XS · S · M · L · XL · 2XL
OpsAgents AI
OpsAgents AI
17 brands, some 830 stores, six Shopify storefronts, and a paid loyalty club holding a full purchase history for every member. Four capabilities that can run on those assets — all of them live today, not a wish list.
Maya is already running on the DREAM CARD surface. The captures below are from the live demo — the same member area, the same promotions out of the club’s CMS, with Maya handling the entire interaction in chat without sending anyone to a form or another page.
Valid for 2 more daysMaya identifies the member by name, reads her active benefits, works out which one expires first, shows the promotion card together with its terms — and asks to redeem. All inside the conversation.
Once Maya reads the purchase history she doesn’t just flag an expiring benefit — she knows which brands are genuinely yours and what is worth offering right now. That is what replaces “My Top 5”: the customer stops picking manually, because the system already knows.
Maya runs on any site, whatever it is built on — on the club member area (Angular) and on Terminal X (Magento) alike, as an embedded widget. On Shopify she additionally has a dedicated integration we built: live catalogue access, cart read and write, and sales optimisation — six of the group’s storefronts stand to benefit. Hebrew-first, a WhatsApp channel, and handoff to a human agent when needed.
In the member area, “My Top 5” asks the member to manually pick which five brands are shown to her first — while her purchase history, by brand, date and amount, sits in that very same system.
The capture on the left is from the club’s live chat today. In one of the app reviews a customer wrote that there is no way to reach a person — this is the screen she was looking at.
The club system already knows how to rank brands by purchase history, and the interface already has a slot for showing personal content to each member. Both components exist — nobody has put a brain behind them yet.
Those three lines are a single description of the sales agent missing from the interface — and she is already running, above.
On the left is the photo live on the product page today. On the right is that exact same item, produced today in OpsAgents AI Studio on our house models, from that same image. Not stock, not a reshoot.
The difference is not only quality. Production runs item by item at a workstation — a brand uploads one item and gets a result — instead of stockpiling dozens of items for a single shoot day, coordinating models and a studio, and waiting weeks. And it costs less.
The same mechanism also produces video, and holds a consistent model across all of a brand’s collections.
Six of the group’s storefronts already run on Shopify. What the partnership means in practice: we do not bolt onto a store from the outside — we work through Shopify’s official AI interfaces, per storefront.
| Brand | Storefront | Platform | Route |
|---|---|---|---|
| FOX | fox.co.il | Shopify | App install |
| FOX HOME | foxhome.co.il | Shopify | App install |
| Foot Locker | footlocker.co.il | Shopify | App install |
| Laline | laline.co.il | Shopify | App install |
| Shilav | shilav.co.il | Shopify | App install |
| Itay Brands | itaybrands.co.il | Shopify | App install |
| Terminal X (+7 brands) | terminalx.com | Magento | Embedded widget |
| DreamCard club | online.dreamcard.co.il | Angular | Widget in the member area |
17 brands running rotating promotions need a landing page per campaign. Instead of queueing with the digital team, you write one line about the campaign and a team of seven agents goes to work: research, layout and art, copy, AEO/LLMEO and SEO, monitoring, and final assembly — while publishing itself stays behind a human approval gate.
Buy now before it’s goneSoftness in the bedroom and style in the kitchen: the FOX HOME end-of-season sale is in full swing, with up to 70% off everything your home loves. Don’t leave it to the last moment.
Shop online
The headline, the copy and the club block are genuine output from the run. The logo and product images were composited in here to show the destination — wiring the builder to the catalogue and brand assets is the next step, not an existing capability.
Without being asked, the team built a block explaining to the visitor that she earns and redeems DREAM CARD points during the SALE as well. That is the link between a brand campaign and the club — automatically.
The publishing agent stayed at “pending human approval”. The run log records: “the page is ready but has not been published”. An agent does not put a page live without approval.
The design studio we built and operate: you write a brief in chat and get back a full design direction, product sketches from several angles, and a gallery — driven by a dedicated design-director agent. The screens below are a real project we ran for one of the group’s brands: Disney-inspired girls’ shoes.
The same engine produces vertical video for TikTok and Reels as well as stills for the feed. The three Reels below were made from the group’s own items — from the same Studio renders you saw above, animated into video. No shoot, no set, no models. The wide film and the Hebrew-voiceover clip come from another campaign already approved by a client, and demonstrate the full production line.
The format for the site, YouTube and in-store screens. Every frame here was generated in the system — no shoot, no set, no models. Playing on loop; tap the speaker for sound.
9:16 format, short cuts, a hook in the first two seconds — the format that works on TikTok and Reels in the Israeli market.
A system-generated character speaking natural Hebrew over an original soundtrack: “Whoa, whoa, whoa — okay, stop everything. Are you seeing this? Limited edition — find a store near you.” No actress, no voice artist, no recording studio and no music licensing. From a client-approved campaign.
Not just an image — a finished post: the crop, the promo sticker, Hebrew copy and hashtags. These are exactly the files Studio produced earlier, with no additional production. One of them is a Reel.
The videos were produced from product images and system-generated models, including original voice and music. The same production line works on any catalogue.
Today a customer does not always type into Google. She asks an AI assistant: “which fashion club in Israel gives cashback?”, “where should I buy bedding on sale?”. Whoever the engine understands gets named in the answer. Whoever it doesn’t simply isn’t there. This is a new field, and among Israeli retail groups it is almost empty.
Content structure, structured data and explicit answers to the questions customers actually ask, so the engine can quote you directly.
Assets language models can read: llms.txt, brand and catalogue descriptions in language the model interprets correctly, and crawl permissions that are set up properly.
Speed, tags, linking and structure. Still important, but no longer on its own — it is the bottom layer, not the whole story.
The campaign-page builder runs an AEO/SEO audit inside the process — on the FOX HOME run the score went from 94 to 100 before the page was even shown. This is not separate work, it is part of production.
A measurable starting point: we check how each of the 17 brands is presented today in AI-assistant answers, fix the underlying infrastructure, and report the change every month.