ShipStation AI review: auto-fill earns its place on the packing bench, the chat agents need a second opinion

The verdict

ShipStation announced a round of AI features on October 28, 2026. Two pieces matter to a small packing team: auto-fill, which learns how you usually set up a shipment and fills in the next one for you, and conversational analytics, three chat agents that answer questions about your shipping data.

Our verdict, up front:

Auto-fill: turn it on. It takes over a repetitive job your team already does by habit, and the mistakes it can make are visible on the screen before a label prints. Put a few guardrails around new products and heavy boxes, described below.

The chat agents: use them to start a question, not to settle one. They're a fast way to find where to look. Before you change a carrier, a box size or a staffing plan, open the report the answer came from and check the number.

Wait or go slowly if your catalog changes weekly, if you ship a lot of oversized or dimensional-weight-sensitive items, or if your team doesn't yet look at the package fields before printing.

What auto-fill does, and why it's the easy yes

ShipStation says auto-fill learns how each merchant configures shipments, including package type, weight and service level, and applies those settings to future orders. The numbers in the announcement are ShipStation's own: 1.4 million labels pre-filled a week, average label creation time down 35%, from about 100 seconds to about 65, up to 97% of suggested values left unchanged, and 69% of active ShipStation sellers with the feature on.

Those are vendor figures, but the shape of the claim is believable. Most stores ship a small number of products most of the time, and whoever packs them already knows "the mug goes in the 8x6x4, first class, 14 ounces." Auto-fill is the system learning the same habit. If you used to build automation rules for this, the difference is that you don't have to write and maintain each rule.

The time saving is real at volume. Thirty-five seconds an order doesn't sound like much, but on a 300-order day it's close to three hours of someone's time.

A suggestion that's right 97 times out of 100 is wonderful, until the three wrong ones are your heaviest boxes.

Where auto-fill needs a fence

Auto-fill learns from what you've done, so it's weakest exactly where your shipping is changing.

• New products. A SKU with no history gets a guess based on something else. Make it a habit that the first few shipments of any new product are set up by hand.
• Seasonal packaging. Gift boxes, bundles and holiday inserts change dimensions and weight. If November's box is bigger than October's, the system's habit is a month out of date.
• Weight and size breaks. An order that sits just under a carrier's weight or size threshold is cheap when it's right and expensive when it's wrong. Those are the orders where a person should glance at the fields.
• Service level promises. If you promised a customer faster delivery at checkout, a learned default for "this product usually goes ground" isn't the right answer.

None of this is a reason to skip the feature. It's a reason to decide, before you turn it on, which orders still get a human look. We've written before about deciding which orders get which labels inside ShipStation; the same habit applies here.

The three chat agents

ShipStation lists three agents, each pointed at a different part of your data:

• Shipping & Cost Optimizer looks at performance and costs by zone, service and carrier, delivery speed, return label costs and opportunities to save.
• Business Overview answers questions about orders, shipments, delivery tracking, top products and returns.
• Warehouse & Inventory Manager reports on picker performance, order processing times, batch efficiency, stock levels, low-stock alerts and reorder status.

This is a genuinely useful idea for businesses without anyone who builds reports. "Which zone got more expensive last month?" is the kind of question that used to mean an export and an afternoon in a spreadsheet.

The caution is the usual one with any chat interface over business data: the answer is only as good as the question and the definitions behind it. "Return label costs" might include labels that were created and never used. "Picker performance" depends on how your team scans, and a picker who covers the packing bench at lunch will look slow. Treat the agent's answer as a pointer to the right report, then read the report.

Be especially careful with picker performance. Numbers about individual people need context before anyone acts on them.

What we'd still like to know

The announcement doesn't say which ShipStation plans include the agents or whether there's a usage limit, and it doesn't describe how the agents define their metrics or whether you can see the query behind an answer. Those are fair questions to ask your account rep before you build a weekly routine around them.

It also doesn't say whether auto-fill shows which past orders a suggestion came from. That would make spot checks much faster, and it's worth asking for.

One more thing to settle inside your own team: who owns the defaults. When auto-fill is learning from everyone who packs, a new hire's habit in week one can become next month's suggestion. Pick one person who reviews a sample of auto-filled orders each week, corrects the product's saved weight or box when it's wrong, and tells the bench when something changed. That's ten minutes a week, and it keeps the system learning from your best packer rather than your busiest day.

Who should switch on what

Switch on auto-fill now if you ship a stable catalog of products in a handful of box sizes. Agree which orders still get a manual check: new SKUs, oversized items and anything near a weight break.

Try the agents with a known answer. Ask them something you already know from last month's reports. If the answers match, you've learned how far to trust them.

Hold off on decisions about people based on agent answers alone. Use the numbers to start a conversation.

If your shipping runs across several tools and you want the pieces wired together with sensible checks, our fulfillment operations team does that work. Get in touch and tell us which order type gives your packing team the most trouble.

Related work: Spin Market.