“The success of an AI product depends on how real users can interact with its capabilities”

In the Author Spotlight series, TDS editors talk to members of our community about their career path in Data Science and AI, their writing, and their sources of inspiration. Today, we are happy to share our conversations with them Dr. Janna Lipernkova.

Dr. Janna Lipernkova is an AI expert, entrepreneur, and author The art of AI product development. With a PhD in computational languages, you combine a deep understanding of technology with business strategy to help organizations turn AI into tangible results. Janna has founded and led many resources in language communication, data, and intelligence – including Anacodefocused on the evolution of business AI, too Of course you don'tan AI platform that supports business agility. Through his thought leadership and consulting activities, Janna continuously shaped and refined his comprehensive approach to AI development and integration.

He calls “AI strategic play” a set of mental models that help teams agree on what to build and why. What models often open to clarify in the upper rooms, and why they meet?

One of the biggest challenges in senior living is communication. People mean different things when they talk about AI, which prevents assassination. I use three mental models to create an orderly common ground that allows us to move forward without excuses and misunderstandings.

I usually start with AI At At Treewhich helps us map the world of AI use cases. Managers often come in with a combination of curiosity and curiosity – hype – “We need to do something with AI” – but not a clear vision of where the value actually lies. The default method that many groups take when building a chatbot, but these projects rarely take off (cf. This document). Opportunity Medicine breaks this mold by systematically finding AI use cases and providing a systematic, objective basis for prioritization.

As soon as we have clarity on what and why to build, we move to How and fill out AI System Blueprint. This model helps map the details, models, user experience, and governance issues of the considered AI system. It has great potential in multiple stakeholder areas, where business, data science, and compliance teams need a shared language. Blueprint turns the complexity of AI into something tangible and present – we can draw it, talk about it, then analyze it and clean it up.

Finally, I present AI Solution Space Map. It increases the discussion that goes beyond today's big technology – especially large models and agents – and helps groups looking at the complete area of ​​types of solutions, from classical ml to hybrid structures, to recover programs based on recovery. This broad perspective keeps us committed to delivering the right solution, not just fashion.

Together, these models create the journey that governs where successful AI products come from: Discovering opportunities, program design, and continuous testing. They agree with the management because they block strategy and execution.

According to your writing, domain technology is important in building AI products. Where have you seen domain knowledge change the entire nature of an AI solution, rather than just improving accuracy in games?

One clear example where domain technology reinvented the solution was a Logistics project that initially started predicting shipping delays. When domain experts joined, they repeated this problem: Delays were not random events but signs of serious business risks such as deep risks such as provider pain, or network overload. “AI experts” could see patterns.

To incorporate this background information, we extend the data layer beyond travel times to include supply hazard signals and dependency graphs. The architecture of AI evolved from a single prediction model to a hybrid system that combines prediction, knowledge graphs, and rule-based reasoning. The user experience was extended from predicting returns to risk situations with suggested reductions, which could be more effective for professionals.

Ultimately, domain knowledge not only improved accuracy, but reframed the problem, the design of the program, and the value of the business acquired. It turned the AI ​​model into a real decision support tool. After this experience, I always insist on domain experts who join during the early stages of the AI ​​program.

In addition to his post on TDS, he also wrote a book: The art of AI product development: delivering business value. What were the most important trends that changed your approach to building AI products (especially anything that surprised you or overturned a previous belief)?

Writing this book prompted me to think about all the pieces of theoretical knowledge, practical experience, and my conviction and it was good into a practical framework. Since the book needs to stay relevant for the ages, it also forced me to distinguish between the basics on the one hand, and the hype on the other. Here are a few of my lessons:

• – It's the headI learned to find business value in technology. Often, we live between midnight oscounts – either chasing AI for AI's sake, or relying solely on user-driven discovery. In the first case, you don't make a real value. In the second case, who knows how long you will have to wait for the “perfect” Ai Ai problem to come to you. In fact, the sweet spot lies in the middle: You use the unique power of different technologies to unlock the value that users have not heard, but cannot say from the great importers like Steve Jobs and Henry Ford, who created new experiences and asked. But to do this successfully, you need that magic of technical expertise, courage, and understanding about what the market needs.
• – LekayoI realized the importance of user experience for the success of AI. Many AI projects fail not because the models are weak, but because the intelligence is not clearly identified, defined, or implemented. The success of an AI product depends on how real users can interact with its capabilities and how much they trust its results. While writing this book, I was revisiting design classics, such as Don Norman's The Design of Everyday Objects, and I always asked myself – how does this apply to AI? I think we are in the early stages of a new UX era. Chat is an important part, but it's really only part of the complete equation. I am very happy to see the development of new user concepts such as productive UX.
• -ThreeAI systems need to evolve through cycles of feedback and improvement, and that process is never truly complete. That's why I use the metaphor of the basic person in the book: Brushing, soaking, reading further. Teams that communicate freely and iterate frequently tend to deliver much higher value than those that expect a “perfect” model. Unfortunately, I still see many teams that take too long before delivering the first foundation and spend not enough time doing it right. These programs may make it into production, but adoption will not happen, and they will be banned like other AI experiments.

For teams deploying the next AI feature quarter, what practices would you recommend, and what are the key pitfalls to avoid, so they can stay focused on delivering real business value rather than chasing real business hype instead of rushing?

First, as above, great is the art of Iteration. Post early, but do it profitably – release something useful enough to gain user trust, then promote it unfairly. Every interaction brings you new data, and every piece of feedback is a new sign of training.

Second, keep a broad perspective. It's easy to get an idea of ​​turnen around the LLM or release model, but the real innovation often comes from the way you combine technology – Retrieval, Reasoning, and Domain Logic. Customize your plan to deliver, and continuously monitor AI solutions and potential developments (see also our upcoming AI Radar).

Third, test with real people early and often. AI products live or die by how people perceive and use them. In-house demos and hands-on testing can't replace the messy, surprising installation and feedback you get from real users.

Your long-form writing (book, deep sweat) avoids hype and institutions in bringing value to organizations. What is your method of choosing topics and writing about these topics helps you to understand better?

Writing has always been my way of thinking out loud. I use it to learn, process complex ideas, and generate new ideas. I usually go with my gut and write in ways that I truly believe and have seen work in real organizations.

At the same time, in my company, we have our own “secret sauce.” Over the years, we have developed an AI-driven system to monitor new trends and innovations. We serve a few select customers in industries such as aerospace and finance, but of course, we also use it for our own purposes. That combination of data and intuition helps me see topics that are relevant now and may be more important in some months, but also two or three years down the line.

For example, at the beginning of 2025, we published a report on the state of the AI ​​business, and almost all the predictions from it have been carried forward like the whole year. So, while my writing is accurate and personal, it is put into evidence.

To learn more about Janna's work and stay up to date with the latest articles, you can follow her on TDS, the full site, or LinkedIn.