The use of Artificial Intelligence is changing every industry, and during this week’s Webinar, INMA’s Greg Piechota took a deep dive into some of the ways news media companies can harness it.

Coming one day before the release of the his INMA report AI Guide and ChatGPT Promptbook for News Marketers, Piechota — lead of INMA’s Readers First Initiative — gave members a glimpse into what the future holds. Members can watch the replay here.

While most news media companies are using AI in some form, Piechota said publishers are “now entering the final most advanced stage of AI data automation where actually it’s not only about dynamic payrolls, it’s not only about the dynamically managed customer journeys. We are talking about automated communication via chatbots via help in creating marketing assets.”

The advantages include being able to do more with fewer resources and hopefully, he said, sell more subscriptions while engaging and serving customers better.

One of the big concerns around AI has been that it will replace humans, but Piechota said it cannot function without a pair of human helping hands.

“You need a human touch to adjust, to edit, to revise, to fact-check the outputs of those AI systems,” he explained, noting the versatility and ability of the tools are “absolutely stunning.” Yet at the same time it demands oversight for quality control.

“It requires human intervention in the beginning and in the end. It is not truly creative. It is not plagiarising content from other books, but it is basically a language model that tries through statistics to imagine what would be the most relevant answer. But in the end it is not truly creative. You need a human touch.”

What AI can (and can’t) do

Thomas Davenport, professor of IT and management at Babson College in the United States, joined the Webinar in a pre-recorded interview to talk about the potential and limitations of generative AI, or AI that can generate novel content.

Davenport is co-author of the book All in on AI and was quick to assure that AI depends on quality guidance from humans to live up to its potential. AI tools such as ChatGPT, Writesonic, and Jasper AI are content generators that respond to prompts created by a human.

Those prompts offer the parameters of what information is needed and serves to guide the tool in its search. They can be used to write everything from social media posts and blogs to marketing e-mails and subscription offers.

“There is an important human component involved in effectively using these systems. You have to have a good prompt in order to get a good outcome,” Davenport said. “So there’s some kind of prompt engineering skills and maybe even entire jobs that will come about as a result of this.”

Additionally, the content and information generated need editing: “They need editing even more than humans do because both the formats and the factuality of the content is sometimes problematic.”

What a user is able to get out of generative AI tools is directly proportionate to what they put into it, Davenport said. As a result, writing prompts is emerging as a new field of engineering, and users also can buy prompts from an online marketplace.

News publishers would be wise to learn more about the science and skill of writing prompts, as “it will become an important aspect of either existing jobs or a new job.”

The future is here

To prepare for these changes, publishers should be experimenting with the technology and learn how to onboard it for use in their companies. It is a versatile tool that will aid marketing departments and newsrooms, and Davenport said there are three things it can do from a business perspective:

1. Create a new business model. This could be a new strategy or a new set of products and services.
2. Transform operations by improving such important areas as cost, efficiency, and productivity.
3. Influence customer behaviour and craft content that readers are looking for.

What’s important for media teams right now, he said, is that they keep senior executives informed about AI, educate themselves on how to use it, and understand what its potential and limitations are:

“Every company should be doing this today. We’re all potentially in the AI business now, and if you want to compete with these digital native companies — where really thinking about AI is pretty pervasive — I think education is the most powerful weapon.”

Artificial Intelligence, real-world results

Lukas Görög, data strategist at Die Presse in Austria, joined the Webinar to share some of the real-world applications for ChatGPT, showing how it could be used in different departments.

He echoed Davenport’s message that good outcomes depend upon good prompts.

“You can really make many things work when you define a good prompt,” he promised. “If you put good prompts at the beginning, you will have good results. [But] if you write a bad prompt, you might not be so satisfied with [your results].”

Görög shared examples of how he used ChatGPT for “assignments” across multiple areas:

Marketing

For a marketing campaign, Görög told the ChatGPT that he wanted it to serve as campaign manager and create five Google ads to promote subscriptions. He showed how he also included context to improve the outcome. In return, he received five ideas that were “pretty well tailored to our audience.”

For a second marketing prompt, he simulated a marketing manager and asked for 15 keywords to use in a Google ads campaign and was again pleased with the outcome. Another use in marketing, he said, would be to “explicitly ask to optimise the performance of the payroll of the newspaper page with some suggestions for three marketing offers.”

Newsroom

The newsroom offers many opportunities for leveraging AI, from creating story titles that are SEO optimised to tailoring topics for specific departments. “It saves a lot of time from the editor’s perspective,” he said. “It can create very tailored ideas where the audience is considered.”

ChatGPT can also be used as a community manager on social media, responding to hate speech “in a non-violent way” and reminding commenters of the guidelines. Görög also showed examples where he asked ChatGPT to serve as a social media manager and write captions for Facebook, LinkedIn, and Instagram. Each was written in the tone that was appropriate for the format.

Product

Product managers could get help in optimising welcome messages and creating welcome popups for new subscribers. “You can also ask ChatGPT for help, and it’ll give you a lot of ideas on how to do that,” he said.

Data

A data analyst wanting to analyse to a competitor’s content could write a prompt that includes the name of the competitors. Görög used a case where it looked at The New York Times. “I asked what I would like to improve and it generated nice results, which I can use for my next analysis,” he said.

As AI becomes more integrated into operations, Görög said he equates what is happening now with the introduction of computers in the 1970s: “In that case, everybody who was not able to control and to work with a computer couldn’t find a job later on. So I think we all should start learning how AI can help us improve our jobs.”

Greg’s Readers First newsletter is a public face of a revenue and media subscriptions initiative by INMA, outlined here. Subscribe here.

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