Turning Ideas into AI Use Cases: The Product Manager’s Perspective
The following contribution comes from Medium and is authored by Dilyana Evtimova.
Given the proliferation of artificial intelligence tools and functionalities, any Product Manager (PM) will sooner or later have the opportunity to work on an AI feature. Here are five things I learned when working on an AI use case for the first time:
1) Relax. While AI technology is new, user needs are not.
When I was assigned to work on AI, I got very excited and overwhelmed with information about this new technology:
I started reading about AI every day (the TLDR AI newsletter is quite good);
I contacted friends who work in AI and asked them many technical questions;
I enrolled in some AI courses (Machine Learning Foundations for Product Managers on Coursera offers a good overview, although it doesn’t cover practical product management much);
I watched numerous YouTube videos (the most stimulating was “AI and the Future of Humanity” by Yuval Noah Harari).
Gathering information from diverse sources and formats helped me catch up quickly, but it also made me realize that while AI technology is new, user needs rarely change overnight.
This understanding allowed me to focus on learning from the perspective of the user needs we wanted to meet and the business value we sought to generate, making my learning experience much more practical and memorable.
From then on, as the project progressed, whenever I encountered a new or unclear situation, I made sure to consult with the brilliant data scientists, project managers, and engineers at the FT. Special thanks to David Djambazov, Ares Kokkinos, Krum Arnaudov, Matteus Tanha, Evgeni Margov, Ognyan Angelov, Desislava Vasileva, and others, who were always willing to teach me and share their knowledge.
2) Start by solving a problem, not by creating a model.
No user wakes up with a desperate need to use an AI tool. As the saying goes, “Nobody wants a drill. They want a hole in the wall.” Many organizations rush to use AI in their work or in their user products. This responsiveness of companies to new technologies is admirable; however, there is a risk that it will lead to the creation of unnecessary AI features.
If you are not creating an AI-based solution and there is a way to validate your idea with a simpler product testing technique (such as false door testing, conditional statements, etc.), this could provide you with valuable insights more quickly. In my experience, we focused on AI use cases to scale successful features we had created, but we didn’t start with an AI model.
For example, we interviewed students to understand why they weren’t claiming their free subscription to the Financial Times in the US.
It turns out that many FT readers found the articles jargon-laden and difficult to understand. So, we created the FT Definitions feature in collaboration with our editorial team, which compiled 200 definitions of business and financial terms. This feature increased readership among students by up to 86% among those who opened at least one definition (and 25% did).
While the feature was successful, it was nearly impossible to scale by manually writing and updating the definitions. Therefore, the team explored the possibility of using AI to scale it. In short, our process was this: we ended up using AI, but we didn’t start with it.
3) Ensure you have the necessary user data to power your AI feature.
Needed data:
Write on Medium
How is tracking currently handled in your product?
Where is it stored?
How accessible is it?
What is the right dataset for a specific problem? These are questions you’ll address in every meeting.
In the initial phase, we typically conduct a technology exploration to understand the simplest way to develop functionality based on the problem we want to solve. For an AI use case, there’s an additional step: exploring the data tracking (or «data features» in AI terminology) we have implemented and understanding whether it’s sufficient to address the user problem we’re trying to solve.
For example, we recently launched a playlist version with manually curated content in the app.
In addition to manual selection, we looked for ways to personalize the playlist each user receives in FT.
The first step was to review the audio data we already collect in FT. It turned out we had quite a few (for example, how many articles users listen to on average; where they start listening; whether they pause or skip articles). However, we discovered we were missing data on «listening quality»—data that would ensure someone is actually listening meaningfully, not just hitting play and pausing immediately afterward.
Not having this information meant we lacked a universal success metric for optimizing model training. Since models go through many rounds of A/B testing and feedback loops, having a single metric like this is important for the team to stay focused on a single goal.
4) Train, Feedback, Retrain, and Repeat
Deciding which data features to include in model training is, in a way, like hypothesizing that optimizing user behavior will lead to a better model outcome. An AI model has a vast number of data features. Therefore, I discovered that developing an AI feature involves making far more assumptions than the typical development of a single feature. This means incorporating more internal feedback and qualitative research into the product development process than in a standard feature. For a news organization like the Financial Times, I believe editorial feedback is important to ensure AI features better reflect the FT’s identity and preserve our unique brand.
Furthermore, I found myself seeking more qualitative feedback from users, as AI experiences are so personalized that a variety of users are needed to test them and share their opinions to ensure their needs have been met. This is very different from having a limited number of conditions and ensuring you observe users as they go through each user flow.
5) Record your assumptions from the start.
Given the large number of assumptions made when training an AI model, it is crucial to record them and list the possible iteration scenarios for the feature. I recommend starting to record these assumptions from the very beginning of the AI use case. This will not only serve as a reminder for you, but it will also help the analytics team analyze the feature’s performance and know which performance data to focus on.
For example, we observed that the AI playlist model (built in collaboration with an external agency) was generating UK-centric playlists for US users. Rather than rushing to fix it, we realized it was a mistake to assume that Americans wouldn’t be interested in news about British companies. Therefore, we kept the model as is and will validate later whether it performs equally well in different regions. If it doesn’t, we’ll consider how to add geographic weight to the playlists, but it’s not a definitive statement that we need to improve before conducting any tests.
It’s incredibly helpful to be part of an organization like the Financial Times, where we have the ability to quickly gather user feedback, whether through A/B testing or interviews. While not a skill specific to an AI organization, it’s critical for developing effective AI models.
In conclusion, working with new technology can be an exciting experience, so enjoy it and don’t rush things. Be sure to ask plenty of clarifying questions of your team and users, as this will be invaluable when developing new AI features!
3 Ways to Leverage AI to Boost Creativity in Your Business
The following contribution comes from the CO US Chamber of Commerce website, which describes itself as follows: CO— is the U.S. Chamber of Commerce’s digital platform for small businesses and is dedicated to helping entrepreneurs across the country start, manage, and grow successful businesses. We provide timely and practical information and resources for entrepreneurs at all stages of growth through specialized content, exclusive interviews with business leaders, and virtual and in-person events.
The article is authored by Erica Sweeney, a contributor.
According to a new study, this technology can function as a creative collaborator. Here’s how to integrate it.
When you think of artificial intelligence, you probably think of its ability to automate tasks or generate quick answers to questions. But new research suggests that this technology can boost your creativity by acting as a personal collaborator.
Creativity is essential for small business owners.
However, the hectic pace of running a business often makes brainstorming difficult, or you may simply be experiencing a creative block. AI can be a tool to help.
Here’s everything you need to know about the latest study and how to integrate AI to boost your creativity.
The benefits of using AI to drive creativity
In the study, published in the Association for Computing Machinery’s journal Transactions on Interactive Intelligent Systems in October 2025, more than 800 people participated in an online experiment where they used an AI system to design virtual cars.
What can membership do for your business?
Get tools to stay informed, competitive, and connected by becoming a member of the U.S. Chamber of Commerce. Membership gives you direct access to expert policy analysis, economic updates, and exclusive resources designed to help your business thrive. From in-house analysis in Washington, D.C. From exclusive discounts to expert support, U.S. Chamber of Commerce membership helps you adapt to change and seize new opportunities.
Join today.
AI-powered tools generated virtual design galleries featuring a variety of potential car designs, including «high-performing examples, unusual ideas, and some deliberately imperfect ones,» according to a press release from Swansea University.
When participants were shown AI-generated design suggestions, they spent more time on the car design task, produced better designs, and felt more engaged, the study authors observed.
«It wasn’t just about efficiency. It was about creativity and collaboration,» said Sean Walton,
associate professor of computer science at Swansea University (Wales) and lead author of the study.
