I had the good fortune to hear The Washington Post’s chief technology officer, Vineet Khosla, speak in San Francisco at INMA’s Media Tech & AI Week a few weeks ago. This is a person who helped build Siri and went on to work at Uber. Now, he’s helping the team at WaPo navigate chaotic AI-driven, Trump-driven, Bezos-driven times.

Here are some of the insights Khosla shared:

The changing user experience for Audience V3.0

A big technological shift in the news industry will change at least two of the three processes of creation, delivery, and consumption, Khosla said. GenAI has changed all three at the same time, with changes in news consumption leading the way. 

“People want to be more involved with the news,” Khosla said. “They’re not passive and waiting for us to sit and deliver that news to them. People have queries, they have questions, they have curiosities. They already know what’s going on. Sometimes they want details. Sometimes they just want skim. They want to consume it in different forms, and we have to accept that one size does not fit all. So we really have to change our consumer experiences there.” 

Unlike the shift from desktop to mobile, this shift occurred swiftly, within two weeks, because news consumers already had the devices they needed in their pockets.

Learning from comments

“This is one place which, when I joined this media industry two years ago, I felt this is so silly. All newsrooms I saw pushed comments all the way down to the bottom because they said, man, it’s a cesspool of hate. And then at the same time, you go, do you guys realise Twitter and Facebook and all of these companies have used your content and the conversation around it to build their growth? So why don’t we invest in that?”

The Post now uses AI for comment-moderation. “We actually went from rejecting 3% of the comments to rejecting 15% of the comments. It’s doing a better job because it was not just spot checking. It started making a safer community for our commenters to come in and engage and not get abused.” 

You can read about the Toronto Star using GenAI for their comments here.

The Post also built a feature that summarises comments, which led to another interesting user experience, Khosla said. 

“We started noticing there’s some people who are coming in, they’re clicking on the comments icon, they’re reading the summary to decide whether they want to read the rest of the article or not.

“So it’s almost like, is the peanut gallery interesting enough and exciting enough? Are people activated by what’s happening in this article? Oh yeah, that’s worth my time. So as much as headlines matter and as much as summaries matter, these types of experiences also matter, and we were able to unlock it because we started investing in using GenAI to do more and more with us.”

Learning from audio

“We have finally gone to full audio mode. Now you can have all of these conversations natively on the device, and now we’re in the stage of learning from our users,” Khosla said. “Does it change queries when we type versus when we talk? Does the time they engage with us start to change?”

It’s fine for an answer engine to say, “I don’t know”

When people ask questions of The Post’s answer engine, it replies with “I don’t know that answer” about 20% to 30% of the time. 

“For a typical tech company producing this type of product, this would be death. They would never do that. But we were very happy with that. We leveraged only our own journalism, and if we did not find evidence in it, we were OK saying we don’t answer.

“For every answer we produced, we had to be mindful of the fact that we, unlike other companies that are producing these engines, truth is the differentiator.”

You can hear more about Khosla’s framework and tools to position legacy media in an AI-first ecosystem by accessing  INMA’s Media Tech & AI Conference (San Francisco, October 2025) on demand here.

If you’d like to subscribe to my bi-weekly newsletter, INMA members can do so here.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .