A practical AI strategy framework — Beyond the hype
January 15, 2026
AI-Built Systems — Shipping Faster with AI
January 24, 2026
This is part of the blog series, “A practical AI strategy framework — Beyond the hype”.
AI powered experiences is where most companies start their AI journey — since the ROI feels immediate. This is about AI helping users write, summarize, search, or make better decisions inside a product. Say, you are using a Customer Engagement product to send a Halloween SMS campaign, and AI helps you write copy for the message. This was fairly novel in 2023, aka an ‘excitement attribute’, but has quickly become a ‘threshold attribute’ — talk about rapid decay of delight. Nowadays, it is fairly standard for users to expect AI assistance embedded throughout the product experience. And it is not just about improving text, but picking templates, smart approvals or generating multimedia content.
The pace of AI innovation is unrelenting. What felt innovative six months ago is now table stakes. I’ve seen teams spend months building AI features that users try once and never touch again. And in almost every case, it comes down to the same three mistakes.
First, the AI solves a problem the user doesn’t actually have. It was built because it was technically possible, or the company wants to push AI adoption — not because users had a problem to solve. You may — or may not — recall Microsoft’s ‘Recall’ feature (pun intended). It was supposed to be a big deal for Copilot+ PCs, but now is opt-in only. Recall was built because modern NPUs (Neural Processing Units) finally made local real-time image processing possible. So, Microsoft deviced a feature that takes continous screenshots of your digital activity, and make it searchable or “recallable” with on-device AI. While this was great technically, the “creep” factor was an overwhelming problem. Users needed better search, not 24×7 surveillance. After strong backlash, this feature was re-engineered and made opt-in only.
Second, the AI requires too much handholding. If it takes longer to correct or process the AI’s output than to do the work yourself, users will quietly abandon it, or may actively resent it. My pet peeve is the UX pattern of “tell me what you want and I will generate it for you”. The products typically say, ‘tell me about your campaign’, ‘tell me about your design’, ‘tell me about your report’ — but mostly what’s generated is just-not-good-enough! Then you need to go back to the regular creation workflow, take bits and pieces from what was generated — and try to cobble together what you want. It creates non-trivial expectation-reality Gap, triggering negative confirmation, where the user actively resents the product for failing a promise. This typically leads to lower user satisfaction compared to the feature not being there.
Third, the AI isn’t integrated into the real workflow. It lives in a separate tool, modal, or screen — breaking momentum instead of enhancing it. Look at AI video editing: many tools interrupt you constantly to approve a clip or pick a background, forcing you to ‘manage’ the AI instead of creating. The better approach is systemic AI. You define your brand profile upfront, record your content, and let the AI act as an invisible producer in the background. It doesn’t ask for permission at every step; it stays out of the way so you can stay in the flow.
The teams getting this right do a few things consistently.
They start with a real user pain point. They prototype quickly to see if AI actually improves outcomes. And they embed AI directly into the flow where users are already working.
AI-powered experiences have officially graduated from ‘wow factor’ to ‘utility.’ Winning teams don’t build AI to show off their technical prowess; they build it to disappear into the background. They know that if a user has to stop and ‘manage’ the AI, it is not going to be very useful.
Stop asking what AI can do, and start asking what your users hate doing. Solve that powerfully using AI, stay out of the way, and you’ll find the ROI everyone else is just chasing.
In my next blog, I’ll look at the engine room: AI-built systems — and how teams are using AI to ship products at a speed that was impossible six months ago. Happy exploring — wishing you a steady flow and zero friction on your AI voyage 🖖
