Within the Writer Highlight collection, TDS Editors chat with members of our neighborhood about their profession path in information science and AI, their writing, and their sources of inspiration. Right this moment, we’re thrilled to share our dialog with Dr. Janna Lipenkova.
Dr. Janna Lipenkova is an AI strategist, entrepreneur, and writer of the ebook The Artwork of AI Product Improvement. With a PhD in Computational Linguistics, she combines deep technical perception with enterprise technique to assist organizations flip AI into tangible outcomes. Janna has based and led a number of ventures on the intersection of language, information, and intelligence — together with Anacode, which focuses on enterprise AI transformation, and Equintel, an AI platform that helps company sustainability. By way of her thought management and consulting actions, Janna is constantly shaping and refining her complete methodology for AI growth and integration.
You name your “AI Technique Playbook” a set of psychological fashions that assist groups align on what to construct and why. Which fashions most frequently unlock readability in government rooms, and why do they resonate?
One of many largest challenges in government rooms is communication. Individuals imply various things once they discuss AI, which blocks execution. I exploit three psychological fashions to create a structured frequent floor which permits us to maneuver ahead with out excuses and misunderstandings.
I often begin with the AI Alternative Tree, which helps us map the panorama of attainable AI use circumstances. Executives typically are available with a mixture of curiosity and hype — “we have to do one thing with AI” — however not a transparent view of the place worth actually lies. The default path most groups take from there may be constructing a chatbot, however these tasks hardly ever take off (cf. this text). The Alternative Tree breaks this sample by systematically uncovering potential AI use circumstances and offering a structured, goal foundation for prioritization.
As soon as we have now readability on what and why to construct, we transfer to the how and fill out the AI System Blueprint. This mannequin helps map the info, fashions, consumer expertise, and governance constraints of the envisioned AI system. It’s particularly highly effective in multi-stakeholder environments, the place enterprise, information science, and compliance groups want a shared language. The blueprint turns the complexity of AI into one thing tangible and iterative — we are able to draw it, talk about it, and refine it collectively.
Lastly, I introduce the AI Resolution House Map. It expands the dialog past at present’s dominant applied sciences — primarily massive language fashions and brokers — and helps groups take into account the complete house of resolution sorts: from classical ML to hybrid architectures, retrieval programs, and rule-based or simulation-driven approaches. This broader view retains us grounded in delivering the appropriate resolution, not simply the trendy one.
Collectively, these fashions create a journey that mirrors how profitable AI merchandise evolve: from alternative discovery, to system design, to steady exploration. They resonate with executives as a result of they bridge technique and execution.
In your writing, area experience is vital in constructing AI merchandise. The place have you ever seen area data change your complete form of an AI resolution, slightly than simply enhance accuracy on the margins?
One vivid instance the place area experience utterly reshaped the answer was a logistics mission initially began to foretell cargo delays. As soon as the area consultants joined, they reframed the issue: delays weren’t random occasions however signs of deeper enterprise dangers similar to provider dependencies, regulatory bottlenecks, or community fragility. We “AI consultants” weren’t in a position to spot these patterns.
To include this area data, we expanded the info layer past transit instances to incorporate supplier-risk alerts and dependency graphs. The AI structure advanced from a single predictive mannequin to a hybrid system combining prediction, data graphs, and rule-based reasoning. The consumer expertise was expanded from reactive delay forecasts to danger situations with advised mitigations, which had been extra actionable for consultants.
Ultimately, area data didn’t simply enhance accuracy, however redefined the issue, the system design, and the worth the enterprise obtained. It turned an AI mannequin into a real decision-support software. After that have, I at all times insist on area consultants becoming a member of in in the course of the early phases of an AI initiative.
Along with your posts on TDS, you additionally wrote a ebook: The Artwork of AI Product Improvement: Delivering enterprise worth. What are a very powerful takeaways that modified your personal strategy to constructing AI merchandise (particularly something that stunned you or overturned a previous perception)?
