What can we learn from the book "Co-Intelligence" by Ethan Mollick about the future of decision intelligence products? 📚

The age of AI is going to bring us new challenges and opportunities

Apr 17, 2025

Hello friends 👋🏼,

I recently read Ethan Mollick’s great book Co-Intelligence, and I wanted to dedicate an article to it because it brought on a few thoughts.

The idea is to explore the things we can learn about the future of information and insight products through some of the areas mentioned in this book.

If you are curious to understand how generative AI will change our ways of working and beyond, I highly recommend reading this book. In addition, Ethan Mollick has his own Substack, “One Useful Thing” that you can sign up for.

Let’s dive right into some interesting insights.

The cover of the book Co-Intelligence by Etan Mollick.

🏁Introduction: Gen AI’s Role in Reshaping Information Consumption

The way we generate, analyze, and act on information is undergoing a major transformation. Ethan Mollick’s Co-Intelligence explores how artificial intelligence (AI) is not just a tool but a future collaborator in knowledge work. This could have some interesting implications for the future of information and insight products.

As of right now, we mainly see information and insight products as “analytical tools” or “software tools” but not so much as “analytical collaborators” or “data assistants”. That’s a major shift that could alter the existing ways of working around the development of information and insight products in software development teams and working with such products as users.

Let’s start to unpack it a little. Here’s my take on it.

1️⃣ - We’ll have to become “AI supervisors” at work with a new generation of decision intelligence products

Ethan Mollick describes how certain companies try to improve the outputs of generative AI models and reduce their bias with Reinforcement Learning from Human Feedback (RLHF). Many people earn their living nowadays on platforms such as Outlier.ai, where they can review and improve AI outputs. Ironically, sometimes, more bias is introduced in the process because the reviewers are just humans, and humans are, you guessed it, biased.

I think it’s a positive thing to try to improve the quality of generative AI models. I assume it’s probably easier on smaller and more carefully selected data sets. What it tells me is that one very possible outcome of the higher degree of AI use is that expertise in certain domains will be more important than ever, so we can assess the biases of information products that have a higher degree of AI algorithms embedded in them.

Yes, it’s easier to ask an AI-based information or insight product to spit out information in response to our prompts, but we as humans might need to have higher confidence in our human expertise to evaluate the output.

If you have a hard time imagining such a future, look into how pilots have re-learned to fly as machine supervisors ever since auto-pilot was invented. I could find some articles all the way back from 2011 that spoke about concerns that too much auto-pilot flying time eroded pilots’ skills to fly a plane manually.

I hope that in the future, information and insight products will allow for manual filtering and data exploration instead of leading users to conclusions in a way that could give them tunnel vision. It is totally up to us as product managers and designers to make sure this doesn’t happen.

2️⃣- What if we leveraged AI’s coaching capacity?

Ethan Mollick suggests a few ways for humans and AI to collaborate. One of them is to use AI as a tutor or a coach. This is an interesting opportunity for those working on developing decision intelligence products and customer journeys around them.

As of right now, I’ve seen a few embedded AI solutions that try to make you look into a particular area, such as the insights on Google Analytics (“Metric X is 30% higher than what it was last month”) or the assistant on LinkedIn (“You might be a good candidate for position X because of Y”).

What if we flipped the role of AI and made it more like a coach or a tutor that forces you to think together with them? I think that this could help those who work with information and insight products do a better job by suggesting areas to look into and explore in the data/information without leading them to conclusions right away. This could lead to an entirely new generation of decision intelligence products and a completely different customer experience than what we know today.

3️⃣- Data literacy is about to become more important than ever

These days, SQL is the “lingua franca” of most data practitioners. In the future, we might have more opportunities to query databases with plain language instead of code. When this happens, data literacy is going to become one of the most important skills for anyone working with data

In other words, having a good level of both prompt and data literacy is going to be crucial to working with future information and insight products. This way, we’ll be able to make the most out of these future products, filter, and explore data.

🔮 Conclusion: Be prepared for a shift in how we design, develop, and use new human-centred information and insight products in the future

The advancements in generative AI models could lead to new mental models and expectations around digital products in the future, including decision intelligence products.

Users will become more comfortable with AI-based solutions, and the power users of tomorrow are going to be the first ones to treat these solutions as digital collaborators, just like the type of co-intelligence described in the book by Ethan Mollick.

Therefore, it’s good to start flexing our muscles and imagine how AI-based decision intelligence products could look in this future and plan for the design of significantly different user experiences around them that will cater to “co-intelligent” mental models.

My hope is that we’ll be able to create experiences where humans and machines can double down on what they each are good at, and that our capabilities as humans won’t be reduced by machines.

I’d love to hear your opinion on the topic. Let’s start a chat in the comments below.

If you know anyone who works on decision intelligence products or is interested in this space, invite them to subscribe to Signal to Product at this link (signaltoproduct.substack.com) or share this post with them.

If a friend sent it to you and you’d like to learn more about this fascinating space of decision intelligence products, subscribe for free to Signal to Product now to get my articles, case studies, and more directly to your inbox.