Can we talk today about a couple of issues that pop up frequently in my discussions and Ask Me Anything sessions with INMA members? As regular readers of this newsletter will correctly guess, they are both about how best to use AI.

One is the question of which projects to pick when it comes to using AI. There are many to choose from, so how do you get started on something that will not prove jarring to your audience? We present an intelligent framework for this.

The second is the question of calculating return on investment (ROI) when it comes to AI use cases. News industry leaders will swear up and down they do not want to prematurely snuff out experimentation by focusing on ROI — yet, those who are a rung or two down the ladder tell me that they absolutely need to include that when they make a business case for any AI project. We shed some light on what we’re doing wrong when it comes to ROI.

Sonali

A playbook for including audiences in AI strategies

How does a news brand go about picking an AI project to pilot? Most are (rightly) worried about losing their audience’s trust by building something that alienates exactly the people they are trying to reach, connect with, and monetise.

There is often a real disconnect between how excited news organisations are about using GenAI and how news consumers feel about it. This terrific chart on audience interest from the Reuters Institute for the Study of Journalism at Oxford University illustrates this well:

As you can see, news brands are planning all sorts of AI-related personalised consumer-facing products — but their audience seems to be saying they are less certain about them.

To be sure, audiences themselves often do not really know what they want. As many who have run a pilot of user-driven personalisation have seen, people pick topics they think they want to read about and then … don’t read about them. 

As Henry Ford is famously believed to have said, “If I had asked people what they wanted, they would have said faster horses.” 

Still, I was intrigued to find a newsroom leader who actually approaches the question of which AI projects to undertake by first considering what the audience wants rather than what the business or the newsroom deems important. It seems like the logical sequel to the user needs model many newsrooms are now considering and implementing.

Aldana Vales, audience experience director at Gannett, developed a playbook for including audiences in AI strategies as her capstone project at the Executive Program in News Innovation and Leadership at the City University of New York’s Newmark J-School.

Most AI strategies are still top-down, technology-led, and tool-based, she pointed out. Instead, news organisations should consider an audience-centric approach.

First, undertake an audience listening process. Conduct a survey of your news consumers and ask them how they feel about AI. 

Understand where they are — and that there might be varying mixes of nervousness and excitement.

• The first stage is usually pessimistic and involves consumers who have little direct experience with AI and get their knowledge from science fiction or news headlines. 

• The next stage is skeptical. “These users have tried AI, but the results fell short … . These moments confirm their belief that AI is overhyped and unreliable,” Vales said.

• As they grow familiar with the technology, they become cautiously engaged. “They know the limits, but they see the value.”

• And then finally, they grow supportive of the technology as they come to believe that AI is helping, not replacing, journalism.

Then, design the right features according to where your audience is:

Vales shared advice for each type of audience:

• “If your audience is mostly pessimistic, don’t start with the flashiest AI tool. That’s likely to trigger even more fear or rejection. Instead, begin with low-risk assistive tools like transcription, translation. And be transparent. Build trust first. Here, you’re earning the right to go further.”

• “If they’re skeptical, you need to show them the point. Start with use cases that offer clear, tangible benefits. These users haven’t been convinced so focus on functionality.”

• “If you find engaged users, this is your chance to experiment. Try bold pilots: Q&A bots, conversational explainers. Invite them to shape these features.”

• “And if you find supportive users, congrats. These users trust the newsroom and see AI as a way to deliver more relevant, accessible, or powerful journalism. So deepen that relationship. Ask how AI can serve them better.” 

In addition to this thoughtful step-by-step guide, Vales has excellent advice: “Don’t think about this as a checklist. Think of this as a starting point to move from uncertainty to strategy when thinking about how to adopt AI.” 

ROI and AI: Why is this so hard?

I have seen a fair amount of hand-wringing about the question of calculating ROI when it comes to AI.

Leaders consider it imperative to use the technology, while team members are a mix of enthusiastic, skeptical, and afraid in varying proportions across the industry. This leads to spotty adoption — which means any cost savings are patchy and any ROI is not clear.

