AI Can Help You Understand Your Audience. Here’s How.

Branislava Lovre:

Welcome to AImpactful. Newsrooms today have access to more audience data than ever before. But knowing the numbers is not the same as knowing the people behind them. Today, I’m speaking with Khalil, an audience development consultant who works with newsrooms around the world, helping them better understand their audiences and make more informed editorial decisions. We talk about audience relationships, the User Needs Model, and how AI and synthetic personas are changing the way newsrooms think about their audiences. Khalil, thank you for joining us.

Khalil A. Cassimally:

Oh, thank you. Thank you so much for having me.

Branislava Lovre:

Khalil, much of your work is about helping newsrooms connect better with their audiences. When we talk about audience connection today, what does that really mean beyond clicks, reach, and dashboards?

Khalil A. Cassimally:

To me, audience connection is when audiences, or people, feel like journalism sees them. I guess what I mean by that is when people feel that journalism is helping them understand what’s happening, feel something about what’s happening, and do something about what’s happening. It is when people feel that the context, the situation they are in, is being respected. I think all of that leads to that connection. And I suppose, from our perspective, so from the perspective of newsrooms or the industry, I think it really means more empathy, more compassion, more humility, and allowing ourselves to challenge our assumptions by listening better to people and better understanding the people we are meant to serve. So all of those things lead to that connection, that big thing that people feel.

Branislava Lovre:

You spoke about empathy and the importance of really listening to people. This is where the User Needs Model becomes so useful. Instead of asking only, “What should we publish?”, it asks, “What do people need from us?” What changes when a newsroom starts from that question?

Khalil A. Cassimally:

The User Needs Model is one method, one tool, that allows newsrooms to center themselves around audiences, around people. There are other tools as well, such as solutions journalism and change-centric journalism. But I think you’re right that rebuilding those meaningful audience relationships leads to quite a lot of good things. One of those things, which isn’t talked about as much as it probably should be, in my opinion, is the clarity it creates within a newsroom. For audience-informed, audience-centric, or audience-first newsrooms, the objective is actually quite clear: we want to serve the audience in whatever way we can.

When the objective is clear, there is clarity within the newsroom, and decision-making becomes much easier. We know that this clarity is important not only for efficiency, but also for sustainability. So I think all of that plays into the sustainability question, which is obviously very important. I’m talking a lot about humility, but that is something you see much more in newsrooms when they center themselves around audiences, around people. The need to understand people better, and the curiosity to learn about their needs, interests, and problems, become very important for newsrooms and news organizations, because that understanding allows them to address those needs and interests and find solutions to those problems.

And that allows them to be valuable. I think creating that value is a kind of North Star, a guiding line for many newsrooms. So, again, that alignment becomes very important. Everyone will be pulling in that direction. And, as I said, it becomes a necessity, a key element of sustainability as we move forward with AI.

Branislava Lovre:

For a newsroom that is just starting to explore AI, what is a simple and low-risk way to use it for audience insight?

Khalil A. Cassimally:

AI is a very powerful tool for helping people in the newsroom deepen their understanding of the people we are meant to serve. And it does that well. We can use AI to become better listeners. If we look at data analysis, for example, AI can help us conduct qualitative and other complex forms of data analysis, drawing on a wide variety of data sources and data types to unlock new insights.

But what are we doing when we analyze data? We are listening better. Data is information, and we are able to process information better. That is quite important. The shift from “we’re just looking at numbers” to “actually, we’re listening better to develop a better understanding” is a powerful one, and AI is really useful on that front. As I said, AI is great with different types of data. It’s not just numbers or quantitative data; it’s also qualitative data, which, for a long time, has mostly been disregarded because it is so complex to work with and analyze.

But AI does a great job of connecting behavior with meaning, because qualitative data helps explain why people are doing what they’re doing. So with AI, we can make more sense of qualitative data and better understand why people are doing what they’re doing, what their intentions are, and what their expectations are. And again, that all comes back to deepening our understanding of people, which then allows us to serve them better, which is what we want.

Branislava Lovre:

You’ve described AI as a way for newsrooms to look beyond the numbers and explore the needs, questions, and motivations behind the data. Synthetic personas take that idea a step further by creating AI-based versions of different audience groups. What exactly are they, and why are newsrooms beginning to experiment with them?

Khalil A. Cassimally:

Synthetic personas are essentially AI-generated representations of audience types and audience segments. They are informed by as much audience data as we have, including people’s needs, interests, and problems, as well as research insights and behavioral patterns. So they are representations created using the rich amount of data we have about our audiences. I would say there is a buzz around them in our industry now for several reasons. With ChatGPT, AI bots, chatbots, and so on, AI synthesis is much more accessible. That is one element contributing to the emergence of synthetic personas. We also have a great deal of audience data, which allows us to create rich personas.

