In the Declaratory Ruling adopted February 2, 2024 and released February 8, 2024, the Commission confirmed that the TCPA restrictions on the use of an artificial or prerecorded voice encompass current AI technologies that generate human voices, so calls using those technologies require the prior express consent of the called party absent an emergency purpose or exemption. The ruling is short, it is plain, and it predates every vendor page on the first results page for this topic. If you want the architectural version of that decision, the tradeoffs between AI voice agents and rule-based call automation are a separate choice that consent comes before.

There is a related question worth separating out. Whether a given dialer counts as an automatic telephone dialing system is its own analysis, and it turns on how the system selects the numbers it calls rather than on how fast it calls them. That question applies to human-staffed calls too, and it is one of the TCPA rules that apply to outbound calling in general. Put it to counsel with your actual configuration in hand.

What Written Consent Must Say Before an AI Voice Reactivation Call

Most CRMs record consent as a checkbox, and most reactivation campaigns are approved on the strength of that checkbox being ticked. The rule does not describe a checkbox. It describes a document.

Prior express written consent is defined at 64.1200(f)(9) as an agreement, in writing, bearing the signature of the person called, that clearly authorizes the seller to deliver advertisements or telemarketing messages using an automatic telephone dialing system or an artificial or prerecorded voice, and that includes the telephone number to which those messages may be delivered. The agreement has to carry a clear and conspicuous disclosure that signing authorizes those calls, and that signing is not a condition of purchasing anything. An electronic or digital signature counts where it is valid under applicable federal or state contract law.

The Telemarketing Sales Rule asks for the same substance in its own words, in the prerecorded message provisions of 16 CFR Part 310. The express written agreement has to be obtained after a clear and conspicuous disclosure that its purpose is to authorize prerecorded calls, obtained without making it a condition of purchase, evidence the person’s willingness to receive prerecorded calls from a specific seller, and include that person’s telephone number and signature.

Hold your reactivation list against that. The record said yes to a demo request form in early 2024. Did that form name artificial or prerecorded voice? Did it name the number you are about to dial, or just whatever number the contact typed that day? Is there a signature? Or a server-side timestamp somebody will have to reconstruct from a log once a complaint arrives? For most teams the honest answer is no, no, and maybe.

That is not a reason to abandon the campaign. So what should you store instead? The artifact, not the flag. Keep the form text as rendered, the signature or its equivalent, the number submitted, the timestamp, and the page it came from, all attached to the contact record where the dialer and the auditor can both reach it. A consent field that cannot produce the document behind it is a note to yourself. It is not evidence.

The Revocations Hiding in Every Dormant Reactivation List

Every old list carries revocations that nobody wrote down, and the older the list, the more of them there are. Vendors never raise this part. It is also the part most likely to generate the complaint that ends the program.

Under 64.1200(a)(10) a called party may revoke prior express consent, including prior express written consent, using any reasonable method that clearly expresses a desire not to receive further calls or texts. A keypress or interactive voice opt-out during a call is reasonable per se. So is replying stop, quit, end, revoke, opt out, cancel or unsubscribe to a text. So is using a website or phone number the caller designated for opt-outs. If the reply uses other words, the caller still has to treat it as valid when a reasonable person would understand it that way. You may not designate an exclusive means of revoking, and every request made in any reasonable manner has to be honored within a reasonable time not to exceed ten business days from receipt.

Paragraph (a)(11) goes further. Revoking through some other channel, such as a voicemail or an email to a number or address meant to reach the caller, creates a rebuttable presumption that consent was revoked once the called party produces evidence that the request was made.

Now walk the queue. A rep dialed this contact in 2024. The contact said take me off your list. What did the rep do? Marked the deal closed lost, typed a note, and moved on. Closed lost was the field that paid them. The suppression field was somebody else’s job. Eighteen months later that contact is a perfect reactivation candidate by every segmentation rule in the first section, and calling it is the single worst thing this campaign can do.

So run the revocation sweep before the campaign, not during it. Where do those revocations live? In free-text call notes, in email bodies, and in inbound text replies. Search all three for the per se words and their obvious cousins, then promote every hit into the suppression field. Fix the capture path next, so the following one lands in a field instead of a note. If an opt-out is only recorded when a rep remembers to record it, you do not have an opt-out process. You have a habit.

Opt-Out Mechanics an Automated Reactivation Call Has to Carry

Say the consent clears and the call goes out with a prerecorded message to induce a purchase. What does the call itself have to do? The Telemarketing Sales Rule answers that in the prerecorded message provisions of 16 CFR Part 310, in mechanics concrete enough to design against.

