Robo-journalism is coming – but how will it affect your brand?
Last month, Mumbrella and Hotwire hosted an exclusive roundtable discussion on the challenges and opportunities of AI-generated reporting. Here, we recap the panel’s findings.
Mylan Vu, the managing director of global communications agency, Hotwire, has a theory about how customers have changed in 2019. “There’s no longer a brand safe from judgement by consumers,” she says. In fact, Hotwire’s research suggests Aussie marketers – more so than any other nation – are more likely to manage a crisis at one point or another. On reflection, it’s no surprise that cynical consumers are holding businesses to account because they’re being bombarded with more and more news and advertising.
All of which means, of course, that journalists now have more power and influence than ever before to make or break a company’s fortunes.
Yet, soon, marketers may have to face an altogether more unpredictable threat because editors are commissioning AI robots to generate stories. What was once the stuff of science-fiction is now a reality to help alleviate the workload of depleted newsrooms. In 2014, for instance, an algorithm called Quakebot helped the Los Angeles Times publish a story within three minutes of an earthquake. The Washington Post’s AI tool, Heliograf, wrote 850 articles in its first year, generating 500,000 views. And closer to home, Guardian Australia published its first scoop written by a robot in January this year – a detailed, important piece about political donations falling significantly.
“Robo-journalism is something brands need to plan for,” says Vu. In a world where news will be generated instantly, she adds, businesses must prepare for every eventuality. “It’s one thing to have a policy, media or messaging training in place, but you must have a plan B ready, too. For brands, the best reaction is to have content ready to go. In the event of, say, a data breach, you need guidance, advice and an FAQ on standby.”
But how, though, does robo-journalism actually work?
“Machines can write because they look at data,” explains Jodie Sangster, the CCO of artificial intelligence experts IBM Watson. “AI can correlate information, put trends together and understand what they are. It can also understand how to compose a sentence because you teach it how to compose a sentence. It understands what a noun is because you’ve taught it to do that.”
But while the technology is already there, public sentiment towards robo-journalism still has a fair way to go. As evidenced by a barrage of anxiety-ridden clickbait articles on the topic, humanity is still not used to the idea of robots working alongside us. So when The Guardian published its first, it was careful to make it clear who – or what – was behind the story. ‘ReporterMate’ won the byline, instead of the journalist behind the robot, Nick Evershed.
It’s also why Hotwire commissioned research to find out just how fearful the public is about all this. It concluded, maybe surprisingly, that we’re happy to read reports written by robots – so long as publishers are upfront about it. We’d also rather have them deliver breaking news rather than anything more weighty. “If you’re a day trader reading financial results, you don’t care whether the analysis is there or not,” argues Vu. “Speed and immediacy of information is now consumers’ number one priority. They want to be the fastest to know something; the first to share.”
Speed, then, must be married with transparency. “When people are talking with chatbots but don’t understand they’re talking with a robot, they get quite upset about it,” says Sangster. “It might be a bit the same with journalism. ‘I don’t mind all of this,’ they think, ‘but just tell me what I’m dealing with here, rather than trying to masquerade it as something else.’”
But what the report also argued is consumers do expect human beings to be doing the stuff that requires some humanity. “I watched Waleed Aly interviewing the New Zealand PM about the Christchurch attacks,” adds Vu. “And it was so incredibly moving. Why? Well, because his first question was, ‘How are you?’ and you could see the pain and sorrow on her face. A robot can’t evoke that – you can’t teach AI to ask somebody how they’re coping after a terrorist attack.”
Whether the public likes it or not, robo-journalism is here to stay. With newsrooms shrinking thanks to declining ad spend, dwindling circulation figures and continuous consolidation, reporters are actively seeking out automation and artificial intelligence to help them with the tasks they no longer have time to do. Writing down recorded interviews, for example, is something almost universally despised by journos.
“Newsrooms are more sparse than ever,” says Paul Wallbank, Mumbrella’s news editor. In 2019, journalists just don’t have time to go through MP3 recordings, he says, writing down every word. “But today AI programs can come back with, if not a perfect transcript, then one good enough for you to listen and pick out the right bits.”
On the other side of the media fence, some comms jobs are already being replaced by machines. But for Vu, the takeover couldn’t come soon enough. “They’re going to take our jobs, but please, take them!” she jokes. “Because for the vast majority, particularly in comms, jobs are going to change for the better.”
One role she won’t miss is the creation of ‘clips’, physical newspaper clippings created by cutting and glueing newspaper coverage into scrapbooks, which were physically posted to the client only less than a decade ago. “If by some terrible chance you had an article on both sides of the paper, you had to go and buy a second newspaper to print the other side,” she recalls. “I don’t know a single PR in the world who’s going, ‘Man, I wish I kept that job. I really loved glue sticks!’”
These days, instead of spending time cutting and sticking into scrapbooks, Vu’s team are free to think creatively, allowing them the space to produce strategic work for their clients. Instead of stealing jobs, automation has allowed the quality and depth of work to flourish. “Now clips are something everyone on my team spends ten minutes doing every morning and ten minutes doing before they leave the office.”
In the machine and human ecosystem, the “creative stuff” is currently left to the homo sapiens. But will we ever get to the point where machines can tackle those tasks for us, too? Sangster isn’t too sure. “Humans are good at empathy, emotion, dreaming and all of those softer things which we can’t necessarily explain,” she argues. “Machines don’t have emotion. Machines are good at crunching data, analysis, looking for patterns, looking for correlations between sets of data or images or content.”
Analytics, though, is one area where AI is already helping journalists and comms professionals test, refine and adapt their work based on past performance data. IBM Watson’s Sangster points to a system where a journalist can ask AI to analyse their most popular headlines. Within a matter of five seconds, she says, a robot can run through every piece you’ve written in the last year and explain what worked and what didn’t. “So you’re starting to hone down the craft better. And I think that’s quite exciting because we can get better at what we do for the audience, which is what we’re there for.”
Pre-internet, journalists had no idea how many people were reading their articles, or for how long. Instead, they relied on gut instinct to decide which ones to lead on or spike. “But today we know exactly how many people have opened that story, and how many people come back to it,” says Wallbank. “So we have data behind us to say, ‘Right, this Married at First Sight or Channel Ten story will go gangbusters. This other one isn’t.’ That does affect the way that you prioritise stories and resources.”
Beyond that, machines are already helping journalists land scoops. Wallbank points to an experience in the Mumbrella newsroom when the team’s analytics software revealed several people were searching for a specific name and opening an appointment story from several years ago. “The question you ask is, ‘Why is everyone going to that story?’ So we rang that person and they told us they’d announced their resignation. And bang, you’re first to market. You’ve got a scoop because analytics gave you an early warning. Analytics backs a lot of the judgment now. You’ve got to prove yourself because the numbers are there, and that does change the game.”
This data-based approach to creative thinking also helps PRs provide better feedback for their clients, which in turn helps them test and refine their work. “It’s become increasingly common for our team to get feedback from journos along the lines of ‘When you wrote it this way, it got lots of clicks,’” says Vu. Her team can then feed that data back to their clients, allowing brands to strategically create a cycle of click-worthy and valuable content. “That openness and sharing of data you have access to is trickling down through to brands.”
So although the robots are well and truly here, machines are being drafted into work alongside humans – informing rather than replacing. “That’s really where we need to be going with AI,” concludes Sangster. “It’s not about, ‘The machine’s going to do my job for me’, it’s about, ‘How can I leverage AI to be better at what I do?’”