AI can analyze the data your small business generates, such as sales trends or customer feedback, and quickly develop new ideas for new products or new ways to attract customers.
How to Leverage AI for Creativity
As a small business owner, here are three ways to use AI to boost creativity:
Consider AI as a collaborator
AI works best when used to generate ideas that you can leverage and develop. According to MIT, AI enhances human creativity when it incorporates people’s own thought processes, planning, self-evaluation, and revision. AI should not replace these human functions.
Use AI tools, such as ChatGPT and others, to generate ideas for slogans or marketing campaigns, find ideas for a specific question, predict industry trends, or answer the question, “What’s the unique approach?” on a specific topic.
Overcome creative block
Small business owners need to be agile and constantly innovate, which often means tapping into their creativity. But when you’re feeling uninspired, AI can help you break through the block.
If you have general ideas for new services or initiatives, generative AI tools can help you refine them, according to Harvard Business Review. You can ask it to offer variations on a single idea—perhaps something you hadn’t considered. Or, when you need to write witty emails or social media posts, AI can take your basic ideas and create a draft. Of course, you’ll still need to review it to make sure it reflects your own style.
Leverage AI’s efficiency.
Automation is a common use case for AI, as it can analyze data and generate results quickly. This also benefits creativity.
For example, according to New York University, AI can analyze the data your small business generates, such as sales trends or customer feedback, and quickly develop new product ideas or strategies to attract customers. AI can also help small businesses efficiently write content, such as blog posts, marketing materials, or descriptions of new products. But again, be sure to reread and revise everything you create to ensure it’s accurate, relevant to your business, and reflects your personal style.
Seek inspiration from the experience of recognized experts. However, before making any business decisions, consult with a professional who can advise you based on your specific situation.
How AI Is Revolutionizing New Product Development
The following article is from Forbes magazine and was written by Michelle Greenwald, a former contributor. She is an expert in corporate, systematic, and creative innovation, specializing in product development.
When it comes to AI in new product development and marketing, we are in an early stage full of experimentation and important learning. To explore the latest and most promising use cases, I spoke with leading professionals from IBM, The Estée Lauder Companies, Inc., P&G, Smart Design, and GenexAI.
Mathematics, Algorithmic Magic, the Art of Asking Key Questions, and High-Quality First-Hand Data for a Competitive Advantage
AI Strategy at The Estée Lauder Companies (ELC)
With extensive prior experience in R&D, global innovation, and category strategy, Raheel Khan, Senior Vice President of Foresight and Growth Intelligence and current Chief AI Strategy Officer at The Estée Lauder Companies (ELC), is a pioneer in co-creating a wide range of cutting-edge AI and GenAI use cases and tools with partners such as OpenAI, Microsoft, Google, and Adobe. Raheel discusses the mathematics and magic required to successfully implement GenAI, as well as the art of asking the right questions to elicit answers and insights.
ELC leverages the valuable resource of millions of first-hand consumer conversations to extract valuable information, which, according to Raheel, gives them a competitive edge in their industry. The company has worked to create tools that democratize access to information so that more people within the organization can monitor it in real time. One example is their interactive AI trend detection tool, which identifies early signs of rapidly growing trends, such as the recent Scandi Girl makeup craze from Scandinavia that went viral. ELC’s tool monitors detailed metrics and relevant content at a local level.
It generates not only new product ideas but also valuable insights into how to optimize or modify existing products. Furthermore, it can inform the creation of advertising copy, product functionality, potential performance claims, the brand’s key message or slogan, ideas for packaging and visuals, and can uncover emerging customer segments. Predicting product stability is another area where ELC’s AI can significantly accelerate the process.
AI-Generated Hair Dyes
Other examples of product development in the beauty industry include hair dyes, whose colors are mathematically measured using AI. Dye combinations are then matched with color performance to generate millions of possible combinations. AI analyzes a vast array of possibilities based on brand objectives and identifies optio
