// CASE STUDY · MARKETPLACE · $2B+ GMV
Zulily turned its operators into builders.
ZuSuite, the white-labeled Agyntx instance running inside Zulily, did not just replace four legacy tools. It made every team in the company faster, and the engineering org started shipping its own AI workflows on top of it.
legacy systems replaced
data sources unified
internal tools built by Zulily teams
ZuSuite · ZuSuite wordmark powered by Agyntx
// THE SETUP
A marketplace at scale, with 14 disconnected systems.
Zulily runs at marketplace scale. 139 million addressable email contacts. 4.7 million daily active recipients of the daily blast. $7B-plus lifetime revenue across women's, home, and footwear. The catalog turns over fast. The decisions move faster.
By early 2026 the operating reality looked like most retail at scale: BigQuery as the warehouse of record, Braze as the messaging spine, Shopify on the storefront, Tapcart on mobile, Yotpo for loyalty, plus a long tail of analytics dashboards, attribution tools, and custom scripts gluing the pieces together. The data was there. The decisions were not.
The merchandising team was opening eight tabs to plan a Monday email. The creative team was rebuilding the same brand guardrails in three different tools. The engineering org was writing one-off integrations every week that nobody had time to maintain. And the AI tools the team had layered on top were doing exactly what every operator hates: producing confident-sounding output with no source, no audit trail, and no way to trust it.
“We did not need another dashboard. We needed a brain that could read all of our systems, run our actual playbooks, and tell us when it was guessing.”
Senior leader, Zulily
// THE DEPLOYMENT
March 2026. ZuSuite goes live.
Zulily deployed Agyntx in March 2026 as a private, white-labeled instance: ZuSuite. The brand stayed Zulily-purple. The architecture became Agyntx.
Connected on day one
- +BigQuery (ZulilyBeyondData, 381M rows)
- +Braze (89M deliverable contacts)
- +Shopify storefront + order data
- +Tapcart mobile app
- +Yotpo loyalty tiers
- +Internal merchandising + pricing
Replaced in 90 days
- -A standalone customer data integration tool
- -A separate AI copywriting subscription
- -Two competing brand affinity engines
- -A dashboard subscription nobody trusted
Built net-new on top
- *Push notification engine (6 options/send, scored)
- *Brand Planner workflow
- *Merchandising Workbench
- *Perfect Order Dashboard
- *Pulse-sync pipeline into Braze
ZuSuite · Architecture diagram: BigQuery, Braze, Shopify, Tapcart, Yotpo flowing through ZuSuite/Agyntx
// THE UNLOCK
The result nobody saw coming.
Once the platform was live, Zulily's own teams started building on it. Engineering did not have to commission ZuSuite v2 from a vendor. Merchandising did not have to wait six months for a tool. Creative did not have to file a ticket. They built their own tools.
Today, more than a dozen internal applications run on top of ZuSuite. Some are simple: a daily standup digest for the merch leads. Some are more ambitious: a vendor CRM rebuild that addresses years of fragmented vendor, brand, and supplier data. The point is not the count. The point is that the same operators who used to wait for software now ship it.
“The week ZuSuite went live, our team stopped asking IT for tools and started building them. That is the cultural shift. The technology is the floor.”
Engineering leader, Zulily
“I get six push notification options before my coffee finishes brewing. Each one is on brand, each one is scored, and I know which one is a guess. I have not opened our old tool in two months.”
Marketing operator, Zulily
ZuSuite · Push notification generator UI, 6-option grid with brand voice scores
ZuSuite · Merchandising Workbench: live margin, sell-through, brand affinity
// THE PROOF POINTS
First 120 days.
lift in email-attributed revenue per quarter
push notification production time, down from 45 minutes
legacy SaaS subscriptions eliminated
internal tools shipped by Zulily teams in Q1
The AI Confidence Protocol in production
Every output generated by ZuSuite ships with a verified, modeled, or inferred breakdown. When ZuSuite is sure, the team moves fast. When ZuSuite is guessing, the team knows to verify. No more confident-wrong AI.
ZuSuite · AI Confidence Protocol output sample with verified/modeled/inferred ring
“Other AI tools tell you what they think you want to hear. Agyntx tells you when it is not sure. That is the entire difference. That is why our team trusts it.”
VP, AI-Driven Commerce, Zulily
A platform pays for itself when it replaces tools. It changes a company when it lets the operators become builders. ZuSuite did both.
If you run a marketplace, a retailer, or a brand at scale and you want to white-label Agyntx for your team,
book a call →Read next
How 27aDay built a fully autonomous ecommerce business on Agyntx →