Personalize portals in Liferay with real-time AI

ICX Consulting helps companies turn Liferay portals into experiences that adapt to each user. Personalizing portals in Liferay with real-time AI means adding a decision layer that reads user behavior as it happens and tells Liferay what to show, prioritize or hide, so every visit becomes more relevant and more profitable.

Picture the monthly operating review at a mid-size insurer. The digital director puts the portal dashboard on the screen and the numbers look healthy. Logins are up, the new design won praise and the project closed on budget. Then the chief customer officer asks the question nobody prepared for. If so many policyholders are logging in, why has the call center volume not moved, and why are renewals still slipping in the same segments as last year? The room goes quiet because everyone knows the answer. The portal greets a claimant in the middle of a stressful week with the same promotional banner it shows a prospect browsing for a quote. It is busy, but it is not useful.

That scene repeats across industries, because most enterprise portals are still built on an outdated logic, the same experience for everyone regardless of who is on the other side. This traditional approach assumes that a unified and homogeneous structure meets every use case, and in today's environment it is obsolete and disconnected from modern digital expectations. Companies that keep building portals under this paradigm face a growing disconnect with their users, who perceive these platforms as impersonal, rigid and unable to respond to the dynamic nature of their real needs.

The commercial stakes are no longer theoretical. McKinsey's research on personalization found that 71 percent of consumers expect companies to deliver personalized interactions, 76 percent get frustrated when that does not happen, and personalization most often drives a 10 to 15 percent revenue lift. For a company that already invested in a portal, that gap between expectation and delivery is the cheapest growth it has left on the table.

Why does a portal that shows everyone the same thing cost you money?

In a world where every user expects quick answers, relevant content and interactions that feel custom-made, continuing to show the same thing to everyone no longer makes sense. Users increasingly value unique digital experiences, where every touchpoint adapts to their motivations, moments and context. Ignoring this demand reduces conversion rates, weakens engagement and erodes brand value. Success today is built on personalized, flexible, customer-centric relationships, in which digital platforms stop being simple channels and become tailored experiences that drive trust, loyalty and competitive differentiation.

The cost shows up where the portal team rarely looks. In banking, a customer who cannot find the status of a loan application calls the branch. In distribution, a dealer who sees the full catalog instead of the parts that fit its machines orders less, or orders from a competitor whose portal remembers. None of this appears in the portal's analytics, yet each failure moves revenue, retention or cost in the wrong direction.

Personalization does not just mean changing a banner or updating text based on a profile. True personalization goes much further. It means designing experiences that evolve in real time, deeply understanding the user journey, anticipating needs at each stage and even surprising users with recommendations or proactive actions before they ask. It is about using behavioral data, context and history to dynamically adapt, prioritize and display content, processes and functionality. The goal is a digital environment capable of learning, optimizing and adjusting its logic as each user interacts, which turns the exchange into a living, adaptive and people-centric experience.

What does Liferay already offer, and what is missing?

Liferay has the foundation to achieve this, but something more is needed, a layer of intelligence that enables it to react. Out of the box, Liferay DXP lets teams define segments of users and create experiences for them. As the official documentation on content page personalization puts it, each experience contains the content and layout that is most appropriate for each segment, and experiences can also support A/B testing.

That is a solid starting point, and many companies have not even used it fully. While Liferay's flexible architecture, open APIs and fragment system already support a certain level of personalization, the real value is unlocked by integrating artificial intelligence, machine learning and real-time automation. This intelligent layer must analyze events, interpret context and make automated decisions about what to show, when and to whom, with zero friction and robust levels of security and performance. With this approach Liferay stops operating as a mere content delivery platform and becomes the core of a dynamic digital ecosystem, where every interaction drives value and strengthens the relationship. Technology then works not only as an enabler but as a strategic driver of innovation, scalability and differentiation in increasingly competitive markets.

>> Integrating Business Processes with AI in Liferay <<

Personalization in earnest, context and urgency

Personalization has evolved from a promise or an advanced feature into an expected standard. If a portal cannot adapt in real time, it loses relevance to more agile, user-centric solutions that respond without friction. Liferay provides tools for audience segmentation and content conditioning, but if that segmentation relies solely on static rules, it falls short of what real-time artificial intelligence makes possible. A static rule knows that a user belongs to the "premium" segment. It does not know that the same user just abandoned a payment halfway through and is about to call.

