Veristage’s Insight Platform Processes Over 50,000 Books Monthly

The chief customer officer of Veristage says publishers now process the equivalent of more than 50,000 books per month through Insight, her company’s AI publishing services platform—and that figure is growing fast.

You’re based in the U.S., but working with more than 80 pub****lishing houses in 15 countries. How are they using it?

Publishers use it for a range of tasks, from proofreading and translation to metadata and accessibility work. The breadth of our clients has given us the experience to build tools people actually need. They range in size from three-person teams up to multinational organizations and cover trade, education, and academic and institutional publishing—a really wide range.

Why can’t a publisher just run a manuscript through ChatGPT or Claude and get the same result?

The point where publishers come to us is when they say, We’ve used ChatGPT or Claude, we’ve done the experimentation, and we’ve hit a wall. If you upload a book into ChatGPT and ask it to proofread according to your house style guide, you’re probably not going to get a good result. It’ll stop eventually, or do the full book inconsistently. We’ve built tools that process the full book using consistent rules and return consistent results. You choose your rule set, press a button, and get the proofread back. And if it didn’t work the way you wanted, you have someone to call to help you fix it.

What shouldn’t publishers use AI for?

If you’re relying on its general knowledge, that can go wrong. For example, if you upload a book into ChatGPT and ask for comp titles, you’ll get comp titles, but they’ll skew toward well-known titles and be older, because the model’s training data has a cutoff of six months to a year ago. The other thing, specific to publishing, is that we don’t advocate that people use it to decide what to publish. We still see that as something that has to be based on human expertise.

What would you like to see evolve with AI a year from now?

A lot of what we spend our time on right now is dealing with file formats. PDFs are still pretty tricky, because they’re static display formats—if you move one letter, the others don’t move with it—and I’d love to see that fixed. But let’s be honest, we’re just at the beginning of this—still in the starting-out period, at least it feels that way to me. The change has happened fast, and it’s hard to know what it might look like once things settle.