Writing the ebook motivated me to mirror on all of the bits and items of theoretical data, sensible expertise, and my very own conviction and construction them into reusable frameworks. Since a ebook wants to remain related for years, it additionally compelled me to tell apart between fundamentals on the one hand, and hype alternatively. Listed here are a few my very own learnings:
- First, I realized find out how to discover enterprise worth in expertise. Usually, we oscillate between two extremes — both chasing AI for the sake of AI, or relying solely on user-driven discovery. Within the first case, you aren’t creating actual worth. Within the second case, who is aware of how lengthy you’ll have to attend for the “excellent” AI downside to come back to you. In observe, the candy spot lies in between: utilizing expertise’s distinctive strengths to unlock worth that customers can really feel, however wouldn’t essentially articulate.We all know it from nice innovators like Steve Jobs and Henry Ford, who created radically new experiences earlier than prospects requested for them. However to do that efficiently, you want that magic mixture of technical experience, braveness, and instinct about what the market wants.
- Second, I spotted the worth of consumer expertise for AI success. Many AI tasks fail not as a result of the fashions are weak, however as a result of the intelligence isn’t clearly communicated, defined, or made usable. The success of an AI product will depend on how intuitively customers can work together with its capabilities and the way a lot they belief its outcomes. Whereas writing the ebook, I used to be rereading the design classics, like Don Norman’s The Design of On a regular basis Issues, and at all times asking myself — how does this apply to AI? I feel we’re nonetheless within the early phases of a brand new UX period. Chat is a crucial element, however it’s positively solely part of the complete equation. I’m very excited to see the event of latest consumer interface ideas like generative UX.
- Third, AI programs have to evolve by means of cycles of suggestions and enchancment, and that course of by no means actually ends. That’s why I exploit the metaphor of a dervish within the ebook: spinning, refining, studying constantly. Groups that grasp early launch and fixed iteration are inclined to ship much more worth than those that watch for a “excellent” mannequin. Sadly, I nonetheless see many groups taking too lengthy earlier than delivering a primary baseline and spending not sufficient time on iterative optimization. These programs would possibly make it into manufacturing, however adoption will seemingly not occur, and they are going to be shelved as one other AI experiment.
For groups transport an AI characteristic subsequent quarter, what habits would you suggest, and what key pitfalls ought to they keep away from, to remain targeted on delivering actual enterprise worth slightly than chasing hype?
First, as above, grasp the artwork of iteration. Ship early, however do it responsibly — launch one thing that’s helpful sufficient to earn consumer belief, then enhance it relentlessly. Each interplay brings you new information, and every bit of suggestions is a brand new coaching sign.
Second, hold a wider outlook. It’s straightforward to get tunnel imaginative and prescient across the newest LLM or mannequin launch, however the actual innovation typically comes from the way you mix applied sciences — retrieval, reasoning, analytics, UX, and area logic. Design your system in a modular manner so you’ll be able to lengthen it, and constantly monitor AI options and developments that might enhance it (see additionally our upcoming AI Radar).
Third, check with actual individuals early and sometimes. AI merchandise reside or die by how people understand and use them. Inner demos and artificial checks can’t substitute the messy, stunning inputs and suggestions you get from precise customers.
Your long-form writing (ebook, deep dives) avoids hype and facilities on delivering worth to organisations. What’s your strategy for selecting subjects and does writing about these subjects enable you to higher perceive them?
Writing has at all times been my mind-set out loud. I exploit it to be taught, course of advanced concepts, and generate new ones. I often go along with my intestine and write about approaches that I actually consider in and that I’ve seen work in actual organizations.
On the identical time, at my firm, we have now a little bit of our personal “secret sauce.” Through the years, we’ve developed an AI-driven system for monitoring new developments and improvements. We offer it to a few choose prospects in industries like aerospace and finance, however after all, we additionally use it for our personal functions. That mix of knowledge and instinct helps me spot subjects which are each related now and more likely to matter not solely in some months, but in addition two or three years down the road.
For instance, firstly of 2025, we revealed a report about enterprise AI developments, and virtually each theme from it has turned out to be extremely related all year long. So, whereas my writing is intuitive and private, it’s additionally grounded in proof.
To be taught extra about Janna‘s work and keep up-to-date along with her newest articles, you’ll be able to observe her on TDS, Substack, or LinkedIn.