The real opportunity lies in adding a dynamic layer that interprets what the user is doing at any given moment, makes automatic decisions and adapts the experience on the fly, without reloads and without manual intervention. In one case, a bank using Liferay as its customer portal integrated AI to identify financial profiles in real time, displaying alerts to users with overdue products and offering proactive recommendations to those with investment capacity. All of this ran on the same backend, yet produced entirely different experiences depending on the context.

It is not about redesigning the portal but about adding a decision architecture that interprets events, consults an AI model and delivers immediate answers on what to display, what to prioritize and what to hide. This can be done with services such as Amazon Personalize, which AWS describes in its developer guide as a fully managed machine learning service that uses your data to generate item recommendations, with API operations for real-time personalization, with Google Vertex AI, or with proprietary models connected by API. All of them are fed by tracking tools such as Segment and orchestrated by middleware that manages the logic without affecting Liferay's performance.

In this way, the portal ceases to be a simple content container and becomes an experience built in real time from user behavior. That change makes all the difference, and it is the difference the insurer's chief customer officer was really asking about.

>> What is Liferay Portal and what is it used for? <<

How flexible is Liferay for customization?

Liferay has far more potential than is typically used. Its modular design, open APIs and fragment system enable much more than visibility rules or conditional templates. The default features are a good starting point, but truly compelling solutions emerge when decision-making is managed outside the portal, so that Liferay focuses on rendering what an external intelligence dictates.

Consider an insurance company that used Liferay as a customer portal and redesigned it so each user saw a unique dashboard based on transaction history. If a claim had been initiated, that information appeared first. If not, the focus shifted to promotions or new services. The key was decoupling the frontend rules and replacing rigid conditions with real-time API queries that delivered the ideal content structure for each profile.

In practice, this is achieved with dynamic fragments, templates built with FreeMarker or React, and embedded logic that queries an external source when the component loads. That source can be a microservice connected to an AI model, which receives the user's context and returns decisions on which blocks to display or which message to highlight. It is not about Liferay making complex decisions itself but about delegating this work to a system designed to learn, adapt and evolve, while Liferay presents what it is instructed to, efficiently and quickly.

A comparison between static and dynamic personalization helps executives see the trade-off. Static rules are cheap to start but expensive to maintain, because every new segment means another rule and another test, and the experience only improves when someone remembers to update it. Dynamic personalization asks for more architecture upfront, yet it improves with every interaction and scales without multiplying the editorial workload. Once that external logic is in place, the next challenge is making sure everything truly operates in real time.

>> Liferay for the digital experience <<

What does real time really mean in a Liferay portal?

Many systems claim to operate in real time, but in reality they respond with delays of several minutes or only in the user's next session. If the goal is a truly dynamic and immediate experience, that is no longer enough. The customer who abandoned a payment needs help now, not the next time they log in.

Real-time operation means detecting events instantly, analyzing them through a trained model and providing a visible response in less than a second. That requires a lightweight, distributed and bottleneck-free architecture, one that can process signals such as clicks, navigation, location or user intent and translate them into visible content without disrupting the user experience.

An internal benefits portal built on Liferay adopted this kind of system. When an employee accessed the portal on payday, the site highlighted savings promotions or relevant benefits. On other days the focus shifted to corporate news or internal tools. This was made possible by an event engine powered by Kafka that detected interactions, sent them to a model in SageMaker and received as a response the exact structure to display. That response was transformed into data Liferay could consume and render, all in less than 300 milliseconds and without page reloads.

The infrastructure included an event broker, a decision service and a caching system such as Redis or Cloudflare Workers for the most frequent responses. Everything remained outside Liferay's core, which made scaling and updates easy without touching the portal every time something changed. A simple flow diagram, from the user's action to the personalized experience on screen, helps clarify the whole journey, and understanding that flow is essential because that is where the real impact of this logic lies.

How can AI act as a content co-pilot?

