How to Automate Bookmanager

A bookshop stocks one copy of ten thousand different things

Bookmanager is the point of sale and inventory system a large share of independent bookstores across North America have run on since 1986. It is built in Kelowna, British Columbia, and aimed at shops that sell books rather than at retail in general.

The inventory problem it solves is unlike anything else in retail. A clothing store carries a few hundred products in a range of sizes. A bookshop carries thousands of individual titles, most of them one or two deep, all competing with next season's, and a good number returnable to the publisher for credit if nobody buys them. The buying decision gets made once per title, months before publication, off a catalogue, and the shop lives or dies on how often it was right.

Bookmanager holds the till, the on-hands, the special orders, the customer accounts, the receiving, the returns and the ordering that feeds all of it.

Special orders and returns fill the hours between customers

The counter is the straightforward part. The rest of a bookseller's day is a set of small open loops.

Somebody special-ordered a book eleven days ago and rings to ask where it is, which means finding out whether the wholesaler ever shipped it. A title on order since March is still on order, and nobody has established whether it went out of print. A publisher's returns portal has an authorisation waiting and the window closes. A rep's catalogue arrives for a season nine months away with three hundred titles in it. The prize longlist was announced this morning, and by Friday you will either have sold twelve copies or told twelve people you can order it in.

Some of those live on a website. Some of them are simply news.

Bookmanager

Prize news before the first customer Longlists and shortlists read the hour they publish and set against your on-hands, so the reorder goes in ahead of the rush.

Local demand you can actually see Course adoptions, festival programmes and library events read from the pages that publish them, in time to order for the term.

Backorders that stop being open-ended Wholesaler order status checked for you, with anything that has quietly gone out of print flagged rather than waited on.

Pubstock answers who has it and at what discount

Bookmanager is unusually strong on the buying side, which is the part booksellers care about most.

Pubstock gives live stock availability and pricing from well over a hundred publishers, wholesalers, distributors and remainder companies, so cost, discount and availability at each vendor sit on one screen before you commit, and a title that has switched suppliers surfaces as something to return elsewhere. Titlelink puts a database of around thirty million titles and cover images behind the system, so bibliographic data is not typed in, in-store or on the webstore. The Webstore is integrated rather than synced: an online order arrives in the software directly, and a special order can be raised from it with the customer notified automatically when the book lands. Returns are built out of sales history, invoice age and the date code printed on every shelf label, and anything marked for return comes out of available on-hands.

There is also anonymised sales data shared across hundreds of participating independents, which is a genuinely unusual thing for a system this size to offer.

What none of it can do is tell you why a title is about to start selling.

The reason to reorder gets announced somewhere else

Demand for books arrives as news, hours before anybody walks in asking, and that is the gap worth closing.

Prize longlists and shortlists read within the hour they are published and set against your on-hands, so the order goes in while the wholesaler still has stock rather than after four people have asked. Adaptation announcements and release dates for the film or the series, which move a backlist title that has sat quietly for six years. The weekly bestseller lists, checked against what is on your shelf on Wednesday morning.

Then the local layer, which a chain cannot see. Course reading lists published by the university and the schools nearby, read before term rather than during it. The library's event calendar and the festival programme, so a visiting author's backlist is in the building on the day. A local writer with a review in the regional paper this weekend.

The open loops on the other side too. Backorders on the wholesaler's own site, with anything that has quietly gone out of print flagged instead of waited on. Returns authorisations on publisher portals, with the ones about to expire at the top.

Keep the writing human. A reply to a one-star review is drafted by the machine and posted by whoever runs the shop.

A bookseller's judgement should not be spent on refreshing pages

None of that reading needs a bookseller. It needs somebody to open the same dozen pages every morning and set the four things that changed beside what is on the shelf.

WebRun is an AI agent that works a real Chrome browser, signed in as you. It opens Bookmanager, a wholesaler's site, a prize announcement page or a publisher's returns portal, reads what is there and gives you the list with the ISBNs on it.

Because it works the browser rather than an API, a prize jury's press page and a university course listing are as readable as any vendor system, and neither will ever publish an integration. It runs on your schedule in your own private environment, you can watch a run and stop it, and anything sent to a customer is drafted and waits.

The workflows below are already built, and each one names what it opens.