We released our AI PIM a couple of months ago, and our customers are currently putting it to the test.
Many companies are designing AI PIM to enrich product content, and that’s a valid approach. We do that, too. But what sets Kodaris apart is that we have our own PIM team—people who enrich product data directly for customers who need it done for them.
We Built It for Ourselves First
We built our PIM platform years ago. We released it internally, used it ourselves, and polished it through real, daily use. Our own PIM team has been running on it for years, and they are our pickiest users because they have direct access to our engineers.
That same team is now the first group training on AI PIM, using it every day, and already requesting enhancements. We’re getting real, firsthand feedback fast, from the people who feel it first.
How AI Is Changing the Workflow
The work itself is changing. Our PIM process used to run through three stages: a team that collected and updated product data, a QA1 team that did the first pass, and a QA2 team that signed off before publishing.
The process hasn’t disappeared; it’s evolved. AI handles the heavy lifting—data entry, research, and verification—shifting our team’s role from manual input to critical review. We aren’t removing the process; we’re optimizing the system.
Can AI Review Its Own Work?
Can AI review its own work? To a point, yes. We give it templated pages and clear direction on structure, style, and voice. We can even run a separate agent to review the first agent's output.
But here’s the reality: you’re publishing product specs that your customers depend on. AI is still fundamentally matching patterns. It isn't creatively thinking, even when its knowledge base makes it look like it is. Because it's built to complete the task, it can occasionally invent facts if that's the fastest path to an answer.
That’s not a hypothetical. I recently heard about lawyers facing disbarment for citing AI-fabricated case law in real lawsuits. The AI needed to support an argument, so it invented one. That’s the same failure mode we watch for in product data—wrong specs, different stakes.
"Hands-Off" AI Is Still a Work in Progress
This is why we keep a human in the loop: a basic check that the output looks right and that the AI didn't decide to hallucinate its own data. We apply that same standard everywhere at Kodaris—not just PIM, but engineering and marketing, too.
A lot of companies sell a "hands-off" dream. I believe that’s still a fantasy, and it may never be fully true. And that’s a good thing. True innovation isn’t about replacing the human; it’s about giving them the tools to do better work.
Can AI speed up this work by 80%? Absolutely, and we're seeing that already. But speed and full autonomy aren't the same thing.
What We're Building Toward
AI is accelerating this industry, ours included. We'll keep iterating on AI PIM and finding ways to drive down cost and turnaround time. But we're also going to keep the human review tooling available. The goal is efficiency, but it’s also ensuring you stay compliant and the content is high quality.
As we move through this AI journey, we’ll keep having this discussion. How much should we hand off, and how much should stay in human hands?