One of the key differentiators of incorporating artificial intelligence into Liferay is its ability to go beyond simple block selection. AI can act as an active co-pilot, suggesting content, reordering elements, proposing new topics or detecting patterns the editorial team may have missed.

This applies to both internal and public portals. A training portal can use AI to recommend courses or articles based on each user's progress, interests and performance profile. A self-service site in telecom or utilities can prioritize help articles that match the most frequent questions within a segment, and it can detect intent to abandon and react in real time with an incentive, a question or an alternative support channel. AI can also work with the content team by suggesting new pieces based on user behavior or by producing small variations of the same message for different profiles. Instead of one version of a campaign, the team publishes variants aligned with each segment without increasing its workload.

This completely changes the way content is planned. Instead of linear campaigns, teams think in terms of content ecosystems that adapt and evolve with user behavior. AI does not replace the team, it enhances it with speed, context and responsiveness, and above all it enables scale without sacrificing personalization.

Why personalization starts with design, not code

Real-time personalization is not only a technical transformation. It requires a complete rethinking of user experience design. When content can change for each individual, design can no longer follow a rigid or closed logic. It must be modular, flexible and prepared to adapt without breaking.

That means working with reusable components, layouts engineered to scale with dynamic content and a visual hierarchy that adjusts to context without sacrificing clarity or function. The design and UX team needs a deep understanding of the personalization logic, so it can anticipate scenarios and propose solutions that accommodate many possible combinations rather than a single visual narrative. If a block can be replaced by an automated recommendation, the space must be ready to show a banner, a card, a list of items or a notification, and all of them must work together without compromising the overall experience.

The approach also affects navigation, calls to action and conversion paths. When every user sees something different, it is not feasible to plan a single flow. Several journeys must be designed, and every route must lead to outcomes that match the portal's objectives. In this landscape, design stops being a final stage and becomes part of the personalization strategy from the outset.

How do you scale personalization without losing control?

When personalization works, it is tempting to deploy it across more cases, segments and variables. Scaling without a clear strategy, however, creates its own problems. The more complex the logic becomes, the greater the risk that a poor recommendation damages the user experience or even violates business policies.

One B2B portal with multiple product lines saw AI-driven personalization improve form conversions while also generating recommendations that did not match current commercial conditions. For a sales director, that is the nightmare version of personalization, a portal promising a price or a product the sales team cannot honor. Avoiding such mismatches requires a continuous improvement cycle that monitors decisions, retrains models and validates impact against business metrics.

That cycle depends on automated pipelines that control how models are trained, deployed and deactivated, and on clear metrics that show what works and what does not. Tools like Datadog, Prometheus or Grafana help monitor performance, while MLflow tracks model versions and results. A/B testing makes it possible to trial changes without affecting the entire user base, and feature flags let the team activate or deactivate specific personalizations depending on outcomes, load or context. Routine maintenance covers log analysis, review of the data used to train models, validation of business rules and updates to critical thresholds. With that structure in place, scaling does not mean losing control. It means multiplying impact without compromising quality or system stability, and that is what turns real-time personalization into a business asset rather than an attractive feature.

What this means for the C-level

For the chief executive and the chief customer officer, the decision is not whether to buy another tool but which customer moments the portal should own. The best candidates are the moments where a relevant answer either keeps the customer or saves a costly contact, such as the claim in progress, the overdue payment, the renewal window or the dealer reorder. Each of those moments maps to a lever the leadership team already manages, whether it is converting a visit into a sale, retaining a customer who was about to leave, serving at lower cost or building the loyalty that turns a portal into a habit.

For the chief financial officer, real-time personalization deserves the same discipline as any investment. It should be funded against a baseline, measured with a control group and judged on revenue per active user, self-service completion and cost to serve, not on logins or page views. When the company already runs Liferay, the portal, the content and most of the data exist, and what is missing is the decision layer and the governance to run it.

For the chief commercial officer, the lesson is governance. The commercial team must own the rules AI cannot break, from pricing and eligibility to credit conditions, so the portal sells the way the company wants to sell.

