If you're running an ecommerce team right now, you already know the feeling of finally getting your digital shelf under control and then everything changing again.
Product content used to have one job: to show up correctly on a retailer's PDP. Now, in the world of agentic commerce, it has to show up correctly and get chosen by an AI agent that's shortlisting products on someone's behalf.
I talked about this tension at this year's Digital Shelf Summit, and I want to dig into it further here, because I don't think the answer is asking your team to do more with what they've got. Instead, it should feel like they’re getting a promotion — moving them from doing the work to directing it.
Shoppers used to start with a keyword and the search bar. Now, more and more of them are starting with a command or phrase, like "find me a durable chew toy for my puppy that isn’t plastic" or "what's the best sunscreen for sensitive skin," typed into a chat window.
And an agent, not the shopper, is deciding whether a product is worth showing.
This changes who sees your content first and who your content has to convince first (spoiler: it’s AI agents). Instead of wondering whether your PDP is optimized (which is still a valid question), you also need to ask whether your product data even qualifies to be the answer.
If an agent can’t understand what your product is, who it’s for, and what it does, it’s not going to shortlist it, no matter how good your PDP looks to a human.
Last year alone, Salsify completed over 768 million automated tasks (a 50% increase year over year), and our customers collectively published more than 5 billion products.
This goes to show that teams are running flat-out just to keep their digital shelf fed. Add agentic complexity on top of that, with the same team and the same manual workload, and you’ll quickly find it’s not sustainable.
So it's no surprise that when I talk to customers, they’re not asking me whether they should be leveraging agentic automation — they’re leapfrogging that question entirely and asking how to get started and how to govern it.
I want to caveat this by saying these are still open questions industry-wide in 2026. Anyone who tells you they've fully solved agent governance is getting ahead of themselves. We're all figuring this out together in real time.
What I have found, though, is that the AI itself isn’t the tricky part. The models are very capable. The hard part is helping teams trust a new way of working, redefining what "my job" means, and building guardrails that keep people feeling confident.
Think about what logging into your PIM used to mean. It meant data entry, writing formulas for a computed property, and reconciling a spec sheet against what's already in the system. This is changing. Not that you’ll be logging in less, but rather, you’ll be logging in differently.
As I said on stage at DSS earlier this year: Instead of logging in to do your work, you're logging into Salsify to define the strategy, set the goals, validate the work the agents are doing, and provide direction.
I like to think of this as a promotion. Your team spent years becoming excellent at the tactical work, and now that expertise gets to be used for something bigger. Instead of being the one entering the data, you're the one deciding what "done" should look like, and making sure the agents doing the work are meeting that bar.
I want to be really clear about one thing, though: You’re absolutely not handing over the keys and walking away. Your team is still setting the strategy and deciding what good looks like. They’re still accountable for what actually ships. The agents just do the repetitive work.
To show this isn’t theoretical, take Liam Buchanan and his team at Kerry Group as an example. Using Intelligence Suite, they took their new item setup process from days down to a single day, which is an 85% reduction in time. As a result, his team gets hours back to spend on work that still needs human judgment and input.
Quick context, in case the term's new to you: Product experience management (PXM) is the system your team already uses to manage product content, from specs and images to how that content gets published across retailers. The agents I'm describing are built into these systems rather than a separate tool bolted on.
In other words, PXM has to evolve from just being a tool to becoming a team of domain-specific intelligent agents that work alongside you and under your direction. Here’s how it can do that.
Right now, a lot of content creation starts from scratch every single time. Someone writes a prompt, explains the brand voice again, re-pastes the style guide, and hopes the output sounds right.
With PXM agents, you don’t need to re-explain that context every time. Your brand guidelines, tone of voice, and style rules get stored once and applied automatically, regardless of whether an agent is writing product copy, translating it for a new market, or generating images.
If you’re a little nervous about handing off work to AI, just know that validation doesn’t get skipped just because a task is automated.
Think about the basics that used to eat up so much of your team's time, like checking a spec sheet against what's in your system, making sure a listing still meets a retailer's requirements, fixing a mapping as soon as a retailer changes their schema.
These are exactly the tasks that PXM agents can handle continuously, catching issues as soon as they crop up.
Today, getting a product live means working through a task list: fill in the fields, run the formulas, check the mappings, submit for syndication, fix what bounces back. Every step needs someone to do it.
With Salsify, you don’t. Instead of handing your team a task list, you tell the system what you want to happen, like "launch this product globally" or "get this SKU ready for Amazon."
From there, our event-driven architecture completes the next steps. The agents work across your data, your workflows, and your retailer connections to make it happen, checking accuracy and compliance as they go.
And you decide how much of that you want to watch. You can set checkpoints wherever you want someone on your team to weigh in.
Let's go back to the question I raised earlier: How do you govern this safely?
The thing about AI agents is that they don't know when they're wrong. An agent doesn't second-guess itself. If it's working from outdated specs, an unapproved product description, or data that's just … off, it won't hesitate. It'll answer confidently anyway with that wrong or out-of-date information.
So the real question to answer here is: What’s the agent working from? An agent is only ever as good as the data it's grounded in. Feed it messy, unverified, out-of-date product content, and you'll get messy, unverified, out-of-date results. But if you feed it governed, brand-verified data, it can make accurate calls on your behalf.
That's why this is so important. Whether it's your own team's agents, a retailer's agent, or a third-party agent like the ones ChatGPT or Amazon are building — this is agentic commerce in practice — they all need to pull from somewhere.
This is also why we didn't just bolt AI features onto our existing system. Salsify was built so that agents work natively inside the same platform where your product data is created, checked, and kept accurate.
The content, the constant validation, the outcome-first workflows all run against data your team already trusts rather than a separate AI layer working off a stale export somewhere else. The system of record and the system doing the work are the same system.
This might feel like a lot to take in — so let's make it practical.
Start by looking at where your team is still doing the agent's job for it, which is probably the manual data entry, the repetitive checks (aka, the tasks nobody really enjoys). Pick one that happens a lot and causes the most friction, and hand that one off first.
You don't need to have it all figured out on day one. Start small, see what works, and build from there.
At the end of the day, this isn't about needing fewer people on your team. It's about giving the people you already have more leverage, and taking the work that was eating their time and handing it to agents.
We're moving toward building your AI workforce — or, as Rob Gonzalez, co-founder and chief strategy and innovation officer at Salsify, would say, giving everyone in the room a promotion, because in this world, speed matters and humans alone can't keep pace.