Why Fixing UX First and Then Strategic AI Rollout Drive Real Results
By Nisal Tharanga
The conversation in every leadership meeting today seems to circle back to the same question: how do we bring AI into our product? And for good reason. Artificial intelligence is no longer a future consideration. It is a present reality reshaping how digital products serve customers, how businesses operate, and how organisations compete. The pressure to act is real, and I understand it. Competitors are making moves, customers are expecting smarter experiences, and the promise of efficiency, personalisation, and scale is difficult to ignore.
But here is what I have observed across multiple digital product implementations, particularly in financial services and banking: the organisations that rush to integrate AI into products with unresolved user experience issues rarely see the returns they expect. Instead of unlocking value, they compound existing friction. Users who already struggle with basic tasks are now presented with intelligent features they neither asked for nor understand. The result is not innovation. It is confusion layered on top of frustration.
This is not an argument against AI. Far from it. AI will be essential to the next generation of digital products, and organisations that delay too long risk falling behind. But the question is not simply whether to invest in AI. It is whether your product has earned enough user trust to carry that investment forward. In my experience, sequencing matters more than speed. When the foundation of user experience is solid, AI amplifies value. When it is not, AI amplifies problems.
In this article, I want to share a practical perspective on why getting your UX fundamentals right is a prerequisite for meaningful AI integration, what the most common experience gaps are that I see being overlooked, and how organisations can introduce AI in a phased, measurable way that builds user trust rather than testing it. This is relevant for any industry building consumer facing digital products, and especially for banking and financial services, where user confidence is not just a design metric. It is directly tied to adoption, retention, and revenue.
The Business Case for Fixing UX First
Before discussing AI readiness, it is worth understanding what poor user experience is already costing your organisation. The numbers are difficult to ignore. Research by Forrester shows that every dollar invested in UX can return up to a hundred dollars. On the other side, a study published by Amazon Web Services (AWS) estimates that poor user experience costs businesses approximately 1.4 trillion dollars annually in lost revenue. These are not design
statistics. They are business performance indicators that directly affect customer acquisition cost, retention, and lifetime value.
A study by McKinsey confirms that organisations improving their customer journey satisfaction can increase cross sell rates by 15% – 25% and boost share of wallet by 5% – 10%. For a financial institution or any consumer facing platform, that is not a marginal gain. It is a strategic revenue opportunity sitting inside the experience your users already have. And yet, many organisations continue to underinvest in experience improvements while allocating significant budgets toward new technology layers.
What concerns me further is a pattern I have observed across the industry: organisations launching multiple apps or platforms under a single brand to serve different functions. In banking, for example, a customer might be expected to use one app for everyday transactions, another for investments, and yet another for a specific service such as insurance or digital wallets. From a business perspective, the logic may seem sound. From a user perspective, it creates confusion and unnecessary complexity. People recognise and trust their bank by name. When they are asked to navigate between products with unfamiliar names and different interfaces, it fragments their experience rather than enhancing it. Based on multiple user researches I have conducted for many products, users consistently express discomfort when they encounter product names that differ from the brand they originally trusted. That disconnect, even if subtle, introduces hesitation and friction at the very moments where confidence matters most.
Research by PwC reinforces this reality: 32% of customers will leave a brand they once valued after just one poor experience. When you multiply that across fragmented product touchpoints and inconsistent journeys, the risk compounds quickly. The question leaders must ask is not how much AI will cost to implement, but how much the current experience is already costing in lost engagement and missed opportunity.
Why AI on Top of Poor UX Makes Things Worse
When a user interacts with a digital product, they are extending trust at a fundamental level. Can I navigate this? Will it do what I expect? Is this reliable? These are the baseline questions every user silently asks before they commit to using a product regularly. AI introduces a second, distinct layer of trust on top of this. Is this recommendation accurate? Is the system making decisions on my behalf? What happens when something goes wrong? For AI to deliver value, users must first be comfortable with the product itself. Without that foundation, adding intelligence does not elevate the experience. It complicates it.
The global data supports this concern. A comprehensive 2025 study by KPMG and the University of Melbourne, surveying over 48,000 people across 47 countries, found that only 46% of people globally are willing to trust AI systems. Nearly half of respondents said they do not feel they understand AI or when it is being used. This is not a technology gap. It is a trust and literacy gap. Through the user studies I have carried out as part of engagements I have had with
AI powered product creation, people have shared mixed feelings about using AI driven features. Some see the potential but hesitate because they are unsure how the system arrives at its suggestions. Others are open to trying but quickly disengage when the experience feels unpredictable or difficult to control. These observations from real users confirm what the KPMG and University of Melbourne study highlights at a global scale: willingness to use AI does not automatically mean willingness to trust it.
In a market like Sri Lanka, where a significant portion of the population is still building confidence with digital products and where many users have recently transitioned from branch based interactions to mobile platforms, this gap is even more pronounced. AI literacy is limited, and users are naturally cautious about features they do not fully understand. Consider a user who already struggles to locate a specific service within their banking app. Now introduce an AI assistant that proactively suggests financial products or automates the categorisation of transactions. For a user who has not yet built confidence in the basics, this does not feel like innovation. It feels overwhelming.
The leadership takeaway here is clear. AI investment without user experience readiness is not a growth strategy. It increases your customer acquisition cost and accelerates churn among the very users you are investing to retain. The smarter path is to treat UX readiness as a prerequisite, not an afterthought.
