From industry giants like The Washington Post testing automation to cover the Olympics to Scandinavian media groups using automated texts to stay relevant in local communities, the media industry is waking up to the opportunities of deploying robots to generate automated content.

But there are still more questions than answers, such as does automation pose a threat to journalism? What are the key success factors in creating real value? And how do publicists overcome complexity and start using automation in everyday operations?

On Wednesday, INMA members learned the possible solutions to these questions in a Webinar presented by Thomas Sundgren, chief commercial officer at United Robots.

Sundgren led with a look at the challenges that media publishers face in today’s digital age, such as growing digital subscriptions, advertising in post-cookie reality, combining reader revenue and advertising, doing better with less, and maintaining a journalistic presence in a crowded digital space.

“To replace old revenue with new revenue, there’s a tremendous amount of work to be done,” Sundgren said. “I think it’s important to acknowledge that digital subscriptions and reader revenue are huge content opportunities for media companies. The challenge is to exploit that opportunity with new and different distributions strategies of content, and also handle that opportunity with a limited number of resources.”

When the industry looks at possible solutions to that, Sundgren said it’s important that such solutions be scalable for publishers from the giants to the small, local newspapers.

Artificial intelligence (AI) and automation

“When it comes to AI and automation, the toolbox is really quite extensive,” Sundgren said. “They can do many good things for readers and media companies.”

Of all of these potential solutions that automation can bring, the focus of this Webinar was on serving readers and newsrooms with automated content. Sundgren focused on three specific areas.

• Natural language generation.
• Robot texts.
• Automated quality content based on structured data.

“We’re not talking about scraping the Internet. We’re talking about finding trustworthy, transparent, correct, and consistent data to create quality news content,” Sundgren said.

Media challenges

There are still many questions that need answers, however:

• How can the automated opportunity work for all publishers — not just the big ones?
• Are there threats, and if so, how do we handle them?
• How do newsrooms overcome thresholds and start using automation in daily operations?
• How does the industry go from testing to easy-to-board, standard solutions?
• What are the value drivers, business cases, and key success factors?

It’s the last question that this Webinar addressed. “AI and automation should be a driver for journalism, not the end goal itself,” Sundgren said.

Value drivers of automation

There is a “golden rule” for the value for automation could be expressed by a pyramid.

“This is the same logic that should guide the conversation for any tech,” Sundgren said. There are some specific value drivers that publishers should consider.

• Comprehensive coverage: giving readers content that wouldn’t otherwise get done (time restraints).
• Enhancing journalism: freeing up time for human reporters by letting robots do the legwork on volume reporting.
• Speed in breaking news: serving readers with local or national news, and newsrooms with alerts.
• Growing digital subscriptions: driving registration, conversion, and retention at low costs.
• Digital advertising: growing data-enriched premium ad inventory at low costs.
• Personalisation: creating relevancy through volumes of auto-generated, granular content.
• New products: testing, scaling, or aborting with minimal cost and effort.
• Scaling and expanding presence: growing journalism and business by using automated content as a volume base.

“If media companies let automation do the legwork, they can let human reporters do the real jobs — unique reporting, investigative reporting, feature reporting.”

Sundgren added, “When it comes to digital subscriptions, automation can be a value driver if you do it the right way.”

Newsroom case studies

Sundgren presented newsroom examples of publishers who used automation to enhance their coverage. The Washington Post uses automated content to cover special events such as the Olympics, elections, and high school sports.

“This means hundreds of thousands of page views and ad servings for the Post at a low cost,” Sundgren said. “The value for them is advertising, comprehensive coverage, and they can free up reporting time for features.”

Another example is Bloomberg, which uses automation to create and distribute texts reporting company financials. “They’re putting out thousands of texts per quarter on earnings, through automation. It reduced their article production time from 30-plus minutes to just a few.”

The Guardian Australia started automated texts on donations made to political parties. The system takes structured data and turns it into news stories, which frees up time to do feature reporting and creates coverage that might not otherwise get done. Perhaps most importantly, they are serving the community with this important information.

Key success factors

Sundgren shared some other examples being done by Associated Press, NTM, Aftonbladet, and MittMedia. Using these case studies, he launched into the key success factors of these case studies.

• Don’t “over-tech. “It’s easy to let automated content become pure tech projects. It must be regarded just as any other kind of content.”
• Set value objectives. It’s easy to forget that objectives must be applied here as well.
• Focus on metrics. Connect and understand the data, then iterate and evaluate.
• Embrace volume. “I don’t think you should be afraid of distributing a large amount of automated text. A large volume is definitely key to having automated content and driving personalisation.”
• Set a distribution strategy. Automated content must be properly distributed and mixed with human-made journalism.
• Focus on persistency. This is long-tail stuff.
• Get everyone involved. Everyone must understand how and why automation is applied, for buy-in and effect.
• Focus on quality and transparency. Everyone will notice if you don’t.

MittMedia did a study to see if readers could tell automated content from human-made content — and they could not. “If quality is paired with transparency, readers will really like it and buy into it,” Sundgren said.

“In a lot of the successful cases we see, automated content is a volume-based complement for unique reporting.”

Conclusion

Wrapping up his presentation, Sundgren offered some key thoughts for INMA members.

• Automated content can be a value driver, but it must be treated like any other content.
• Business and journalism should always come first, then technology.
• Set strategies and objectives, then execute.
• Quality, persistency, volume, and distribution strategy are important factors.
• Most importantly: Get everyone involved from the start.

“I think it’s important to understand that this can bring value to the media industry, and transformation,” Sundgren said.

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