For many years, personalization has been one of the distinctive features of many successful products. Products capable of adapting to user needs had superior appeal for consumers and buyers in general. Companies creating personalized customer experiences were considered outstanding. They justified exceptional business valuations.
Nowadays, with AI, things are changing. Consumers do not consider personalization a premium attribute anymore. They expect it as a default option. They are used to open an AI app, ask a question, and get a personalized answer in a matter of seconds. Personalization has moved from being the exception to the norm.
So, many product companies are wondering how to improve their products now that personalization has become the new status quo.
As explained in a previous post, AI can support product companies with defining and shaping their product strategy. But AI can also be part of a product itself. Being part of the product does not mean by adding a new feature a front-end chatbot. AI agents can work in the back while the front-end provides the same functionality as before. The difference is that an AI agent can address specific customer needs and help customers with their specific job.
TL;DR: AI is moving products beyond personalization toward intelligence. The most significant shift goes beyond simply adapting products to individual users. It makes expert capabilities more accessible, simple, and more tailored to customer needs. This shift creates new opportunities to widen the target audience and, at the same time, to further differentiate products. Still, such a change requires product leaders to put more attention around trust, explainability, and product responsibility.
How AI enables intelligent products
Intelligent products use AI to make complex capabilities accessible to more people. They adapt not only to who the user is. They also adapt to what users are trying to accomplish and their context.
The key difference between a personalized and an intelligent product is:
- Personalized product: it adapts to the user and provides with possible options
- Intelligent product: understands what the users wants and helps them accomplish it
This means that personalization moves from being a front-end interface to becoming a back-end capability that can connect available resources to different outcomes, one for each user need.
Generative AI makes product intelligent not just because of the technical capabilities it enables. They add five specific product qualities: accessibility, simplification, guidance, learning, proactivity.
1) Accessibility
AI reduces competency gaps and malkes knowledge more accessible thanks to better sensing (photos make it possible to translate images to text), interpretation (large bodies of knowledge make it possible to interpret a user’s needs and map them to a possible solution), and option recommendation (suggesting all the possible options so that nothing is left unattempted).
Example: In the medical field, for example, image recognition and automatic interpretation allow pregnant people to take ultrasound images on their own at home. This does not substitute for regular medical checkups. It enables parents to hear the baby’s heartbeat, take pictures, and estimate gestational age and fetal weight whenever they wish or feel the need to connect with their baby.
2) Simplification
Superior products are simple. And AI can simplify user experience by tailoring interfaces, workflows, and information to specific user needs. The old mantra of designing UI according to users can now find its full application. After a few bits of information and a few clicks, an app can understand users’ characteristics and adapt its user interface, content, and information architecture accordingly. This is not just about adaptive UX. Yes, apps already can automatically change their layout, navigation, and features to fit any screen size, orientation, or device configuration. An adaptive app goes one step further. It can change information content andp resentation structure adaptively. As a consequence, users get a better experience. Information and interactions becomes simpler and tailored to users’ needs.
Example: A traditional accounting software might require users to understand accounting rules, know tax terminology, identify relevant forms, navigate complex workflows, and know where information belongs. An intelligent product can adapt the interface and workflow according to the user’s elvel of competence. It can ask a few questions, and fill the required fields automatically.
3) Guidance
AI moves up the bar of knowledge consumption. The internet made information easy to find. AI is making information easier to use.
A user might browse for an answer to their query. they could open several pages, compare them, interpret what they mean, and figure out what to do next. AI can compress much of that process and provide not only information, but also interpretation and guidance.
Instead of telling a user what something is, an intelligent product can help tusers understand what it means for their situation and what they might do next.
This changes the role of information products. The product is no longer simply a place to find information. It becomes an entry point to turn a need into information and finally into action.
Example: In property & casualty insurance, AI can guide customers through the claims process. It can help them understand what information they need, what photos to take, and which documents to attach to file a claim. Instead of simply providing an intake form or a list of requirements, the product can provide a single point of entry, interpret different customers’ situations and guide them through the specific steps needed to complete their claims. This turns information that was previously available in different policy documents and claims instructions into a smooth customer journey.
4) Learning
By adapting information to different users’ needs and levels of competencies, intelligent products enable more effective learning.
Learning is a strong motivator. When users receive information that is relevant to their current level, they engage more because they learn more. According to Nir Eyal’s Hooked model, learning encourages users to perform work that strengthens their engagement with the product.
Example: In language learning, AI can turn a traditional lesson-based app into a more continuous learning experience. Instead of simply completing predefined exercises, users can practice through conversations, receive immediate feedback, understand their mistakes, and progressively adapt the difficulty to their level. Each interaction helps users become more capable, which can reinforce their motivation to return and continue learning.
5) Proactivity
Traditional products wait for the user to initiate an action. Intelligent products, when properly connected to relevant data sources, can detect signals, process information autonomously, escalate when needed, and trigger further action.
This makes it possible to anticipate needs and intervene at the right moment. Historically, many event detection use cases were difficult to implement. This was because of the risk of false positives and false negatives. A false positive happens when the product raises an alert when there is no real problem. A false negative happens when the product fails to detect a real problem.
Intelligent products can take decisions with much greater accuracy and confidence. By reducing false positive and false negatives, they make proactive intervention increasingly practical. They can do it thanks to better data, more capable AI models, and the ability to combine multiple signals.