Five UX Foundations to Get Right Before Introducing AI
If UX readiness is the prerequisite, the next question is where to start. Based on patterns I have observed across successful digital product implementations, and validated through user research across multiple industries, there are five foundational experience areas that organisations must address before introducing AI. These are not aspirational design goals. They are practical, measurable issues that directly impact whether users trust and engage with your product.
- Performance and interface optimisation. One of the most fundamental issues I observe across consumer facing platforms is poor performance. Slow loading screens, unresponsive interactions, and lag during critical tasks such as payments or form submissions weaken user patience and trust before any feature gets a chance to prove its value. The root causes are often a combination of technical debt and poor interface optimisation. Research shows that 88% of users will not return to a product after a poor experience, and load time is frequently the first point of failure. Before any AI feature is considered, the platform must perform reliably and responsively.
- Information architecture and navigation. This is the core of the experience. How services, features, and content are structured must be defined based on how users think and what they are looking for, not how the organisation is structured internally. In my experience, many digital products mirror their company’s departmental structure rather than the user’s mental model. The result is that users cannot easily find the service they
need within two or three steps. Industry research suggests that only 20% to 25% of mobile app users engage with features beyond the basics, often because those features are buried under vague labels or hidden in secondary menus. No amount of intelligent recommendation will compensate for an architecture that users cannot navigate.
- Clear, human language communication. Technical jargon remains one of the most overlooked barriers in digital products, especially in financial services. Error messages, notifications, instructions, and confirmations should be written in plain language that users immediately If a payment fails, the user needs to know why and what to do next, not read a message that says “transaction could not be processed due to a system error.” In the Sri Lankan context, this also means ensuring that communication works effectively in Sinhala, Tamil, and English, meeting users in the language they are most comfortable with.
- Consistent cross channel experience. Users expect the same product visibility and functionality regardless of whether they access a service through mobile or web. In practice, I have seen platforms where certain accounts or services are only accessible through internet banking and not the mobile app. In one instance from my own experience, I needed to delete saved transfer details and discovered there was no way to do it on mobile. Customer service directed me to use internet banking instead. These inconsistencies quietly damage trust and create unnecessary effort for users at moments when the experience should feel seamless.
- Onboarding and first time experience. The first interaction a user has with your product sets the tone for everything that follows. Complex registration flows, excessive information requests upfront, and unclear next steps cause users to abandon before they ever become active customers. The most effective approach is progressive onboarding: ask only what is necessary at each stage, with deeper verification reserved for advanced features or higher value If users cannot get through the front door comfortably, nothing else you build, including AI, will matter.
I recommend that organisations treat these five areas as a UX trust audit. Assess them honestly before approving any AI investment. In most cases, addressing these fundamentals delivers measurable uplift in retention, task completion, and conversion well before any AI feature is introduced.
Introducing AI the Right Way: A Phased Approach
Once these UX foundations are in place, the conversation shifts from whether to introduce AI to how to introduce it responsibly. The mistake I see many organisations make at this stage is treating AI as a single, large scale launch. They invest heavily, build an ambitious feature set, and release it all at once, expecting users to adapt. Research consistently shows this approach underperforms. Organisations that adopt phased rollouts report significantly higher user
satisfaction and lower implementation failure rates compared to those attempting full deployments.
Based on what I have seen work well in practice, I recommend a four phase model that gradually increases AI visibility while giving your team the data it needs to make informed decisions at each stage.
In the first phase, AI should work passively in the background. Use it for analytics, fraud detection as a precautionary measure, and personalisation engines that improve the product without users needing to interact with AI directly. They simply benefit from smarter defaults and more relevant content. During this phase, monitor adoption metrics of existing features to establish a reliable baseline before introducing anything visible.
The second phase introduces AI as an assistive, optional layer. This could include smart search, auto categorisation of expenses, or gentle nudges such as reminders about upcoming bill payments. The key here is choice. Users should always be able to ignore or dismiss AI suggestions without any penalty to their experience. Track engagement rates, dismissal patterns, and whether support queries increase or decrease.
In the third phase, AI becomes interactive. This is where chatbots, voice assistants, or AI guided workflows enter the picture. However, human support must always remain accessible, and AI interactions should be transparently labelled so users know when they are engaging with an automated system. Measure whether users complete AI suggested actions or revert to manual flows. That behaviour tells you more about trust than any survey.
The fourth phase is where AI becomes fully embedded and feels like a natural part of the product. Users no longer think of it as a separate feature. Global examples such as DBS Bank and Bank of America with its AI assistant Erica have reached this maturity, but it took years of iteration, learning, and consistent user feedback to get there.
Across all four phases, one element is non-negotiable: user education. Especially in our country, where AI literacy remains limited, each phase must include clear, simple explanations of what AI is doing and why. Tooltips, short walkthroughs, and contextual guidance are not extras. They are the bridge between a feature being available and a feature being trusted.
The Path Forward Starts With What You Already Have
The most strategic AI investment an organisation can make right now may not be in AI at all. It may be in strengthening the experience users already have. Trust is cumulative. Every UX improvement builds toward a foundation that makes AI adoption possible and sustainable. Sri Lanka’s digital economy is at a meaningful inflection point. The organisations that invest in user trust first and introduce AI with intention and patience will be the ones that lead, not just in technology adoption, but in genuine customer loyalty and long term growth.
(Nisal Tharanga is a Senior Product Designer specializing in AI-driven design with over a decade in user-centered, inclusive experiences and 20+ years in software development)