What 55 DTC brands actually run for CX

We wanted to know what mid-market DTC brands are actually running for customer service. Not what vendor case studies claim, not what shows up on a G2 grid. The tools loading on the storefront right now. So we checked 55 of them across helpdesk, email, returns, subscription, reviews, and the AI layer, then overlaid what each brand’s CX lead is on record saying about AI and human service.

TL;DR

The core is universal. Everything else fragments. Shopify (98%), Klaviyo (93%), and Gorgias (62%) are table stakes. After that, there’s no winner in any category - reviews, returns, subscriptions are all contested, with the top two vendors usually within a few points of each other and a long tail behind.

Size matters fast. The Gorgias-Zendesk conversation starts at 50 employees. At small (1–25), Gorgias is at 67%. At mid (26–100), it’s 47% and Zendesk enters at 20%. Same for reviews - Yotpo leads at small, Okendo overtakes at mid.

Three operator stances exist, and they’re stable. Hard-line human-first (no deflection AI). Pro-AI-with-guardrails (AI for back-office, human up front). Tried it and adjusted (deployed customer-facing AI, now recalibrating). The data shows all three stances represented across the sample, and operators within each group use remarkably similar stacks.

Deflection AI is visible at 1 in 8 brands, concentrated mid-band. 13% of brands have a customer-facing AI chat loading on their storefront. Only 1 of 30 small-band brands runs one. 4 of 15 mid-band brands do. Deflection-AI experimentation is concentrated where ticket volume justifies the integration cost, not at the top of the segment and not at the bottom.

---

The headline numbers

The single biggest surprise wasn’t what leads - it was how quickly it fragments.

Category	Dominant tool	% of brands	Runner-up	% of brands
Commerce platform	Shopify	98%	(none)	–
Email / SMS	Klaviyo	93%	Attentive	24%
Helpdesk	Gorgias	62%	Zendesk	9%
Reviews / loyalty	Okendo	44%	Yotpo	42%
Returns	Loop Returns	25%	Redo	20%
Subscription	Recharge	29%	Skio	4%
Shipment tracking	Route	9%	AfterShip	2%
Fraud / risk	NoFraud	9%	Signifyd	2%
Deflection AI	Zowie	4%	Long tail (no clear #2)	9%

Shopify and Klaviyo are effectively the price of entry. Gorgias is the modal helpdesk but it does not dominate the way Klaviyo dominates email - about a third of brands run something else, with Zendesk the most common alternative and a long tail of Help Scout, Richpanel, HubSpot Service, Intercom, and Re:amaze beneath.

Then the surprises start.

Reviews and returns are the most contested categories

The reviews category is split nearly even. Okendo appeared on 44% of brands, Yotpo on 42%, and the next four tools (Stamped, Reviews.io, Junip, Trustpilot) appeared on between 5% and 16%. Several brands ran two reviews tools simultaneously, usually one for on-site product reviews and one for cross-site trust display. There is no winner.

Returns shows the same shape. Loop Returns appeared on 25% of brands and Redo on 20%, with Happy Returns, ReturnGO, and Returnly behind. Notably, several brands had both Loop and Redo present on the same site. That is a strong tell of an in-flight returns-tool migration - usually Loop being replaced by Redo, sometimes the other way round. If you were sceptical that returns tooling churn was real, this is what the inside of it looks like.

The size cut: Klaviyo holds, Gorgias drops, Zendesk creeps in

We sized 47 of 55 brands; the rest didn’t have a usable team-size signal.

Small

1–25 emp · n=30

Mid

26–100 emp · n=15

Shopify

100%

93%

Klaviyo

97%

100%

Gorgias

67%

47%

Yotpo

57%

20%

Okendo

43%

33%

Recharge

33%

33%

Attentive

30%

27%

Loop Returns

23%

20%

Redo

23%

13%

Stamped

17%

13%

Zendesk

0%

20%

Reviews.io

3%

20%

Source: Handsom independent tech-detection sweep, May 2026

The pattern reads off the heatmap. Klaviyo gets more universal as brands grow. Gorgias drops 20 points and Zendesk creeps in from nowhere to match Loop Returns and Yotpo. By the mid-band the helpdesk decision is genuinely contested. Reviews tooling flips between bands - at the small end Yotpo leads with a 14-point margin, at the mid end Okendo overtakes. Subscription stays roughly flat in penetration - about a third of brands have a subscription product at all, regardless of size, and where they do they’re overwhelmingly on Recharge.

