Monday morning. A supplier sheet with 300 new products, a translation export that was due last week, a list of collection pages that still have no copy, and three dashboards to check before anything goes live. None of it is hard. All of it takes the week.
For years the answer to that week was a better tool. Today we are shipping a different answer: the tools run themselves, and your team reviews the result. And with this update, that pipeline is faster and simpler to run.
What is new: agents do the work, you review it
The Agent Hub is where your catalog work now runs. Agents write product and collection content, translate it, structure your data and prepare it for your channels. Every result lands on one board as a draft, and nothing reaches your store until someone on your team approves it.
It brings four parts together. Triggers decide when work starts. Recipes define what the work is. Knowledge holds your brand rules. The Board is where you decide.
Why we stopped selling tools
The first wave of e-commerce software gave teams better tools: a PIM here, a feed tool there, an AI writer in another tab. Each one saved some time, and each one needed a person to run it. The work did not shrink. It moved between screens.
We saw the same pattern at every retailer we onboarded. The bottleneck was never the quality of a single tool. It was the hours people spent copying, checking and moving data between them. That is not a tool problem. It is a silo problem.
So we flipped the model. Instead of more tools, we built an agentic pipeline. Agents operate the tools, from one catalog. Your team reviews the output and optimises the rules. The job changes from doing catalog work to deciding on it.
How the agentic pipeline works
Work starts on its own. A trigger fires when products are imported, when content or media changes, on an hourly, daily or weekly schedule, or when another task finishes. That last one is what turns single steps into a pipeline:
- A new product is imported, and the product content recipe writes the title, description and SEO copy.
- The next step translates it into the languages you sell in, up to five per run, using your glossary.
- A publish step sends products to Shopify or Magento, as draft or active, your choice.
On the board, every run is one card. Open it and you see the current value next to the proposed one, field by field. Keep what is right, edit a line inline, deselect what is not, then apply. Or reject the run.
Draft is the default for every trigger. For recipes your team has learned to trust, a trigger can run in full auto, within a daily credit cap you set for the whole organization.
Faster and simpler to run
This update is mostly about removing friction from a pipeline that was already doing the work: fewer steps to start it, quicker review, and less back and forth before anything goes live.
- Start any agent task on a selection straight from your Products or Collections list.
- Review descriptions as formatted text, with real headings, and edit them in the same rich editor used on product pages.
- Drop a spreadsheet with SKUs or barcodes into the agent chat and get one review card with current and new metafield values.
- Approve a generated image and it lands in the right gallery position, even if the gallery was reordered in the meantime.
- Fewer false alarms: your own unpublished media changes and variant ordering no longer show up as channel drift.
Two returns: time back and pages that perform
The time return is the one teams feel first. Writing, translating, structuring and moving data between systems now happens before anyone opens the board. Your team's hours go to review and direction, not production.
The performance return is where the work starts to earn. Three recipes carry most of it:
- Collection SEO texts: keyword research per collection, then an SEO title, meta description and page copy written to your content format, linking only to collections that exist.
- Translations per market: your glossary keeps house terms consistent, and links in translated collection copy point to the translated pages, not back to the source language.
- Internal links on existing copy: a separate recipe adds links to collection texts you already have, without rewriting them.
At PL Line, a multi-brand store with 82 brands and 259 collections, category copy was written in four languages and approved by their own SEO agency inside the platform. Content creation costs came down 60%. Read how that worked.
You stay in control
An agent that writes for your brand has to know your brand. Knowledge documents hold your tone of voice, generation guidelines and content formats: structure, length, linking rules and banned words, with a version per language. Content runs do not start without a content format.
Feedback becomes a rule, not a one-off fix. When a reviewer spots a pattern, the rule goes into the format document and every next run follows it. The second review round is shorter than the first.
After publishing, the pipeline keeps watching. When someone edits a product directly on a channel, you see the difference and decide whether to adopt it or keep your version.
New Impact dashboard: what happens to traffic, clicks and conversion
Doing the work faster is half the story. The other half is knowing what it did. The Impact dashboard shows what happened to each product and collection after it went live: Google impressions, clicks and average position, and the orders that followed.
It starts with one fact: the exact moment work went live. From now on, every product and collection Automated Commerce publishes to your Shopify store is logged with that moment. The date is stored once, so a retry can never shift it. That record is the anchor for every comparison.
- Results only count after the change went live, and the go-live day itself is left out.
- Products are compared with their own recent baseline and with similar products that were not touched in the same period.
- Products that have not sold yet stay in the numbers, so a result is never flattered by leaving them out.
- Collection pages are measured over 28 days on each side of the first publish, against your other collections over the same dates.
A first group of retailers already works with the Impact dashboard, and it rolls out more widely in the coming weeks. From then on the conversation moves from hours saved to results you can point at.
Get started
The Agent Hub is live on every Automated Commerce plan: Pro, Scale and Enterprise, and every Shopify publish is on the record from today. Open it, add your content format under Knowledge, and run the Collection SEO recipe on one category. Not on Automated Commerce yet? Book a demo and we will show you the pipeline on your own catalog.