Is Identifying User Problems Easier Post-AI? What Works for Me as UX Researcher

Identifying user problems has always been challenging. You can run user research, analyze behavioral data, and still solve the wrong problem.

On one hand, AI tools (for transcription, data synthesis, synthetic research, etc.) are making the research process easier. On the other hand, no technology can solve the puzzle that matters most: why a person, with a job and pain points you tirelessly aim to solve, keeps running into the same wall in your product.

So what’s the best framework for finding the most pressing UX problems in your product, assessing their impact accurately, and implementing the right solutions?

For this article, I’ll walk through three topics:

1. How AI, synthetic users, and AI agents are affecting user problems in 2026.
2. The 3 key research methods I rely on for finding UX problems.
3. How I solve UX problems with in-app guidance without having to code.

How is identifying user problems changing in 2026?

A user problem is a challenge or an obstacle a user faces when trying to achieve a goal or accomplish a task. This definition hasn’t changed, but the landscape for identifying one has:

• AI agents are creating a new category of user problems: When AI systems interact with software through MCP (Model Context Protocol) and API calls rather than graphical interfaces, they generate a completely different behavioral signal than human users do. Yet, most teams are blind to them. Agents don’t click or scroll; they call functions, and an agent that calls the wrong API endpoint is a different category of problem entirely. This is why Agent Analytics is emerging as its own category.
• AI-assisted analysis has become standard practice in research teams: According to the Maze 2026 User Research Report, 69% of researchers now use AI in some part of their research process, rising 19 percentage points from the prior year. The MRII 2025 global AI report puts 62% of research teams using AI actively, up from 39% the year before. The dominant use cases are faster transcription, thematic coding, and synthesis across large interview sets.
• Synthetic users are being debated: The Lyssna 2025 UX Research Trends report found that 48% of researchers see synthetic users as a top trend for 2026, while 88% identify AI-assisted analysis as their primary AI use case. Synthetic users are AI-generated profiles that simulate user responses without involving real people. According to NN/G’s tests, synthetic users “feel like a flat approximation of the experiences of tens of thousands of people, because they are”. Even Hugo Alves, cofounder of Synthetic Users, says “you’re never gonna stop talking to real people and you shouldn’t.”

My concern about AI usage in research is that it might give teams the false illusion of knowing the truth about users’ problems. So I’ll go over what I think are the three most fundamental methods for finding and validating UX problems, regardless of AI.

Three methods to identify real user problems

No single research method gives you the full picture of a user problem. What users tell you in an interview is shaped by the recency bias. And the numerical behavioral data sometimes doesn’t give you the full view of the users’ actual issues.

The process of creating and confirming hypotheses must follow a thorough research cycle that will help you build data-based conclusions. The methods of running such a process include:

1. Customer research.
2. Sentiment analysis.
3. Behavioral analysis.

Direct customer research: What users say

Customer research covers the methods that put you in direct contact with real users. It’s best suited to finding problems that aren’t explicit: the workarounds they’ve normalized, the features they avoid for reasons they’ve never written down, and the decisions that look irrational in your data but make sense after talking with them.

The core methods in this category include:

• User interviews: One-on-one conversations focused on understanding how users think about a problem space or specific workflow. Best for finding individual insights that behavioral data can’t capture.
• Usability tests: Structured sessions where users attempt real tasks in your product while you observe where they get stuck. Best for finding friction in specific flows.
• Focus groups: Group discussions useful for understanding how users talk about a problem and what language they use. Best in early-stage research before you’ve defined the problem space.
• Contextual inquiry: Observing users in their actual work environment, which almost always surfaces constraints and workarounds that scheduled interviews miss. Best for understanding hidden problems that users usually don’t mention.
• Diary studies: Longitudinal research where users record their experiences over time. Best for problems that develop gradually rather than appearing on first contact with a feature.