This is the pattern an operator should sanity-check their own stack against. If you are running Gorgias and your brand is hiring its 60th employee, you are already in the cohort where the next-stack-up conversation is happening. Whether you initiate it or wait for your CX manager to push back is a separable question.

Deflection AI: visible at 1 in 8 brands, concentrated mid-band

The category that gets the most vendor attention - customer-facing deflection AI - has a smaller and more lopsided storefront footprint than vendor marketing implies.

The concentration matters more than the headline. Of the 7 brands running visible AI chat, 4 are in the 26-100 employee mid-band, 1 is in the 1-25 small band, and 2 are unsized. Translated: deflection-AI experimentation is happening where ticket volume justifies the integration cost. 1 of 30 small brands runs it; 4 of 15 mid-band brands do, an 8× concentration. That’s a sharper finding than “everyone is or isn’t running deflection.”

3%

Small · n=30

27%

Mid · n=15

Source: Handsom independent sweep + manual verification, May 2026

But the fact that the visible-storefront footprint sits at 13%, not 50%, not 80%, at this point in the vendor cycle is itself a finding. The narrative that customer-facing AI is universal at mid-market DTC is not what the storefront says. What the storefront says is that mid-market DTC has standardised on a particular Shopify-plus-Klaviyo-plus-Gorgias core, and the dedicated deflection-AI layer is something the segment is sampling at mid-band, not deploying at scale.

What the people on the ground say

The qualitative half of the dataset is the operators’ own words about AI and human service. Three groups emerged.

The human first group

About a quarter of the brands in our sample take an explicit human-first posture on customer-facing AI. KREWE, Levain Bakery, Gunner Kennels, Caskata, Live Bearded, Dr. Killigan’s, and several others. The framing is consistent.

I don’t think you’ll ever call our customer service number and get a robot. It’s not our vibe.

For KREWE, we really want to maintain a hyper-personal level of customer service. We want to make sure that someone is talking to a voice, a real voice on the other end.

Avra Wacks, Senior Manager of Customer Experience at Levain Bakery, described the team’s hiring philosophy in similar terms: “we don’t outsource - we’re people who were hired with the shared mission of giving our customers the highest and most personal level of support possible.” Dr. Killigan’s framing: “when you’re dealing with a computer, it’s more of a transactional experience rather than the relationship-building that CX is about.”

Stack pattern across this group: Gorgias modal but not universal, Klaviyo universal, reviews tooling present, returns tooling present, deflection AI absent.

The deeply integrated AI group

A roughly equal-sized group is actively using AI, but with explicit guardrails on what the AI is allowed to do. The framing is “AI for the parts customers don’t notice” - warranty triage, ticket classification, post-purchase follow-up, internal summarisation. The customer-facing chat is still human.

Making sure that any AI we use is giving people appropriate responses and gives them the same level of service as my human agents. I want a team member who’s going to be able to go outside policy when it makes sense.

We’ve implemented AI to respond to those warranty customers by looking at the product photo and then capturing certain fields based on product issue.

Mugsy: “Our aim is to be able to fully integrate AI in a way that it doesn’t take away from the experience.” Ivy City Co: “if you can use an AI tool that can mimic your brand voice and the kind of empathy that you want to portray to your customers, then it’s an excellent tool.”

Stack pattern: same Shopify-Klaviyo-Gorgias core, with the AI layer either sitting at the helpdesk (Gorgias native), at the warranty/returns workflow (often custom), or in the BI layer (Triple Whale, Northbeam).

The “tried it” group

A smaller but instructive group is brands that did deploy customer-facing AI and are now adjusting. The clearest example in our sample is Happy Mammoth, who runs Zowie on their storefront, and whose CX lead is also on record saying:

Because we put AI everywhere, customers are starting to need human touch. They want to speak with a human.

That is the same brand running customer-facing AI. The contradiction is the point. They are not removing the AI - they are recalibrating where the human-handoff lives. This is the operational shape of the broader reassessment most of the industry is now beginning, captured in one brand’s own words.