• Let the phone ring at least fifteen seconds or four rings before disconnecting an unanswered call.
• Within two seconds after the completed greeting of the person called, play a prerecorded message that promptly provides the disclosures the Telemarketing Sales Rule requires for a telemarketing call.
• Follow those disclosures immediately with the opt-out announcement. On a call a consumer could answer in person, the message has to tell them they can use an automated interactive voice or keypress-activated opt-out mechanism to make a do-not-call request at any time during the message.
• Build the mechanism so that it automatically adds the number to the seller’s entity-specific do-not-call list, immediately disconnects the call once invoked, and stays available at any point in the message.
• On a call that could be answered by an answering machine or voicemail service, give a toll-free number that connects directly to an automated opt-out mechanism doing the same three things, accessible at any time throughout the campaign.

Now read that list next to a typical conversational agent spec. How does the usual agent handle an opt-out? As an intent to detect, somewhere after the greeting, after the context line, and after the first qualifying question. The rule wants something else. It wants a mechanism the person can invoke at any time during the message, announced near the top, that ends the call by itself. Those are different designs. One is a conversation feature. The other is a circuit breaker.

Separately, keep the calling-hours rule in the scheduler rather than in the playbook. 64.1200(c)(1) runs on local time at the called party’s location. On a dormant list, what are you really trusting for that? An area code that may be years out of date. Verify the time zone against something more durable than the number before the campaign starts.

Sort the Reactivation Campaign List by Consent, Not by Last Touch

Everything above collapses into one change to how the segment gets built. What should the sort key be? Not the last-touch date. Sort by what you can prove.

Three buckets. Which one does a contact go in? The lowest one it qualifies for.

• Suppress. Any revocation signal in any channel, any seller-specific do-not-call request, a registry match with no relationship and no written consent, or a number you cannot confidently attach to the person. Suppression is not a judgment call, and nothing in this bucket gets a call of any kind.
• Human dial only. The relationship is live on the shorter of the two clocks, but there is no signed agreement naming the number and authorizing an artificial voice. This is the biggest bucket on almost every real list. A person can work it. A synthetic voice cannot.
• Synthetic voice eligible. A signed written agreement exists, it names this number, it authorizes an artificial or prerecorded voice, and no revocation has landed since. You can produce the document today without asking anyone.

The fields that make this work are unglamorous, and they belong on the contact record itself rather than in whatever spreadsheet the campaign owner is keeping this quarter.

• Consent artifact ID, and the telephone number that consent actually names.
• Whether the consent text covers an artificial or prerecorded voice, stored as its own field rather than inferred from the general consent flag.
• Consent timestamp, plus revocation timestamp and the channel the revocation arrived on.
• Last transaction date, last inquiry date, and days remaining on the shorter of the two clocks.
• Entity-specific do-not-call flag and the date the number was last checked against the registry.

Build that once and the campaign query writes itself every quarter. Skip it and someone rebuilds the segment by hand each time. That is how a suppressed contact gets called twice.

Where a Human Dial Beats an AI Voice Reactivation Call

Notice what the sort produces. Which bucket is biggest? Almost always the middle one, because signed agreements that specifically name artificial voice are rare outside of businesses that already run prerecorded campaigns at scale. So a reactivation program is mostly a human-dialing problem wearing an AI headline.

That is an economics question, not a compliance one. Why do teams reach for a synthetic voice on dormant lists at all? Pickup is low and rep time per connect is high. So attack that number directly. The PowerDialer works a list in single-line or multi-line mode, so the rep is not clicking between records. AI Human Voice Detection sends live answers to the rep instead of machines and dead air. Calls, texts and outcomes log to the CRM automatically, so the disposition and the suppression flag land in a field rather than in somebody’s memory. Kixie is business calling and texting with CRM automation. It is the layer that makes a human reactivation dial affordable, and on this list that is the compliant path.

A tool is not the strategy here. The sequence is. Build the consent fields, sort into three buckets, then pick the dialing mode that fits each one. Buy the tool after the buckets exist and it removes work. Buy it before and it just dials the wrong list faster.

Some conversations should never be automated, whatever the consent record says. Complaints, cancellations in progress, anything involving a vulnerable customer, complex negotiations, and your largest accounts all start with a person. A synthetic voice is most credible when it makes one small assessment and hands off. It is least credible when it is asked to replace a process nobody has written down.

What to Measure in an AI Voice Reactivation Campaign

A booked meeting is a lagging indicator wearing a bow. Define every metric before launch, and keep the three categories below separate from each other so that a weak number in one cannot hide inside a strong number in the next.

• Reach. Attempts, valid-number rate, answer rate, and connects per rep hour. These tell you whether the list is real.
• Conversation. Meaningful conversations, qualification completion, transfers accepted by a person, and transfers that failed. These tell you whether the handoff works.
• Outcome. Reactivated contacts, callbacks, meetings booked, meetings attended, and closed revenue, each counted separately. Booked and attended are different numbers and the gap between them is the finding.