How can one wrap their head around this, then? 

I came across some useful advice recently suggested by Gartner, a company that does technology research and consulting work, which pointed out that fewer than 15% of the enterprises that have adopted AI have actually created any value.

Productivity does not equal value. Time saved is not money saved. 

“If you save 20% of a person’s working day, there is productivity leakage. And it is not until they use that new time they have got to produce revenue or to fundamentally change business processes across multiple full-time employees (FTEs) that you are actually going to get savings,” said Gartner’s distinguished Vice-President/Analyst Frances Karamouzis.

“We’re telling people to shift their focus from focusing on FTEs to focusing on processes … . Start planting the seeds for this because processes and the real benefits that come are from increasing the business outcomes — either because you reduce the latency of the process, you reduce the cycle time, you reduce the error rate, you somehow created some efficiency or some kind of efficacy, or you’re now creating new business agility.”

The broader the use case, the harder it is to show value. For example, rolling out an AI assistant across the company is a broad use case. It will not be equally useful to everyone on staff, and everyone will not use it the same way. 

Deployment does not equal adoption. Enterprises underestimate the change-management effort that comes along with AI implementation. 

“If you don’t focus on change management — changing their behaviour, changing their performance metrics, changing a lot of things about the workflow — the process you will enter into this sort of realm of what we call the inevitable sort of brick wall. And this brick wall is where we see a lot of clients now, where they started doing some efficiency based projects with AI, and they’re sort of struggling, first of all, to calculate or recognise value, but secondly, they’re struggling to get to the next level of, what’s the next use case? What’s the next big thing I should work on?” Karamouzis said.

Most companies are looking for efficiencies with AI. They are simply trying to maintain competitive parity or do basic work that is considered “table stakes” rather than using AI to gain a differentiating competitive advantage and gain market share or undertake game-changing business transformation. 

They are aiming only for low-hanging fruit and incremental improvements rather than trying to expand their revenue, market share, or profitability. 

This is not a place where it is easy to create and calculate value, though. Any return on investment here is highly correlated with execution.

Companies should instead create a portfolio of use cases where 80% should be in the category of creating competitive advantage and differentiation, with 5%-15% each in the categories of table stakes and business transformation, Karamouzis said. 

“If your focus is ROI, you want a large preponderance to (that category).”

Why don’t more companies do this? Because these use cases require collaboration across an organisation through multi-disciplinary, cross-functional teams, and it is easier to operate in siloes. 

This means that the portfolio of use cases instead ends up being 80% in basic table stakes work, she said.

In that case, companies should think about their AI initiatives in terms of “return on employee” instead, Karamouzis said, which means taking into account new digital dexterity skills or lower attrition rates among staff due to higher satisfaction with work instead of purely financial benefits.

Worthwhile links

• GenAI at Axel Springer: German news giant Axel Springer plans to double its value in five years with AI.
• GenAI at the BBC: The British Broadcasting Corporation is looking into publicly testing two GenAI pilots.
• GenAI and bots: Cloudflare is letting publishers charge AI bots for crawling their sites for training data. 
• GenAI and vibe coding: You no longer need products to scale, says Semafor editor Gina Chua. 
• GenAI and data points: Interesting article that points out that many of the metrics we are using are no longer relevant in a zero-click world.
• GenAI and revenue: The new flex: A low percentage of your business tied to ads on Web pages.
• GenAI and brainstorming: You still need to use your brain to come up with good ideas because everyone using GenAI for that has similar ideas.
• GenAI and optimisation: SEO is old hat. Companies are now helping brands ensure answer engines are picking up their content.

About this newsletter

Today’s newsletter is written by Sonali Verma, based in Toronto, and lead for the INMA Generative AI Initiative. Sonali will share research, case studies, and thought leadership on the topic of generative AI and how it relates to all areas of news media.

This newsletter is a public face of the Generative AI Initiative by INMA, outlined here. E-mail Sonali at sonali.verma@inma.org or connect with her on INMA’s Slack channel with thoughts, suggestions, and questions.

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