AI Transformation: A Guide to Product Strategy and User Experience

A business owner opens a finance app to understand next month’s expected collections. They switch between reports, adjust date ranges, export data, and build a spreadsheet. The information is available. Most of the work of making sense of it still falls to the user.

Now imagine a different experience. The product considers upcoming receivables alongside overdue payments and highlights what needs attention, with supporting records. The user can adjust assumptions, explore the details, and decide what to do next.

This is the shift that matters when we look at AI transformation through a product lens: how does the work required to reach an outcome change for the user?

At VOYA, this question guides how we evaluate AI’s value in digital products. Producing content faster can be useful. A more meaningful transformation happens when a product better understands its users, helps them make decisions, and reduces unnecessary effort.

What is AI transformation?

AI transformation is the process of using artificial intelligence to redesign the value a company’s products deliver, the way work gets done, and how decisions are made. At the product level, the aim is to help users meet their needs with less effort, appropriate control, and a measurable benefit.

The scope varies by business. In a SaaS product, it might begin with simplifying setup. In a financial product, it could turn fragmented transaction data into an understandable assessment. In an operations platform, it might help teams identify exceptions earlier.

McKinsey’s March 2025 research identifies workflow redesign as a prominent organizational practice associated with reported financial benefits from generative AI. The findings suggest that the value of AI tools should be considered alongside changes in how people work. This is an association, not a guarantee of the same outcome for every company. [1]

Our takeaway for product design is clear: the AI roadmap and the user journey need to be developed together.

How does AI change a product’s value proposition?

A product’s value proposition explains why someone chooses to use it. AI becomes strategically relevant when it strengthens that reason.

A reporting product, for example, might promise to bring all your data into one place. With AI, that promise could evolve into helping you identify the changes that deserve attention and investigate what is behind them. This shift affects the home screen, notifications, report structure, and the reasons users return.

Similarly, a campaign management product could learn a user’s goals and constraints, then prepare an editable campaign draft. The value depends as much on the user’s ability to assess that draft as on the system’s ability to generate it.

At this stage, we recommend that product teams answer three questions:

• What outcome do people use our product to achieve today?
• What interpretation, comparison, or preparation do they have to do themselves along the way?
• How would the product’s value improve if AI took on part of that work?

If these answers are unclear, the role of the proposed AI feature will probably be unclear too.

How do you choose the right AI use case?

A strong starting use case combines a recurring user need, accessible data, manageable consequences if something goes wrong, and a measurable outcome. The team should also be able to explain why AI would be more useful than a simpler solution.

Google’s People + AI Guidebook recommends looking for the intersection between user needs and AI’s strengths. It also points out that, in some situations, a rules-based solution can be easier to explain, build, and maintain. [2]

For example, if users do not understand the fees on a payment screen, the first improvement might be to display those fees more clearly. Asking users to consult a chat assistant could add an unnecessary step.

Extracting specific information from long documents in different formats, then presenting it for review, is a more plausible candidate for AI. Even here, missing information, incorrect matches, and human review need to be part of the design.

The following questions can help teams evaluate potential use cases:

We use this as a discussion framework rather than a fixed scoring formula. It helps teams assess different ideas through the same questions. An idea with substantial potential user impact may still be unsuitable for an initial pilot because of missing data or the consequences of errors.

What does AI transformation look like in fintech and SaaS?

The examples below are hypothetical product scenarios created to illustrate the approach. They are not VOYA client case studies or reported results.

Financial products: From transaction lists to insights users can verify

A business owner wants to identify customers with overdue payments. In the current experience, they compare transaction records with invoices and prepare a separate follow-up list.

An AI-enabled experience could bring together suggested matches and potential overdue payments for review. Each suggestion would include the relevant records, the date range used, and any missing data. The user could correct a match or reject a suggestion.

Preparation time would be only one measure of success. We would also track incorrect matches, time spent checking the results, and user corrections. A list that is faster to prepare but takes more effort to review may not deliver the expected benefit.

SaaS products: From an empty workspace to an editable starting point

A new user joins a project management product and is unsure how to set up their workspace. They browse templates, create fields, and define their team structure.

An AI-assisted setup could ask a few clear questions about their work and team, then prepare an initial configuration. The user would review the suggested tasks, roles, and stages before completing setup.

The outcome we want to measure is whether users complete their first real task, continue using the proposed structure, and need substantial changes after setup. The number of workspaces generated tells us much less about the value delivered.

Operations products: From reviewing reports to managing exceptions

A manager at a business with multiple locations reviews the same reports every morning. Their goal is to spot changes that need attention, which requires working through the numbers.

The product could surface unusual patterns for review, showing the comparison period, data coverage, and possible explanations. Detailed reports would remain available. Where the data is insufficient, the system would make that limitation clear.

Success would be assessed through the proportion of useful alerts, important events that were missed, and review time. A product that repeatedly raises false alarms may eventually find its notifications ignored.

How do you design trust into AI user experiences?

Trust in an AI experience is supported by helping users understand what the system is offering, where its limits are, and how they can intervene. Designing for trust therefore goes beyond adding an informational message to the screen.

Nielsen Norman Group’s research on enterprise AI explainability emphasizes that the people building, evaluating, and managing AI systems need different kinds of explanations. [3] In product design, explanations should likewise be considered in relation to a person’s task and decision context.