Then add the two measures this campaign type specifically needs. Why these two? Because when they move in the wrong direction they turn into a regulatory problem rather than a bad month. Track opt-outs and complaints per hundred connects, and set the pause threshold in writing before the first dial. Track revocation capture latency too, meaning the time between a person asking to be removed and the suppression field actually changing, measured against the ten business days the rule allows.

Hold out a randomized slice of the eligible list. Compare carefully matched segments if a true holdout is not available. Preserve the original source and campaign identifiers so a reactivation call does not quietly claim revenue an existing sequence created. Why bother? Because without a holdout, a reactivation campaign reports every deal it touched, and dormant lists always contain a few deals that were going to close anyway.

AI Voice Reactivation Campaign FAQs

How old can a lead be before reactivation stops being legal

There is no single cutoff, but there are two published ones. The established business relationship runs 540 days from a purchase or financial transaction and 90 days from an inquiry under the Telemarketing Sales Rule, and eighteen months and three months under the FCC rules. Past the shorter of those, a registered number needs its own written consent before a solicitation. Age alone never authorizes an artificial voice at any point on that timeline.

Does an AI voice agent need consent if a rep stays on the line

The artificial-voice rules attach to the voice that delivers the message, so a call a person actually conducts is a different analysis from a call an agent conducts. That does not make the call unregulated. The do-not-call rules still apply, the calling-hours rule still applies, and whether the dialer meets the automatic telephone dialing system definition is a separate question that depends on how it selects numbers. Decide it with counsel against your real configuration.

Should an AI voice agent leave a voicemail on a reactivation call

Treat the voicemail as a prerecorded message with its own requirements rather than as a smaller version of the call. Under the Telemarketing Sales Rule, a call that could be answered by an answering machine has to provide a toll-free number connecting to an automated opt-out mechanism that adds the number to the entity-specific do-not-call list and is reachable throughout the campaign. Keep the message short, identify the seller, and leave sensitive details out of it.

When should a human take over a reactivation call

Escalate on request, on complaint, on any question the agent cannot answer from verified data, on anything involving money already paid, and on any account large enough that you would notice losing it. Write the escalation list before the pilot and make the agent fail toward a person. A transfer that arrives with no context is worse than no transfer, so pass the disposition and the last two touches with it.

What is the best measure of a reactivation campaign

Attended meetings and closed revenue against a holdout, read next to opt-outs and complaints per hundred connects. One number without the other is how a campaign looks successful for two quarters and then costs you a list. If you only have room for one leading indicator, use transfers accepted by a person, because it moves when the list and the handoff are both working.

Scope and method. Every regulatory statement above comes from the current published text of the cited rule or from the FCC order itself, read in full rather than summarized, and the figures are quoted in the units the rule uses. This article describes federal requirements only; state law adds obligations on top of them and is not covered here. Nothing here is legal advice, and a reactivation program should be reviewed by qualified counsel against its actual list, dialer configuration and jurisdictions before the first call. Reviewed October 12, 2026.

Sources

• 47 CFR 64.1200, Delivery restrictions, U.S. Government Publishing Office. Prior express written consent for artificial or prerecorded voice telemarketing to wireless numbers at (a)(2) and to residential lines at (a)(3), revocation by any reasonable method and the ten business day deadline at (a)(10), the rebuttable presumption at (a)(11), the 8 a.m. to 9 p.m. local calling window and registry prohibition at (c), the definition of prior express written consent at (f)(9), the eighteen month and three month established business relationship windows at (f)(5), and the exclusion of established business relationship calls from telephone solicitation at (f)(15). Reviewed October 12, 2026.
• 16 CFR 310.2, Telemarketing Sales Rule definitions, U.S. Government Publishing Office. The established business relationship definition at (q), measured as 540 days from a purchase, rental, lease or financial transaction and 90 days from an inquiry or application. Reviewed October 12, 2026.
• 16 CFR 310.4, Abusive telemarketing acts or practices, U.S. Government Publishing Office. The do-not-call registry prohibition and its written agreement and established business relationship exemptions at (b)(1)(iii)(B), and the prerecorded message requirements at (b)(1)(v), including the express written agreement elements, the fifteen second or four ring minimum, the two second disclosure window, and the automated interactive voice or keypress opt-out that adds the number to the entity-specific do-not-call list and immediately disconnects. Reviewed October 12, 2026.
• Implications of Artificial Intelligence Technologies on Protecting Consumers from Unwanted Robocalls and Robotexts, Declaratory Ruling FCC 24-17, Federal Communications Commission, CG Docket No. 23-362, adopted February 2, 2024 and released February 8, 2024. Confirms that the TCPA restrictions on the use of an artificial or prerecorded voice encompass current AI technologies that generate human voices, so such calls require the prior express consent of the called party absent an emergency purpose or exemption. Reviewed October 12, 2026.

Sources verified by the Kixie Research Team on October 12, 2026.