Automation & AI

Ecommerce automation that knows what not to touch

The jobs that are too repetitive to do well by hand and too consequential to get wrong.

Most ecommerce automation is sold as magic and delivered as a rule engine nobody trusts enough to switch on. The useful test is narrower: does this remove a job you are currently doing badly because there is too much of it?

Three qualify. Writing listing copy for a thousand products. Watching competitor prices on a marketplace all day. Noticing which customers have quietly stopped buying.

Where AI genuinely helps, and where it does not

A language model is good at turning a set of attributes into a marketplace-appropriate title and description, in volume, in a consistent voice. It is good at spotting that a description contradicts the product’s own specification.

It is not good at deciding what your stock level is, what something should cost, or whether a listing is compliant — those have right answers, and a plausible-sounding wrong one is worse than no answer. So generated copy is a draft you approve, and nothing generated is published to a marketplace on its own.

Repricing, with a floor you set

Marketplace pricing moves all day and the buy box goes to whoever is currently competitive. Repricing rules move your price within a band you define, with a floor that protects your margin, so the automation can win the position without giving away the reason for having it.

The rules are tiered rather than a single percentage, and they round to prices that look deliberate rather than computed.

Image compliance

Amazon rejects images for reasons that are precise, published, and tedious to check by hand across a catalogue: background not pure white, product too small in the frame, wrong aspect ratio. Checking those is arithmetic on pixels, which is exactly the kind of job to automate.

Editing them is not. An automated background removal that took a face off a photograph is a real thing that happened here, which is why the tool reports what is wrong and prepares a fix for review rather than silently rewriting your library.

Customers who drifted away

The most valuable segment in most catalogues is people who used to buy and stopped. They already know you and already converted once. Finding them is a query, not an insight — the work is running it consistently and following up, which is what automation is for.

The principle underneath all of it

Automation is allowed to prepare work and allowed to do reversible work. Anything irreversible — publishing to a marketplace, overwriting a listing, changing a stock figure — waits for a person, or passes through a rule that fails closed.

One catalogue, every channel

Connect a channel and see your own stock reconcile. No card required.