Lightspeed AI agents review: it drafts the purchase order, you still sign it
Lightspeed's AI can now do more than answer questions about your stores. It can draft the reorder, plan a transfer between locations and tag slow stock for clearance, then wait for you to approve. Here's what looks right, what's missing, and who should ask for the beta.
Verdict: promising for multi-location retailers already on Lightspeed, and worth joining the beta if your inventory counts are trustworthy. Announced October 15, 2026, Lightspeed AI's new agent features prepare purchase orders, transfers, product updates and clearance tags, show a preview, wait for approval and log every action. It's an opt-in beta for select customers, with no published price, date or plan list.
Who should read this
If you run two or more shops on Lightspeed Retail, you already know the Monday routine. Pull the sales report, check what's low, look at what the other store has too much of, write a transfer, write a reorder, send them. It isn't hard work. It's just constant, and it's the first thing to slip in a busy week.
On October 15, 2026, Lightspeed announced new agent features in Lightspeed AI aimed squarely at that routine. This review is for retailers deciding whether to ask for the beta, and for anyone choosing a point-of-sale system who wants to know whether "AI that acts" is real yet.
Short version: the design is sensible, the scope is right, and the important details aren't published yet.
What Lightspeed says it shipped
Lightspeed AI started as a chat assistant inside Lightspeed Retail that answers questions about sales and inventory and builds reports. The new features let it prepare actions, too, in the same conversation.
According to Lightspeed's announcement, the agent reasons across your catalog, inventory, sales and purchasing data across locations. It can identify low-stock products and prepare purchase orders, plan an inventory transfer toward the locations where demand is strongest, update product information, tag unsold items as clearance, and surface sales opportunities.
Before anything changes, it shows a preview of the action and waits for your explicit approval. Every action is recorded in an audit log.
The features are being introduced to select customers through an opt-in beta, with "broader availability coming soon." Lightspeed's AI page says Lightspeed AI is available on Lightspeed Retail plans and that access and features vary by plan. No price, release date or plan list for the agent features has been published.
How the work moves
From question to approved action
- Ask
What's running low across my stores?
- Recommend
The agent reads catalog, stock, sales and purchasing data.
- Prepare
A draft PO, transfer, product edit or clearance tag.
- Preview
You see exactly what will change.
- Approve and log
Nothing moves until you approve; the action is recorded.
An approval step is only as good as the attention behind it.
Green flags
It prepares, then asks. Nothing moves until you approve it, and you see what will happen first. For purchase orders and transfers, which cost real money and real staff time, that's the right default.
There's a log. An audit log means that when a strange PO shows up in three weeks, someone can see who approved it and when. That matters more than it sounds once two managers share the job.
It works where the data already lives. The agent is reasoning over the same catalog, stock counts and sales history your POS already holds, not a nightly export to another tool. There's no second system to keep in sync.
The jobs are boring on purpose. Reorders, transfers, clearance tags and product edits are frequent, rule-shaped and easy to check. That's exactly where an agent earns its keep. Lightspeed didn't lead with something flashy.
It starts from a question. You can ask why stock is low before you ask it to fix anything. A recommendation you can interrogate is easier to trust than one that simply appears.
Red flags
An approval step is only as good as the attention behind it. Forty previews on a Monday morning become forty clicks. If the preview doesn't show why (which sales, which period, which store had the excess), people will approve by habit. Ask to see the preview screen before you join.
"Demand signals" isn't explained. Lightspeed says transfers move stock toward the locations where demand signals are strongest. What counts as a signal? Last week's sales? A seasonal pattern? A sale that was really one bulk order? Until that's documented, treat transfer suggestions as a first draft.
It's only as right as your counts. If shrink, unrecorded damage or a skipped receiving step has your stock numbers wrong, the agent will confidently reorder what's already on the shelf. That's not a Lightspeed problem, but it will look like one.
The details are missing. Beta, select customers, no date, no price, no plan list. If you're choosing a POS this quarter, don't count it as a feature you have.
No independent results yet. The announcement has no accuracy figures, and we wouldn't expect any this early. Your own first month is the evidence.
It's Lightspeed only. If your purchasing happens in a separate system or a spreadsheet, the agent sees only part of the picture.
Decision guide
Green flags and red flags
| Consideration | Green flag | Red flag |
|---|---|---|
| Control | Preview and explicit approval. | Approval by habit if the reason isn't shown. |
| Accountability | Every action in an audit log. | No published detail on export or retention. |
| Data | Works on the POS's own stock and sales. | Wrong counts produce confident wrong orders. |
| Availability | Opt-in beta now. | No price, date or plan list. |
How it compares
The honest comparison is with what you already do. Reorder points and a weekly transfer report handle a lot of this with no AI at all; Lightspeed AI's existing ordering suggestions sit in the same family. If your rules are working, the agent mostly saves the drafting.
The other comparison is a custom agent on your own systems. That makes sense when your buying spans tools Lightspeed doesn't see, such as a wholesale portal, a warehouse system or an online store on another platform, or when your rules (minimum order quantities per supplier, which store may send stock to which) are unusual. It costs more to build and to own, and you control exactly what it can touch.
For most single-platform retailers, waiting for Lightspeed's version to leave beta is the cheaper path.
Verdict, and what to ask before you join
Ask for the beta if you run several locations on Lightspeed Retail, your stock counts are trustworthy, and someone owns purchasing. Wait if your counts are shaky. Fix receiving and cycle counts first; it pays off with or without AI.
If you do join, start narrow. Pick one category with steady sales and one supplier you know well, and let the agent draft only those reorders for a month. Compare each draft with what your buyer would have ordered, and note where they differ and why. That gives you a real sense of where to trust it before you let it near transfers or markdowns.
Questions worth putting to your Lightspeed contact: Which plans will include this? What does the preview show, and does it explain the reason? Can we set limits, such as a maximum order value or suppliers it may never use? Who can approve, and can approval be split by role? Can we export the audit log?
We build AI applications and agents that act inside real business systems, with permissions, previews and logs designed in. If you want help deciding what an agent should be allowed to do in your stores, or building one that spans the systems your POS doesn't see, talk to us.
Related work: Spin:Market.
We build alongside the people doing the work.
Bring us the problem you’re thinking about.