<- all tokdocs
2026.09.22@carmel_levytan

The 82.76% show rate in this clip is close to the published average, and the sequence it credits is signed by a different co-founder

Marketing
Watch on TikTok

View on TikTok ->

This is a 26-second vertical clip at 576x1024 (H.264 video, AAC audio, 335 kbps) posted on 2026-09-22 by @carmel_levytan under the channel nickname "Carmel Sells AI," scored to an original sound credited to Carmel Sells AI. At capture it held 2,719 views, 150 likes, 31 comments, 109 saves and 29 shares, and carried the hashtags #aiagency, #aireceptionist, #aiagent, #voiceai and #aiconsulting. I read all 13 frames at two-second intervals and 76 transcript words. Frames 1 through 4 show the speaker talking to camera in a cream-walled room with two framed coffee prints behind her and a mason jar with a glass straw in the foreground, under a burned-in caption reading "My pre-call system that gets our AI agency an 82% sales call show rate." Frame 2 adds a black-and-cyan stat tile across her chest: "Show Rate" on the first line, "(Avg)" in italics below it, and "82.76%" on a cyan bar. Frames 5 and 6 cut to a stack of five overlapping email screenshots, each sent from "Eyal @ ufound AI," with recipient addresses and client names blacked out. The subject lines, top to bottom, read "What we actually do (in plain English)" (06:05 AM), "Why most people fail with cheap AI receptionists" (11:10 AM), "The honest worst-case scenario" (09:04 AM), "'We want a human answering the phone...'" (11:05 AM) and "Quick recap (so you know what you're walking into)" (02:05 PM). Email one pitches "a 24/7 custom built AI dispatcher that helps residential home service owners get more booked jobs without hiring more CSRs, spending more on ads, or losing leads to voicemail" and claims a redacted client "added an extra 150 jobs in 6 months" while another "captured 133 new leads in 2 months." Email two states "The first 30 days usually looks like: the phone gets answered 100% of the time, you see a 75% booking rate, and a 90% call completion rate - not dropped, not half-handled, not 'we'll call you back'." Email three frames the downside as "You spend three months with us, and you only get 2-4 additional job bookings per week, and you see a full return on investment." Email four argues "It's voicemail vs AI - because the human was unavailable, it was after hours, or they were already on another call." Email five closes with "If you're showing up, reply with YES so I know to prep the right materials," signed "-- Eyal." Frame 7 swaps in a messaging thread carrying five outbound bubbles: "Hey John, it's Eyal. Appreciate you making time. Reply STOP and I'll stick to email. Thanks, ufound AI" (02:03 PM), "Our best clients watch this before the call: https:" with the URL blacked out (04:04 PM), a "Jul 16" date divider, "Tomorrow we'll cover how to book more jobs without spending more on marketing. No prep needed." (10:30 AM), a "Jul 17" divider, "Still good for later? Bring an estimate of your daily call/lead volume." (07:31 AM), and "I'm on the google meet now, join when ready:" with the link blacked out, "Laptop or phone works." (10:25 AM). Frames 8 through 13 return to the talking head with no overlay while she delivers the closing line, "so save this and just copy it."

The 82.76% figure is roughly average for booked sales meetings, so it cannot carry the causal claim made over it

The voiceover opens with "My AI agency has an 82% show rate on our sales calls and it's because of this pre-call system that we built out." The on-screen tile in frame 2 reads 82.76%, which rounds to the 82% she says, so the clip is internally consistent on the number itself.

The framing around it does not survive a benchmark check. RevenueHero published show-rate data drawn from 18 weeks of meeting conversion records across customers booking 50 or more meetings a month, reporting an average no-show rate of 15.9% and a median of 13.5%. Those translate to an average show rate near 84.1% and a median near 86.5%. The same dataset puts the top 10% at a 5.5% no-show rate, which is a 94.5% show rate. Measured against that set, 82.76% lands slightly below the average and well short of the top decile.

The tile also carries no scaffolding. It shows no tool name, no date range, no denominator and no comparison period. A system credited with lifting show rate needs a before number, and no frame contains one. The strongest supportable reading of frames 1 and 2 is that the agency's average show rate was 82.76% over an unstated window, not that the pre-call sequence produced it.