That is another element. Finally, in many parts of the industry, understanding why people behave as they do is becoming increasingly important. Understanding people’s intentions and expectations is becoming more important because organizations are being proactive in creating value for audiences. I suppose those three elements combined are making synthetic personas a big thing in our industry.

Branislava Lovre:

Your own work has always been based on research with real people, and you’ve said that you were skeptical of synthetic personas at first. What changed your mind? And where do you think they can actually help a newsroom?

Khalil A. Cassimally:

I’ll put my hand up and say that, when I first started reading about synthetic personas, I didn’t really like the idea. I thought, “I’m generalizing here, but much of our industry is already disconnected from people. So if, instead of talking to people, we now talk to synthetic versions of people, that could disconnect us even more.” But as I started looking into it more, experimenting with it, and using it, I realized that it isn’t an either-or choice. You can do both, and both have their place.

For example, synthetic personas can add value when they help newsrooms ask better questions about the people they serve. When newsrooms stop asking questions, that’s a no-no. When we use synthetic personas to stop asking questions, that’s not great. One way synthetic personas can be useful is by helping us test gut feelings very quickly. We may have many ideas for a new product or specific features within a product, and we can now test those things very easily. For listeners who know design thinking or design sprints, the process is basically this: we think about potential problems from the user’s perspective, narrow down the problems we have identified, brainstorm solutions, and then narrow those down to a couple of options.

We create or implement some of those solutions as a test to see whether they work. Then we scale them or go back and revise them. With AI and synthetic personas, we can de-risk that entire process because, instead of using different methods to decide which problems or solutions to prioritize, we can run them through synthetic personas and get much more informed insights that allow us to make better decisions.

Branislava Lovre:

But there is another side to this. Where do newsrooms need to be careful? What are the biggest risks, and what should synthetic personas never replace?

Khalil A. Cassimally:

The most straightforward “no” would be using synthetic personas and no longer talking to humans. That would be a no-no, and a lot of risk comes from it: disconnection, false confidence because AI systems are very convincing in their outputs, and stereotypes embedded in training models. Those pitfalls do arise. But again, if we use synthetic personas to reinforce humility and curiosity, that is a good thing. Doing only synthetic research isn’t the solution. But combining synthetic research with human research, so that they complement one another, is something I think more and more people should at least experiment with.

Branislava Lovre:

So it is not really about choosing between synthetic research and human research. It is about understanding how the two can work together. When newsrooms try to bring audience insight, the User Needs Model, AI tools, and synthetic personas into the same process, what do they most often get wrong?

Khalil A. Cassimally:

When we think about AI tools, synthetic personas, and so on, I think there is an assumption that, ultimately, we will be able to do more. We will be able to do things more rapidly, which means we will be able to do more things. But, as I said, a deep understanding of audiences is about humility and curiosity. And it takes time; it takes work. The more we learn about the people we serve, the more we know and the more questions we have. It’s like that famous quote where you climb a hill and then see so many more mountains.

So the more we understand, the more questions we have and the more we need to uncover. Synthetic personas are tools for creating stronger audience connections, and there is an opportunity to see audience insight, the User Needs Model, AI tools, and synthetic personas as one system with a common objective. But it isn’t going to lead to less work. It could potentially lead to more work, but the end goal is to create a better understanding of people.

Branislava Lovre:

There are so many AI tools available now that it can be difficult to know which ones are actually useful. For journalists, editors, and product teams, how can they avoid chasing every new tool and decide which ones are worth exploring?

Khalil A. Cassimally:

This doesn’t apply only to AI tools. I think it applies to any tool. Tools alone aren’t helpful; they become helpful when they are used in service of an objective. So the way I approach this is to start by clarifying what we want to achieve. What is the objective? Then I move forward from there. With the objective fixed, I ask, “What data do I need in order to know whether we’re moving toward the objective or perhaps moving away from it?” Once we know what data we need, we work backward. Then we think, “What tools do we need to collect or synthesize that data?” That is a general way of thinking about the process of tool selection, and it certainly applies to AI tools as well.

As you say, there are so many AI tools. It seems like there are dozens of new ones every day, perhaps even more. You can’t keep up with all of them, and it’s very easy to become overwhelmed. But if you have a very clear objective, you can look for the AI tools, or other tools, that allow you to deliver on that objective.

