Would you trust an AI to add items to a shopper’s cart?
Hey Hunter, Sheraz here (CTO/Co-founder). I’m exploring a shift from “support chatbots” to a sales agent that actually helps shoppers decide and places items in the cart when they say yes.
What we’re testing that’s different:
Decisioning, not deflection: compare two products in-chat with real specs/stock.
Zero training ingest: it learns catalog, price, and inventory automatically.
Timed nudges: smart bundles/discounts only at the moment of intent.
One outcome: add to cart right inside chat + clear funnel/AOV analytics.
I’d love your take on three things:
Boundaries: where should an AI stop (e.g., add-to-cart ok, checkout no)?
Trust signals: what proof would you need (copy, receipts, guardrails, undo)?
Onboarding bar: if this took <10 minutes to connect, would you try it?
If helpful, here’s a 30–45s demo clip:
I’ll reply to every comment today. If you run a Shopify/Woo store and want a private walkthrough, drop your URL.
Replies
if it’s <10 min via Shopify app + auto-ingest, I’m in. need an A/B kill-switch + easy rollback.
happy to test on a demo store FYI.
@shabb_katoch Absolutely—yes to all three:
Install: <10 min via Shopify app, auto-ingest products/variants/price/stock.
A/B: built-in split (0/25/50/100), sticky buckets, instant kill-switch.
Rollback: one click back to your current chat; clean uninstall; action logs kept.
Similar to vibe banking idea, it's beyond crazy. Me would not trust AI shopping for me at all and that decision has to be made from my own, millions of successful demo would not persuade me :)
@rickyguo Totally fair, Ricky. It’s human to resist new tech—most of us felt that way about online payments at first—but over time clear value + safeguards win. Our line is assist-only: add to cart on explicit “yes,” never checkout, with mini-receipt and undo. Would a simple reorder-only mode ever feel okay, or still a hard no?
@sheraz_abdul_hayee Graet idea to have a simulated mini-receipt without actually checkout and I guess would be a yes if the checkout process is still managed by human.
@rickyguo Thanks excited to share the prototype next week with you!!!
Really interesting. I'm a shopper that adds an abundance of items to my cart while browsing, then removes several of them prior to checking out -- almost using the cart as a mini favorites list. I'd be into this if AI-added items truly match the parameters of my current add-to-cart pattern. Otherwise it just feels like an annoying ad; might as well be a pop-up. Example: if I'm adding running shoes in the $50-$100 range, size 8, in black -- don't add boots size 8, or runners $200, etc.
Even more -- if there were too many slop items added, it's entirely possible I'd abandon cart rather than deal with removing them all.
@leah_madden really appreciate the thoughtful comment “cart as a working shortlist” is spot on. We design to avoid slop: same-category only, exact size/color, merchant-set price band (e.g., $50–$100), act only on clear intent, max one agent add per session, never touch existing items, and show a mini-receipt in chat with Undo + “Why this?”.
Curious, if we match your parameters perfectly, would you prefer the agent to replace a worse match already in your cart or only add alongside?
@sheraz_abdul_hayee At this stage in my comfort with this type of feature, I wouldn't want the agent replacing a product. But assuming best-case scenario for the recommendations, and I'm able to bank a few shopping experiences where the agent is truly adding value, I could see evolving to accept more from the agent.
If your vision is that eventually I can communicate to the agent: "I need 3 pairs of pants, 2 shirts, and a pair of boots" and it populates my cart with best-matches and swaps -- I'm so here for that!
@leah_madden Thanks for showing great interest. I am sure our upcoming demo video is going to answer some of your queries and may even want to try it!
This is really interesting! I’d be okay with add-to-cart, but probably not full checkout — having an undo or confirmation step would help build trust. Seeing clear pricing, discounts, and an easy way to review the cart would make me comfortable.
@viktorgems Thanks for the feedback, we are launching again on product in upcoming days. Stay tuned and dont forget to follow us.
DeepTagger
Add-to-cart feels fine, but checkout should stay human-approved. A quick "undo" or mini receipt in chat would build trust, as it keeps things helpful rather than pushy. Curious though, how are you thinking about nudges vs. being too aggressive?
@talshyn totally with you.
Boundary: AI stops at add-to-cart; checkout is always human-approved but it can assist in filling the details if human approves it.
Trust: instant mini-receipt in chat (item/price/size/returns) + undo window (e.g., 30 min) and an action log.
Nudges vs. pushy: In our tests, the first nudge (on a clear intent signal) drives most of the gain; a second one spikes dismissals and bounce.
It keeps the experience predictable: help once, then go quiet after a “no.”
It’s configurable: you can set 0–2 nudges, per-page/collection rules, and a cooldown (e.g., 24h or until new intent).
Triforce Todos
If it’s really a 10-minute setup, I’m sold, just curious if it’s self-serve or dev-heavy.
@abod_rehman Its a plug and play system, from Zanderio setup, integration to your store and product ingestion all done under 10 minutes that how seamless our product is. You will be able to test its feature soon!
absolutely. I do not see a reason not to add
@lucia_fernandez2 Great you are gonna love the prototype then!
Interesting concept! I would like to know "why" behind each suggestion, + set spending limits.