Where to start, a strategy to evolve without friction

Implementing real-time personalization with AI does not have to be an enormous project. A seasoned team usually starts on a Monday morning with one question for the business owner of the portal, which single page or module would change a number the executive committee cares about if it became relevant to each user? The answer is usually one of the most sensitive touchpoints in the user experience, the homepage, the user dashboard, the search function or a recommendations module, and impact can be measured from the outset.

From there the team writes a concrete hypothesis, for example that prioritizing content related to the user's most recent action will increase conversion by 15 percent, connects that section to an AI model or a basic external logic, and measures the result against a control group. Pilots like this validate the value proposition, but they also let the team refine the technical architecture, establish maintenance processes and train people without risking the overall stability of the portal.

In parallel, someone needs to map the available data sources. Which events are already being captured, what user information can be used without compromising privacy or compliance, and which signals help build context? With that foundation the company can begin with simple models that evolve over time. It is equally important to define roles and processes early. Someone must review the outputs of the AI, someone must decide when to retrain a model, and the business rules that must always be respected need an owner, so automation does not create outcomes that contradict strategic objectives.

Real-time personalization is not only a technical endeavor, it is a cultural transformation. Like any transformation, it requires a clear roadmap, a committed team and an iterative approach that allows continuous learning without stopping progress. To structure the first pilot, the ICX Persona Playbook Toolkit helps define the segments and moments that matter, and the Information Architecture Toolkit helps design the modular page structure that dynamic content needs.

Liferay and the future of personalization

The technology is here, the tools are available and the demand for personalized experiences will only increase. If a portal is built on Liferay, there is no need to rebuild from scratch, but it is time to raise expectations. Integrating real-time artificial intelligence is not just a technical upgrade, it is a strategic decision. It is the way to turn a static portal into a living platform that responds to context, adapts to behavior and puts the user at the center of every decision.

Every interaction is an opportunity to add value, increase conversions or strengthen the relationship with users. That only happens when content stops being generic and becomes relevant, and when the experience stops being the same for everyone and responds to each person's needs in real time. If a portal still shows the same thing to everyone, it is time to rethink the approach, because context will not wait and neither will users.

How ICX helps

ICX Consulting starts where the insurer's operating review ended, with the customer moments that move revenue, retention and cost to serve. Using ICX Compass, our consultants identify which portal journeys deserve personalization first and define the business rules and KPIs that will govern them. Our enterprise portal and artificial intelligence consulting teams then design the decision layer around your existing Liferay platform, and UX analytics keeps every pilot measured against a control group, so each new segment is added because it pays, not because it is possible.

ICX can also assess whether your current architecture is ready to scale this approach without sacrificing what already works. Want to know which portal moments would pay back first? Request a diagnostic with an ICX consultant.

Frequently asked questions

Can Liferay personalize content without external AI?

Yes. Liferay DXP includes segments and experiences that show different content and layouts to different groups of users, along with A/B testing. These features work well for stable rules such as customer type or region. External AI becomes valuable when the experience must react to behavior within a session, such as an abandoned payment or a new claim.

Does real-time personalization slow down the portal?

It should not, if the decision logic runs outside Liferay's core. A well designed architecture uses an event broker, a separate decision service and caching for the most frequent answers, so the portal only renders the response. Teams should set a latency budget from the start and monitor it continuously, because a personalized page that loads slowly loses the benefit.

What data is needed to start personalizing a Liferay portal?

Less than most teams expect. Many pilots start with navigation events, the user's most recent transactions and basic profile attributes the company already holds. The essential step is confirming consent, privacy and regulatory constraints before using any signal, especially in banking, insurance and healthcare, where the rules on customer data are strict.

How do you measure the return of AI personalization in a portal?

Compare a personalized group against a control group that sees the standard experience, and track business outcomes rather than clicks. Useful indicators include conversion on the targeted module, self-service completion, contacts avoided, renewals and revenue per active user. A pilot that cannot show movement in one of these within a few months should be redesigned before it is scaled.

A portal that already knows who your customers are should never greet them as strangers. If you want your Liferay portal to speak to every user with real relevance, an ICX consultant can help you find the first moment worth personalizing and prove its value with your own numbers.