What we do
Draft replies to order status and returns tickets, ready for your team to review.
AI agents for retail and ecommerce brands. The agent does the lookups and drafts the work, and a person on your team approves anything that moves money or changes a listing.
See a Monday morning inboxWhat it costs
$2,500five-day sprint
Full refund if we do not find 10 staff hours a week. The fee comes off the build.
- Buildsper workflow, half to start and half when the acceptance criteria pass
- from $7,500
- Running it after launch, optional
- $1,200 a month
Who we work with
We have no retail AI case study to show yet. Adidas is a QueryNow client from before our AI offer. QueryNow has shipped 200+ production systems since 2014.
Case studiesWhat we want from you
Tell us the task that eats your team's week.
Three requests from a normal Monday morning
Pick one from the inbox to see what the agent prepares before anyone on your team touches it.
Tracking says delivered Friday but there's nothing here. Order 48213. Can you resend?
Prepared by the agent
- Order 48213 matched, $86, shipped Oct 14
- Carrier: delivered Oct 17, no photo on file
- Policy: reship allowed after 48 hours
- Reply drafted, reship waiting for approval
- Today
- Someone looks up each order across several systems and writes much the same reply all day. Exceptions wait until a lead has time.
- With the agent
- Reads each ticket, looks up the order and the carrier events, drafts the reply or prepares the refund exception with its evidence, and queues it for one-click approval.
Examples use invented names and figures.
Your team approves refunds and retailer disputes
The agent does the lookups and the first draft. A person on your team decides anything that moves money or speaks for your company. Refunds and reships outside your policy wait for approval. So do retailer disputes and anything sent to Amazon under your name, such as an appeal.
Some things the agent never handles. It works from order references rather than card details, and card numbers that shoppers paste into messages are removed from its inputs, a control we test before launch. It never writes or posts a review, and it never puts a new claim on a product page unless a person approved it first.
When a category such as order status replies has run cleanly for long enough, you can decide in writing to release it without review. Until then, every draft waits in the queue with its evidence attached.
Schedule of services
Fixed prices. Nothing monthly until an agent is live and you have seen it work.
| Service | What you get | Terms | Price |
|---|---|---|---|
| AI Workflow Sprint | A working prototype and a fixed build quote, with the staff hours each job costs you, ranked. | Full refund if we do not find 10 staff hours a week. The fee comes off the build. | $2,500five days |
| Build | One production workflow connected to your systems and email, with an audit log and a staff-hours report. | Half to start, half when the acceptance criteria pass. | from $7,500per workflow |
| Run | Monitoring, model updates, fixes and a monthly report of hours given back. | Optional, and only once an agent is live. | $1,200a month |
Start with one week
Tell us where your team loses time. We reply within 4 business hours.
A working prototype on your own documents for $2,500. Full refund if we do not find 10 staff hours a week.
- Day 1 is a 90-minute session where we watch how the work runs today.
- By day 4 your team is trying a working prototype and telling us what is wrong.
- On day 5 you get the staff hours each job costs, ranked, and a fixed build quote.
15 minutes with Mahesh. He will ask about how many order-status tickets and retailer deductions your team handles each week, and which helpdesk and order system you use. You leave with a rough hours estimate and a straight answer on whether a first project makes sense. No pitch deck.
Questions retail and ecommerce brands ask
Our helpdesk already has an AI agent. Why would we need this?
Keep using it for the in-policy answers it closes. The tickets it hands off are the ones that need data from outside the helpdesk, such as your 3PL or the carrier, or a decision outside policy. We build the agent for that work, with an approval queue and the evidence attached.
Does the agent see card numbers?
It should not, and we design and test for that. The agent works from order references rather than card details. Card numbers that shoppers paste into tickets or chats are removed from everything the agent reads or logs, and any system that receives the original message stays in your PCI assessment. The agent can prepare a refund for approval, and only the approved refund action touches your payment system.
Do we have to tell customers it is AI?
If a bot talks to shoppers on its own, plan on saying so at the start. Several states, including California and Utah, have disclosure rules, and the EU AI Act chatbot rule has applied since 2 August 2026. Replies your team reviews and sends go out from your team, and counsel should confirm your policy for those. Our guide on chatbot disclosure laws has the details.
We sell to big-box retailers through SPS Commerce. Can it work our deductions?
Yes. The agent reads remittances and portal exports alongside what your EDI provider already sends, matches each line to its evidence and drafts the dispute. A person approves and submits. The sprint checks how each retailer accepts disputes.
We are a 25-person brand. Is this for us?
It can be. Fit depends on volume and handling time more than headcount. As a rule of thumb, a support team of three, or one person who works retailer deductions, often has 10 hours a week of routine work. Estimate payback by setting the hours saved, at your loaded labor cost, against the build and running costs. Full refund if we do not find 10 staff hours a week.
Guides for ecommerce and operations teams
Where AI saves hours at a consumer brand and where a person still has to decide, written for operations leads rather than technologists.
All four guides- Amazon search suppressed: why listings get suppressed and how to fix them from your own product dataA search-suppressed listing loses its place in Amazon search until someone notices. A fix typed into Seller Central often lasts until the next feed. Here is how to find the cause and fix it in the product data every channel reads from, with a person on the approval.
- Vendor chargebacks and retail deductions: how suppliers work them before the dispute window closesEvery short-paid remittance is money a retailer has already kept. Getting it back means matching each line to its order and proof of delivery before the retailer's window closes. Here is a method, and where an agent fits.
- AI customer service for Shopify brands: automate where-is-my-order, keep refunds with a personOrder status and return requests come up every day at a Shopify brand, and helpdesk bots answer some of them. Measure their share in your own inbox before you estimate savings. Here is where the rest of the hours go and which decisions belong with your team.
- AI chatbot disclosure laws for ecommerce brands in 2026: California, Utah, Maine and the EUA brand that sells nationally already has customers in states with chatbot disclosure laws, and since August 2026 the EU has its own rule. One design choice covers most of them. Here is what each rule says, with a checklist at the end.