Example: An industrial equipment product traditionally tells operators when a machine has failed or when scheduled maintenance is due. An intelligent product can continuously monitor signals such as vibration, temperature, pressure, energy consumption, and operating conditions to detect patterns that indicate an emerging failure. Rather than waiting for the machine to break down, it can estimate the likelihood of failure, determine how critical the issue is, and proactively recommend or schedule maintenance.
Implications for user experience
Given the five different ways AI is changing products, what are the biggest implications when it comes to personalizing user experiences?
From services to products
When knowledge barriers are removed and information becomes more accessible and tailore to users, products can deliver value that previously could be delivered only through bespoke services. Apps can cast a wider net and provide value to customers who previously could not afford access to specialized expertise. This translates to a wider addressable audience and greater potential for conversion and retention.
Professional human expertise does not disappear. On the contrary, it becomes more focused on high-value, high-stakes, and highly personalized needs. Products can support the bulk of users while also becoming an additional segmentation layer for professional services, which can become more specialized and focus on customers with more complex needs and willingness to pay.
From point solutions to follow-along solutions
Apps and tools can create value not only by addressing different types of users, but also by supporting users end-to-end throughout their customer journey. They are no longer just point solutions; they become follow-along solutions.
For example, exercise coaching apps do not just create training plans or recommend which exercises to do. They can provide suggestions when they matter most: when you go to sleep, before training, or before and after a race.
This does not mean that products should try to do everything at once (training, nutrition, mental coaching, and so on). Doing so would likely overwhelm users with information. Instead, products can address specific user needs and objectives more holistically. It means addressing the entire user journey rather than only at specific touch-points.
From one-to-many to one-to-one experiences
AI enables products to move from program-based one-to-many experiences to increasingly personalized, one-to-one experiences.
Duolingo, for example, has evolved from a relatively standardized language-learning program toward a more personalized tutoring experience, adapting the learning experience to the needs and progress of individual users.
Andrew Ng, founder of Coursera, is also working on LearnVector, which aims to provide a more personalized learning experience to every student thanks to AI, so that anyone can upgrade skills at one own’s pace.
From human explanation to human validation
Even though AI can make expert-level recommendations accessible to a much broader audience, the need for human expertise does not disappear.
Intelligent products using AI to provide recommendations can offer a human-in-the-loop upon request. This can create a new layer of user segmentation.
Products can use AI to serve the needs of the majority of users while offering expert validation to those who want an additional layer of trust. For example, in high-stakes domains such as health, finance, or legal services, users could receive an AI-generated recommendation and, for a premium, have it reviewed, expanded, and “checkmarked” by an expert before acting on it.
In this model, products can target wider audiences at the bottom with AI and create additional value at the top with human expertise. Professional expertise becomes less about serving every user individually and more about providing an extra level of validation to users who want an additional degree of confidence as a premium feature.
New challenges for personalized products
While AI helps products to become more personalized, it also introduces new challenges. When designing intelligent products relying on AI, there are four aspects that need particular attention.
Knowing when to switch to human expertise
There are situations where an AI should hand over to a human expert. These triggers should not be entirely delegated to the AI without clear guardrails, as an AI may overestimate the reliability of its own recommendation, particularly when domain-specific knowledge is limited. One possible solution is to use a separate AI agent to evaluate predefined criteria and determine when a case should be escalated to a human.
Information bias and edge cases
AI systems tend to perform best on patterns that are well represented in their training data or in the information available to them while they are running. This reliance on a vast, yet limited, pool of data can make AI-based products less reliable when dealing with unusual or poorly represented situations. It then becomes important to have controls in place to recognize when a situation falls outside the domain the AI can reliably handle.
Outcome Explainability
Much AI research is geared toward making sure that its results are transparent and can be explained to humans. For products, however, explainability means more than simply exposing how the AI reached a result. Products need to present AI outputs in a way that is appropriate to the specific user and intuitive given their level of knowledge and understanding. Moreover, not all users want to know why an AI provided a particular suggestion or recommendation. Depending on the use case and the type of decision involved, the level and form of explainability needed can vary.
Product responsibility
As users increasingly rely on AI to make decisions, products need to establish clear boundaries around what can be delegated to AI and what remains the user’s responsibility. In other words, a product should not over-rely on AI guidance, but should keep the user in the loop where appropriate and act as a mediating layer between the user and the AI.
Otherwise, users may progressively offload too much of their judgment to the product and accept its recommendations without sufficient scrutiny. They may also assume that responsibility for those decisions lies with the product, even in areas where it does not.
Showing disclaimers to manage users’ expectations may not be sufficient. An intelligent product should clearly guide users on what falls within their responsibility and what is handled by the product.
Wrap-up
With generative AI, products change. The don’t just adapt to users’ requests. They start to understand what they need and adapting to their context. This creates new opportunities for product makers. They can make expert capabilities accessible to wider audiences. At the same time, they can focus human expertise where it adds the most value.
This post provides an opinionated view on what this transition means for product leaders. As there is no one-size-fits-all formula, the right approach depends on the specifics of the market and the product.
To go deeper into any of these topics, or to analyze how these ideas apply to a specific market or product, get in touch.