Branislava Lovre:

You’ve said that teams should begin with the goal, not with the tool. Once they are clear about what they want to achieve, how should they choose an AI tool responsibly? What should they look at before deciding to use it?

Khalil A. Cassimally:

Maybe we can share a link in the show notes or something, but there is an article written a couple of years ago by two researchers that gives us a set of criteria, or a checklist, for external AI systems. They mention around ten factors to consider, including the quality of the training data, the quality of the model, ownership of the training data, data storage, liability, environmental impact, and so on. This is my go-to checklist, or set of criteria, for external AI systems. The article obviously goes into much more detail. But yes, I think it is really useful, and I would encourage people to use it at least as a template if they want to develop their own criteria. That is a good place to start.

Branislava Lovre:

Let’s make this practical. Which tools do you personally keep coming back to for audience research, synthetic personas, or data analysis? What do they help you do better?

Khalil A. Cassimally:

We talk a lot about synthetic personas, so I use a service called AskRally for them. It makes it very easy to build synthetic personas and cohorts. You bring in your data, put it into the service, and it creates the personas. You have a lot of control over how to change things, and then you can talk to the personas and the different cohorts as well. It also has some built-in personas, so you can use those, or at least use them as templates, and tailor them more closely to your own audiences or the personas you want to target. It is a very useful tool.

There is a free tier, and you can do a lot with it, so you can simply experiment. For other audience work, many other tools come to mind. I use ChatGPT quite a lot for data analysis, but there are others, such as Julius.ai. It is quite good for data analysis, especially in terms of its outputs. You can very quickly create visual representations of the analysis, such as graphs. Plus, if you have some knowledge of coding, you can do even more with it.

Branislava Lovre:

And beyond audience research and data analysis, are there other tools that have changed the way you work? Maybe tools that help with everyday tasks, note-taking, or even thinking about how you communicate and listen?

Khalil A. Cassimally:

For data analysis, there is also Julius.ai. It allows you to do quite a lot, especially when it comes to visualizing data and creating clear graphs very quickly. Plus, if you know some coding, you can do even more. It is really meant for people who are more data-proficient. So that is another tool. There are many other tools as well. I’ll mention n8n, which is really powerful. It is like Zapier, but for AI systems, in a way. It allows you to automate processes and add AI systems to a workflow. As a hobby, I post content on social media about Mauritius, especially environmental topics.

I have created an agent that, every week, searches for new scientific papers, academic research, and related material about Mauritius, the environment, the marine environment, and similar topics. It then sends me a Slack message with around seven articles and explains why it chose each one. Then I reply. I am the human in the loop, so I curate the list. I tell it that I am interested in articles one, two, and five. It then goes back and sends me the original papers, if available, for articles one, two, and five. I can go through those papers and create content based on them.

So with n8n, I have automated that process. It can be a useful tool for people in the industry, and perhaps for other kinds of work as well. My favorite tool is probably Granola, which is a note-taking tool. The way I use it, and I have written about this, is to take notes from meetings, which is always useful. But I also look at some of the transcripts from my coaching calls.

When I coach people, I look at the transcripts, anonymize everything, and then follow a process that I learned from researchers. Researchers have used this process, and it allows me to use ChatGPT, together with those transcripts, to improve my own communication and intervention styles. I think that also works for audience work because, if we look at the full cycle, it is about listening more closely to people. Yes, we have the tools, but we need to improve ourselves.

If we become better communicators and better listeners, we can listen better, understand better, and serve people better. By looking at those transcripts and using ChatGPT to help, I can see where I can improve in my conversations, in my coaching, and in the way I talk to people. I can try to act on that, and hopefully it becomes a virtuous cycle. That is one of the key ways I use it as well.

Branislava Lovre:

We’ve covered a lot: audience needs, editorial decisions, data, and different AI tools. But throughout the conversation, you’ve kept coming back to empathy, curiosity, and the importance of understanding people. Before we close, what would you like journalists and editors to remember as they try to understand their audiences better in this AI moment?

Khalil A. Cassimally:

My final message would be that more of us can take a step back, be more humble, more empathetic, and more compassionate toward the people we are meant to serve. We can then look at all the tools available to us, whether AI tools or other tools, that allow us to develop a better understanding of people. I think that is the only way we can really improve journalism: by developing a better understanding of people. That is the foundation for everything that comes upstream and downstream from it.

Branislava Lovre:

Khalil, thank you so much for sharing your experience and ideas with us.

Khalil A. Cassimally:

Thank you very much for having me. It was a pleasure.

Branislava Lovre:

Thank you for watching AImpactful. Follow us for more conversations about AI, journalism, and media innovation. See you in the next episode.