The counts in the voiceover match the frames exactly, and the sender does not

The transcript says the system "sends out these five emails and this text sequence of five text messages." Frames 5 and 6 contain exactly five distinct email panels with five distinct subject lines and five distinct timestamps. Frame 7 contains exactly five outbound message bubbles across two date dividers, Jul 16 and Jul 17. Both counts check out against the pixels.

The attribution does not. Every email header in frames 5 and 6 reads "Eyal @ ufound AI," the first SMS in frame 7 opens "Hey John, it's Eyal," and the last email signs off "-- Eyal." ufound AI's own site lists three co-founders: Eyal Benzaquen Gabriel on sales, Carmel Levytan on marketing, and Yoav Benzaquen Gabriel on technology. The speaker is Carmel Levytan, the marketing co-founder, describing a sequence that runs under her sales co-founder's name. "My AI agency" in the voiceover compresses a three-founder company into one, and the sequence on screen is not sent by the person narrating it.

The company name is also lowercase in the frames and on the company's own site. It is written "ufound AI," not "UFound" or "Ufound."

The performance numbers inside the emails exceed what ufound AI publishes about itself

Email two tells the prospect that in the first 30 days "the phone gets answered 100% of the time, you see a 75% booking rate, and a 90% call completion rate." That 75% booking rate has no counterpart anywhere in ufound AI's public material.

ufound AI's Retell Certified Partner profile, which the company controls, states that its outbound sales agents hit "a 30% answer rate and 20% booking rate" and that inbound receptionists deliver "an average 464% ROI." The 20% figure covers outbound and the email's 75% covers inbound, so the two are not strictly contradictory, and that is the problem: the only booking rate ufound AI publishes with a number attached is the outbound one, and it is 20%. ufound AI's booking page leads with "Add 15-30 New Booked Jobs This Month" and "Only 5 spots available this week," and states no booking rate, answer rate or ROI percentage at all. Pricing on that page is described as fully custom and is never shown.

The homepage carries animated counters rather than static figures. During my fetch they read between 1,372 and 1,461 jobs booked, between $939K and $1M+ in revenue generated, and between 28 and 30+ home service partners. Named case-study clients on that page include Preferred Plumbing Solutions, Advanced Plumbing & Drains, Kraun Electric, Green Unlimited Lawn & Pest, Cleveland Bee Removal, Belmont Solar, SWAT Wildlife Inc., Northern Snow, Union Capital Realty, NIU Clinic and Reliable Transportation Link. The two client results quoted in email one, "an extra 150 jobs in 6 months" and "133 new leads in 2 months," have the client names redacted in the frames, so neither can be matched to a named case study or checked.

The listed integrations on the homepage are ServiceTitan, Housecall Pro, JobNimbus, ServiceM8, Service Fusion, HighLevel and Google Calendar. HighLevel being on that list matters for anyone copying this, because the conversation view in frame 7 with its right-aligned outbound bubbles, calendar-icon date dividers and per-message overflow menus is consistent with a HighLevel conversations pane rather than a native phone messages app.

The SMS leg carries consent and registration obligations the clip never mentions

The first text in frame 7 ends "Reply STOP and I'll stick to email. Thanks, ufound AI." That is two of the required pieces. CTIA's Messaging Principles and Best Practices states that "Standardized 'STOP' wording should be used for opt-out instructions" and that senders "should state in the message how and what words effect an opt-out," which this message does, and it identifies the sender.

The same document says an opt-in confirmation message "should include: (1) the program name or product description; (2) customer care contact information (e.g., a toll-free number, 10-digit telephone number, or HELP command instructions); (3) how to opt-out; (4) a disclosure that the messages are recurring and the frequency of the messaging." No frame shows a frequency disclosure, a HELP instruction or a care contact in the SMS thread. CTIA also calls for support of "multiple mechanisms of opt-out, including phone call, email, or text," while the bubble on screen offers only the STOP reply.

Consent is captured upstream rather than in the sequence. The ufound AI booking form collects business ownership confirmation, yearly revenue bracket, daily call volume, name, business email, mobile phone, website, and an SMS consent checkbox. That checkbox is the consent event that makes the text leg legal, and the clip never mentions it. Copying the five texts without copying the checkbox copies the liability and not the permission.