For these, I see teams facing friction when recruiting users. My best recommendation is to recruit directly within the product via in-app surveys. With Userpilot, for instance, I can target invitations to specific user segments and easily recruit users who are already familiar with our product without resorting to third parties.

Sentiment analysis: What users feel

Sentiment analysis measures how users feel about their experience with your product. Since you can’t interview every user who churned, this analysis lets you at least measure the satisfaction score that preceded their cancellation.

Although this analysis can’t always explain the cause of a problem, it’s great for finding common problems and prioritize solutions.

The core instruments for this method include:

• NPS (Net Promoter Score): Measures overall loyalty and likelihood to recommend. It’s useful for spotting large-scale satisfaction shifts over time.
• CSAT (Customer Satisfaction Score): Measures satisfaction with a specific interaction or feature. It’s best used when triggered during specific points in a workflow (e.g., completing a task, setting up an account, finishing onboarding, etc.).
• CES (Customer Effort Score): Asks users how much effort they had to exert to complete a task. It’s particularly useful for identifying friction in adoption and onboarding flows.
• PMF (Product-Market Fit) surveys: Ask users how they would feel if they could no longer use the product. It’s a standard survey to assess whether you’ve reached product-market fit or not.
• Micro-surveys and exit surveys: Custom surveys triggered at key moments in the product (e.g., after activation, before cancellation, after dropping off, etc). It’s best for capturing more context that a standard survey would miss.
• Feedback widgets: Feedback boxes that let users report problems as they encounter them, without waiting to be prompted. Best for collecting passive feedback and receiving bug reports.
• Social listening: Monitoring forums, communities, and review platforms for unsolicited feedback. It doesn’t involve surveys, but it’s great for capturing problems users describe to each other without the context of a company asking them.

For many companies, sending out surveys required coordination across three or four different tools, which made the whole process slow. For us, Userpilot’s user feedback tools handle collection and analysis in one place, which speeds that up considerably.

Behavioral analysis: What users do

Behavioral analysis captures what users actually do in your product, tracking quantitative data like DAUs, session durations, in-app events, etc. It’s best suited to finding problems that prevent users from achieving explicit tasks (i.e., hitting activation, adopting core features, setting up accounts, etc) or identifying in-app habits that are less effective than power users.

For me, the best methods for behavioral analysis include:

• Funnel analysis: Shows where users stop in a multi-step flow, telling you where drop-offs happen.
• Path analysis: Shows the routes users actually take through the product, including unexpected detours or happy paths.
• Retention cohort analysis: Compares the retention rates of users in different groups. It’s best for revealing the specific actions that predict long-term retention.
• Trend analysis: Tracks activity trends over time to identify problems that develop gradually or surface after a release.
• Session replay: Collects recordings of individual user sessions that show exactly how a drop-off or error state occurred.
• Heatmaps: Shows where users click, scroll, and focus attention across a page. Best for spotting dead clicks or key parts of a page that don’t get enough attention.
• Agent Analytics: Tracks the function calls, API requests, and error patterns generated by AI agents interacting with your product, which requires separate instrumentation from human UX analytics.

An example from our experience: there’s a graph toggle feature in Userpilot’s dashboard that lets users switch between chart views, but users weren’t using it often enough. The session replays showed that users were clicking the toggle repeatedly with no visible response. The button was broken, which led us to perform a quick fix before more users faced that issue.

How to address user problems once you’ve found them

Once research spots real user problems, you might brainstorm and prioritize solutions using an impact-effort matrix.

But usually, we end up piling up too many requests to the engineering team, which means those UI fixes might not ship until the next update.

As an alternative, I’ve found that implementing in-app guidance with a no-code tool serves as a low-effort solution. It helps users navigate your product with minimal obstacles, and you can deploy it in a day while devs keep working on high-impact fixes.