We translate this into four design decisions.

Make the type of output clear. Users should be able to distinguish a forecast, recommendation, draft, or verified record. Presenting predicted collections and received payments in the same visual language, for example, can create a false impression of certainty.

Make the supporting information available. Users should be able to inspect the records and time period behind a recommendation. Showing a source does not guarantee accuracy, but it gives people something they can check.

Make user control visible. Editing, rejecting, retrying, and switching to a manual process should be available within the task flow where appropriate. For actions with real consequences, displaying a recommendation and executing it should be designed as distinct states.

Design for uncertainty and failure. The product needs a defined response when data is missing, records conflict, or an answer cannot be generated. Users should understand what is missing and what they can do next.

These decisions also affect information architecture. States such as “Draft,” “Awaiting review,” and “Applied” need to be consistent across the product. A design system should define these behaviors alongside the visual appearance of AI components.

Does every AI experience need a chat interface?

No. The interaction model should follow the task the user needs to complete.

Chat can be useful when someone wants to describe a goal in their own words. Tables, forms, and previews may work better when they need to compare many records, edit multiple fields, or review the scope of an action.

In a campaign creation flow, for example, the user could begin by describing their objective. The product could then present suggested budgets, audiences, dates, and content in editable fields. A final preview would show the campaign that is about to go live.

Here, conversation helps with one part of the task. It does not require the user to complete every subsequent step through chat.

At VOYA, our priority in interaction design is choosing the format that helps users evaluate the result most comfortably and effectively.

Which metrics measure AI transformation success?

Knowing how often an AI feature is opened is a useful starting point. Understanding whether that usage creates value requires looking at the outcome of the task.

BCG’s 2025 research on AI value also shows that business outcomes from AI vary substantially between companies. [4] We therefore recommend defining outcome measures for each use case rather than treating tool usage as success in itself.

Establishing a baseline before the pilot makes later comparisons possible. Where feasible, compare similar tasks and user groups, and assess whether improvements in speed come at the expense of quality.

For example, if report preparation becomes faster while verification takes longer, reporting only the generation speed would be misleading. Time saved also does not automatically translate into cash savings. Teams need to understand how the capacity gained is used.

How does VOYA approach AI transformation?

At VOYA Digital, we see AI transformation as a shared challenge across product strategy, user research, and experience design. Our approach follows four stages: Understand, Define, Design, and Develop.

Understand. We investigate the tasks users struggle with, where the current product falls short, and the business objective. Interviews, behavioral data, and reviews of existing flows help make opportunities concrete. We also work with the technical team to assess data and integration constraints.

Define. We establish AI’s role in the selected use case, the decisions that remain with the user, and the measures of success. We clarify how the new experience supports the product’s value proposition and define responsibilities around generating recommendations and taking action.

Design. We consider flows, interfaces, explanations, and failure states together. In prototype testing, we assess both task completion and whether users interpret the output correctly. We also test whether they can spot and correct an inaccurate or incomplete suggestion.

Develop. We work with product and technology teams to bring the validated experience into implementation. We clarify the design system, interaction rules, and development handoff, then support improvements informed by live usage.

The first engagement does not have to become a project that changes the entire product. A clearly scoped user journey, a testable prototype, and a measurement plan with explicit criteria for moving forward can provide a strong foundation.

Frequently asked questions

Do you need a complete product redesign to start AI transformation?

No. You can begin with a clearly defined user journey in an existing product. The scope can expand as you assess its impact on other screens, data flows, and ways of working. A complete redesign should be justified by user and business needs.

What is the difference between AI product strategy and AI-assisted design?

AI-assisted design means using AI tools during research or design production. AI product strategy defines the value a product should deliver to users and the role AI will play in creating that value. The two can support each other, but they address different questions.

How should you choose your first AI transformation pilot?

Choose a recurring task with an outcome you can evaluate and manageable consequences if something goes wrong. Confirm data suitability, baseline performance, and the need for human review before starting. Define the criteria for continuing, changing, or stopping the initiative at the end of the pilot.

Why is UX research necessary for AI features?

A technically functioning system is not automatically understood or adopted by users. UX research reveals their needs, expectations, how they assess outputs, and where they want control. It allows teams to evaluate technical performance alongside practical value.

What determines the cost of AI transformation?

The use case, data preparation, integrations, model operating costs, testing scope, and ongoing monitoring all affect the total cost. Human review and correction effort should also be included. A reliable budget starts with a clearly defined user journey and deliverables.

Find out where AI can create value in your product

A useful starting point is to choose a task that demands significant effort from your users today. Then ask: how should our product behave to help them complete it more easily, with greater clarity and confidence?

VOYA Digital connects user needs with product strategy and UX/UI design for fintech and SaaS products. To explore AI transformation opportunities, prioritize use cases, and turn them into testable experiences, get in touch with VOYA.

References

1. McKinsey — The state of AI: How organizations are rewiring to capture value, March 12, 2025
2. Google PAIR — User Needs + Defining Success
3. Nielsen Norman Group — Crafting AI Explanations for Every Role in Your Enterprise, July 3, 2026
4. BCG — The Widening AI Value Gap, September 30, 2025