Groove Life - also running Zowie - uses similar language in their CXplained interview: “I’m 100% on board with scaling AI, but not just letting it go. AI has its place and will be helpful moving forward in an innovative way.” The framing across this group is portfolio-refinement, not reversal.

What the data suggests for your next stack decision

If you’re picking your first stack (1-15 employees)

Default to Shopify + Klaviyo + Gorgias. 100% of small brands run Shopify, 97% Klaviyo, 67% Gorgias. The cost of being non-standard at this scale is real - smaller hiring pool, more brittle integrations, more expensive agency partners. Save your deviation budget for places that matter.

For reviews: Yotpo (57% plurality at small) or Okendo (43%) - either works. Skip dedicated customer-facing AI for now; only 1 of 30 small brands in our sample runs one, and the integration cost rarely pays back at sub-1K tickets per week.

If you’re growing through 50-100 employees

The Gorgias-to-Zendesk conversation is coming. Zendesk shows up at 20% of mid-band brands versus 0% at small. Map your CX team growth over the next 18-24 months before your next Gorgias renewal - switching helpdesks under renewal pressure is the most expensive stack mistake we see.

Reviews flips at this band: Okendo edges Yotpo (33% vs 20%). If you’re on Yotpo and considering a switch, the dual-tool brands in our sample suggest the migration is doable but real.

If you’re considering customer-facing AI chat

Where you are in the size cut matters. 3% of small brands run one. 27% of mid-band brands do. If you’re at mid, your peer experimentation is real and this is a fair time to evaluate. Below 25 employees, the integration overhead probably outweighs the economics.

Two paths worth knowing about: a dedicated vendor (Zowie and Forethought were the ones visible in our sample), or your existing helpdesk’s native AI (Gorgias AI add-on, Zendesk Answer Bot, Intercom Fin). The native option is cheaper to experiment with and easier to roll back. Start there unless you have a specific reason not to.

If you’re picking a returns vendor

Loop Returns (25%) and Redo (20%) are roughly tied. Several brands in our sample ran both simultaneously - the visible tell of a mid-flight migration. If you’re picking fresh, ask both vendors for cases at your size band specifically, and ask each one about the other. Two-vendor demos tell you something the single-vendor pitch hides.

If you’re adding subscription tooling

About a third of brands in our sample run subscription tooling, and where they do it’s overwhelmingly Recharge (29% of sample, vs 4% Skio). Default to Recharge unless you have a specific operational reason - Skio, Stay AI, and Smartrr are real alternatives but most brands at this scale don’t have the volume to justify the deviation cost.

---

How we built the sample

We took the CXplained interview corpus as a starting list of named mid-market DTC brands with a publicly identified CX leader. We pulled 60 brands from it, then did tech-detection across each brand’s storefront, contact page, help page, order-tracking page, and a product page. We matched the raw HTML against a vendor-signature library covering 50+ tools across 11 categories, and probed helpdesk subdomains (help.brand.com, *.zendesk.com, *.gorgias.help) separately. For each brand we also pulled the verbatim CX-philosophy quote from their interview, and estimated team size from a combination of the interview, the brand’s about page, and LinkedIn where it was accessible.

Five brands fell out (bot-blocked or headless storefronts), and we worked with the remaining 55. We sized 47 of those; 8 didn’t have a usable team-size signal.

The detection method picks up tools that load on the storefront, plus helpdesks routed through a subdomain. It does not pick up purely back-office tools that never touch the customer-facing surface, so true prevalence on the back-office side is almost certainly higher than what we report. Confidence per tool detection is rated HIGH (matched on official vendor domain or unique widget pattern) or MEDIUM (matched on generic string). All prevalence figures here count HIGH+MEDIUM combined; HIGH-only would shift each number a few points down without changing the modal-stack picture.

Sources

• Quoted operators. All on-record quotes are sourced to the CXplained interview series: Addison Edmonds, Gunner Kennels; Megan McGarvey, KREWE; Avra Wacks, Levain Bakery; Aaron Herr, Dropps; Dawn Wildes, HydraPak; Julia Ralaimihoatra, Happy Mammoth; Dr. Killigan’s; Mugsy; Ivy City Co; Groove Life.
• Stack composition data. Independent tech-detection sweep run by Handsom, May 2026. Per-brand data available on request.
• CXplained corpus. Interviews with Customer Service Leaders and Interviews with Founders.