Two legal deadlines also apply that the clip omits. The FCC's consent revocation rule took effect on April 11, 2025, requiring senders to accept revocation "in any reasonable manner," listing "STOP," "QUIT," "END," "REVOKE," "OPT-OUT," "CANCEL" and "UNSUBSCRIBE" as examples, and to honor requests within ten business days. Separately, ufound AI's co-founders are based in the Greater Toronto Area and the booking page describes serving home service businesses across Canada, which puts the email and SMS legs under Canada's Anti-Spam Legislation. Innovation, Science and Economic Development Canada's CASL portal confirms that businesses must obtain consent before sending commercial electronic messages and that text message and social platform messaging fall in scope.

The "comment pre call" call to action depends on an automation path whose availability I could not confirm

The description reads "Comment or DM me 'pre call' and I'Il send this to you in full," including the typo "I'Il" in TikTok's own description text. The comment half and the DM half are not equally automatable.

Manychat's TikTok automation guide claims comment-to-DM flows are available for TikTok Business accounts and points readers to a help-center article titled "Set up your first TikTok automation in Manychat (User sends a message)," which is a message trigger and not a comment trigger. I could not resolve the conflict against primary sources. Manychat's help center returned HTTP 403 on both the trigger overview and the setup guide, and TikTok's Business Messaging API reference portal returned only a page title with no endpoint list. Multiple third-party writeups state the TikTok comment trigger is a gated partial rollout limited to a small number of markets, but I have not verified that against TikTok's own documentation and I am not going to repeat it as fact.

What is checkable is the funnel's response shape. The CTA asks for comments, and the metadata shows 109 saves against 31 comments at capture, so saves outran the requested action by roughly three and a half to one. The closing line of the voiceover, "so save this and just copy it," asks for the save, and the save is the action viewers took. A save produces no contact record and no DM thread, so the majority of the clip's engagement generates nothing the agency can follow up on. The 31 comments are the only part of the 2,719 views that converts into a conversation, and that is a 1.1% comment rate.

Key Takeaways

  • Partial correction: The voiceover's "82% show rate" matches the 82.76% tile in frame 2, and that figure is near the published average rather than above it. RevenueHero's 18-week dataset puts average no-show at 15.9% and median at 13.5%, which are show rates of roughly 84.1% and 86.5%.
  • Correction: The clip says "my AI agency." ufound AI publishes three co-founders on its own site, and every email and text in the frames is sent and signed by Eyal Benzaquen Gabriel, the sales co-founder, not by the speaker.
  • Verified: The counts hold. Frames 5 and 6 contain exactly five email subject lines and frame 7 contains exactly five outbound SMS bubbles, matching "five emails" and "five text messages" in the transcript.
  • Correction: Email two's "75% booking rate" has no backing in ufound AI's public material. The only booking rate the company publishes with a number is 20%, on its Retell partner profile, and that one covers outbound agents.
  • Unverified: The two client results in email one, "an extra 150 jobs in 6 months" and "133 new leads in 2 months," have the client names blacked out in the frames and cannot be matched to any of the eleven named case studies on ufound AI's homepage.
  • Unverified: Whether a TikTok comment can automatically trigger a DM for a North American account. Manychat's blog says yes, its help center and TikTok's API reference were both unreachable, and I will not assert either answer.
  • Unstated cost: US senders need A2P 10DLC registration before any of this runs. HighLevel's fee schedule lists $22.50 one-time for a Sole Proprietor or Low Volume Standard brand, $64.00 for High Volume Standard, $15.00 per additional campaign, and $1.50 to $10.00 per month depending on use case.
  • Unstated cost: The consent checkbox on ufound AI's booking form is what makes the SMS leg lawful. The clip tells viewers to copy the messages and never mentions the checkbox, the frequency disclosure, the HELP instruction, the ten-business-day revocation window under the FCC rule effective April 11, 2025, or CASL coverage for a Canada-based sender.
  • Unstated cost: The AI dispatcher being sold inside these emails runs on Retell. Retell's published pay-as-you-go range is $0.07 to $0.31 per minute, assembled from $0.055 per minute of voice infrastructure, $0.015 per minute for standard text to speech, a per-minute LLM charge, and $0.015 per minute of US telephony, plus $2.00 per phone number per month.
  • Unstated: The description claims "a 30%+ close rate" while the voiceover and the only on-screen tile cover show rate. No frame contains evidence for the close rate.

Resources

Published September 22, 2026. Writeup generated from a favorited TikTok.