Here are my favorite methods to implement in-app guidance:

Personalized interactive walkthroughs

An interactive walkthrough is a step-by-step flow that guides users through a specific action inside the product, triggered at the moment they need it. Unlike a static product tour, a walkthrough responds to where the user is and what they’ve already done, which means it shows up only when it’s relevant and disappears once the user has completed the action. This makes walkthroughs particularly effective for onboarding flows and feature adoption, where the main problem is that the user doesn’t know how to start.

For example, when a user lands on a feature page they haven’t activated in Userpilot, it triggers a contextual walkthrough that takes them through each step inline, without requiring them to navigate to a help doc or open a support ticket. The walkthrough closes when the user completes the action and won’t reappear for that user on subsequent visits.

My recommendation is to personalize the experience, that is, to target different interactive walkthroughs for each user persona. For this, segment new users based on their role with a welcome survey, and then trigger a personal walkthrough based on their response. This will increase the chance for the user to experience the AHA! moment and achieve activation.

Onboarding checklists

An onboarding checklist is a structured list of key actions for new users, and it’s tied to the behaviors that predict early retention. Unlike a welcome email sequence, a checklist shows up inside the product and tracks completion in real time, making the path to first value visible and giving users a sense of progress as they work through it.

This works especially well when paired with retention cohort data: if you know which actions distinguish users who stayed from those who churned, you can add those actions to the checklist. Also, you can track completion rate by step to see which task of the checklist is creating friction, which feeds back into your next round of research.

We have Userpilot to implement onboarding checklist that guides new users. For instance, we have a checklist for core setup steps: installing the code snippet, creating their first flow, and publishing it to a user segment. This has allowed me to spot steps where users stall and add more guidance to prevent drop-offs.

In-app resource center

An in-app resource center is a self-serve help hub accessible directly within the product, without requiring users to navigate to a separate documentation site. It can contain articles, video walkthroughs, FAQs, and links to live support, organized by topic or user stage. The goal is to make help available at the moment of confusion, not after the user has already given up and closed the tab.

The most effective resource centers are organized by what users are trying to do, not by product features. A new user searching for help on an onboarding step needs different content than an experienced user troubleshooting an integration, and mixing those two audiences in the same navigation makes both experiences worse. With Userpilot’s resource center, for instance, we can show different content to different user segments, which means the most common questions from product managers will be easier to find for them.

Contextual guidance

Contextual guidance covers the smaller-scale, targeted interventions that trigger at specific moments of friction. Think of tooltips that explain a specific UI element when a user pauses on it, hotspots that draw attention to features users haven’t discovered, and modals that educate users about how to use a tool they’re revisiting.

The key to making contextual guidance work is triggering on behavioral signals rather than time or session count. A tooltip targeting a user who has visited a settings page three times without making a change is far more effective than a tooltip shown to everyone who opens that page once. We use Userpilot’s targeting to set triggers based on user properties, event history, and session behavior, which makes the guidance precise rather than easy to dismiss.

For example, Abrar Abutouq, Product Manager at Userpilot, once found that users were completing the wrong steps in a workflow because one action in the sequence wasn’t clearly labeled. Rather than waiting for a development cycle to redesign the UI, she built a targeting tooltip in a few hours that appeared for users who reached that step without completing the action correctly. In her words:

“Within a few hours, I just created a targeting tooltip and showed it to users and highlighted the correct steps for them to make it clear what to do next. That helped a lot on reducing friction and supporting users in real time without involving our dev team.”

Human judgment is still the differentiator in UX research

The Maze 2026 report found that researchers still rely on human judgment for understanding emotional nuance, making ethical decisions, and framing research questions.

Today, AI is good at handling the operational parts that used to take more time. That is, getting participants recruited through in-product channels, transcribing sessions, coding themes, and distributing surveys. But none of this will matter without the expertise of a proper researcher who knows what user problems look like.

So if you want to apply the methods I explained and solve in-app friction quickly, I highly recommend booking a Userpilot demo. It’s free, and it makes many parts of the research process (like recruiting and sending surveys) easier without coding.