How we’re using AI at Reach

Q: How are you using AI in your business now?

A: To answer that question, we need to split it into our three main ways of using it — commercially, editorially and centrally.

We’ve been using AI commercially for several years now. When I founded our in-house ad tech platform Mantis back in 2019, it was originally an AI-powered brand safety tool, to open up more of our inventory for advertising and keep innocent stories from being blocklisted (eg. football stories with the word ‘shoot’). We then realised it could be used to do so much more, and over the last few years, it has grown to power contextual ad targeting, translation, and content recommendations on our sites via content classification.

Editorially, we’ve been using AI for over two years, primarily via our in-house platform, Guten. It’s an editorial support tool and most commonly used to rewrite content from the wires or from one of our titles to tailor it for another, using the relevant house brand style. Through this work, we were also able to automate much of our CMS, cutting in half the amount of time it takes to upload a story. We’ve also seen our data journalists find interesting ways to apply AI to the sourcing of stories by quickly analysing and pulling out data.

Both of these tools were built and refined by us and for us, and through that journey, we realised that some of these tools might be valuable to other publishers, which is when we expanded Mantis as a B2B proposition and additional revenue stream.

Now, we’re starting to use AI more across all functions, mostly as administrative support. The opportunities are huge, as there are very few people who can’t benefit from speeding up parts of their job, from filing emails to creating schedules to summarising or analysing data.

Q: How do you expect this to change over the next 12 months?

A: I see general business use as an area of massive growth. I’ve been working to set up an in-house ‘AI university’ to upskill everyone to explore how they can use AI to support their everyday tasks. Ideally, we’ll go beyond ad hoc support — which is probably where most people are now — and move to creating agents which will routinely manage and deliver certain tasks (eg. setting up a schedule, providing a report). This isn’t just about saving time but fast-tracking the “admin” bit so that the person then spends more brainpower on the part that adds value — analysing the data and deciding next steps.

Commercially, Mantis will support increasingly powerful ad targeting and provide more data-based insight that we can share with our advertising partners. We’ll be able to activate larger sets of data to then target more granular cohorts, giving us more agility in classifying content. For example, classifying not just the topic or keyword but also how the user interacts with it (eg. is it a ‘how to’ article vs an ‘explainer’).

We’ll also be looking to add more AI-powered B2B products to our stable of proprietary tech, to license to other publishers.

In terms of our content, I’ll be working with our editorial leaders to go beyond productivity and start looking at new ways we can use AI to solve problems. We’ll be using AI to provide more useful insight about what audiences are engaging with today as well as what they’ll be engaging with tomorrow.

Journalists have so many metrics to consider, that we wanted to simplify some of those signals and pare down various metrics (including engagement, scroll depth, brand safety, propensity to consume another page view) to a single, more usable score which takes more of a 360 view into account. For example, while crime stories typically don’t get as many views as a celebrity story, people are much more likely to read all the way down — this tool will capture that kind of insight instantly, embedded right into the CMS.

Of course, we’ll also be seeing more of our audience use AI in their own lives and that’s something we’ll need to reflect in our products. For example, we’ll start to see more publishers embedding AI search within their own platforms. That will mean that for those readers who want to use AI to help them digest content, they won’t have to leave the publisher’s trusted environment.

Q: What are your three top tips?

1. Don’t do AI for AI’s sake. Instead, ask yourself what the challenge is that you’re trying to solve and work backwards.
2. Don’t silo AI in the organisation. Set up the right guardrails and support and let everyone have a go. Make sure knowledge sharing happens so that you can build on each other’s learnings.
3. If you wonder if AI can do it, then try it.