We spent Research Week in San Francisco listening, taking notes, and talking with the UX and market researchers, research managers, human factors engineers, research operations program managers, and product designers who gathered coast to coast. Sessions covered Advanced Market Research, Growth UXR, Great Research, AI and UXR, Up Up and Away!, and Moonshot Research.

‍

‍

‍

What are the key themes and insights that emerged from Research Week?

‍

‍

• Product defines, marketing communicates,  business captures the value, and researchers are translators and connectors.
• Build a culture of influence because influence isn't a moment, it's a structure.
• The insights alone are not enough; it's how you deliver and socialize them that makes them stick.
• Nothing hits like good UX research: compelling clips, stats, and verbatims.
• AI is transforming workflows, enabling large-scale data processing so researchers can focus on high-value work.

‍

‍

Practical takeaways to implement next week

‍

‍

Know who you're translating for and what matters to them

‍

‍

The gap in research most teams face right now is translation. Every function hears research differently. Each has a completely different "so what," as Apurva Luty puts it, so you need to respond and hear different ways, hearing data science in questions, design in critique, marketing in frameworks, engineering in constraints, and leadership in confidence. No one wants to hear: "we need to do more research”, but when you feel rushed for insights, you can always ask upfront if it’s a quick directional and recognize when you need more time for a comprehensive answer. 

‍

Building that bridge doesn’t mean you’re a gatekeeper. Rachel Ousley has seen democratizing access to data play out through more conversations and approaches with better questions.  Map your stakeholders' "so what" and foster an open line of communication because you are working towards the same goal in the end. 

‍

‍

Design for a culture of influence

‍

‍

Influence isn't a moment, it's a structure. Jess Holbrook breaks down direct influence versus indirect influence where direct influence is the central mechanism you present to senior leadership, and indirect influence is about setting the stage. Get ahead of it: know what your organization needs to understand in three months, six months, nine months, and how do we set ourselves up now to do that? 

‍

Build a culture of influence by giving credit loudly and often, saying people's names in the room by sharing wins and shoutouts. At Optimal, we do this through a celebrations Slack channel and quarterly value awards with open nominations through the company. Bring stakeholders, even ones you might not see eye-to-eye with upfront, into conversations where their perspective is genuinely valued, and anticipate what your team needs.

‍

‍

Continue running thoughtful studies with your users

‍

‍

The value of research is clear: to build better experiences, you must listen to the people who are using the product, service, the thing you’re making, and are affected by it. It is in discovery where you, well, discover what users actually want. As Andrew Chamberlain says, those hack projects and rapid prototypes can scale and become new products, and beyond that, research is how you elevate your brand and get invited into new spaces. 

‍

In discovery, know when to screen for behavior and when to screen for demographics. Maybe you’re looking at how people us mobile devices in homes, where one phone does not necessarily mean one owner or one user and in this case, your questions need to be framed openly and intuitively to get insight into your users’ mental models and actions, often different from your own, with people assigning different meanings to the same words. Nicole Naurath uses the example of asking “Do you share a device?” instead of, “How does someone else access this device?” to capture richer, more accurate insights into actual behavior.

‍

‍

Treat delivery like it's part of the research

‍

‍

Research reporting is socialization. Your decks don't have to change, but the artifacts around it do. A compelling clip, a sharp stat, a well-chosen verbatim – nothing hits like it. Nicole Zeng explains UX research as the thing that silences rooms, changes minds, and redirects roadmaps.

‍

Format your findings and discussion for the spaces people already work. Lauren Lin describes sharing insights as stackable and shareable clips on Slack as well as data cards that are downloadable as Figma components. 

‍

‍

Use AI to buy back your time for the work that matters

‍

‍

AI is enabling large-scale data processing that used to take months, which means you can spend less time in the weeds and more time on the work that moves the needle – the  judgment, translation, organizational, goal setting, and influence-building work. AI can handle the volume and scale of your data. However, everyone has a different comfort level with new tools. Nicole Zeng uses the analogy of a lake: maybe you’re diving in headfirst, maybe you’re watching from the shore, or maybe you’re paddling through the waves. 

‍

Break your workflow and explore novel ways of leveraging AI in UX research, then share out your findings and flows, because that's how we make progress as teams, get deeper customer insights, and ultimately make better decisions. It's why we're constantly evolving Optimal, and Optimal 3.0 is built for exactly this: helping product teams discover, validate, and continuously optimize user experiences that drive real business results.

‍

We're in an exciting time and it's moments like this when our industry comes together that we never forget. Stay connected with us on LinkedIn to get the latest updates on our upcoming events!

Artificial intelligence offers remarkable capabilities for UX research. It can process massive datasets, identify patterns humans might miss, and accelerate insights that traditionally took weeks to uncover. But as the adage goes: with great power comes great responsibility.

‍

As research teams increasingly adopt AI-powered tools, we're facing critical questions about data privacy, algorithmic bias, and ethical use of user information. These aren't just philosophical concerns, they're practical challenges that every research team needs to address.

‍

‍

‍

More data, more risk

‍

‍

AI thrives on data. The more information it can access, the better its pattern recognition and predictive capabilities become. For researchers, this creates a fundamental tension. To gain meaningful insights, you need comprehensive user data, but collecting and processing this data creates privacy risks that traditional research methods didn't face at the same scale.

‍

Think about a typical AI-powered analysis:

‍

• User session recordings processed to identify usability issues
• Behavioral data analyzed to understand user journeys
• Interview transcripts processed for sentiment analysis and theme identification

‍

Each of these activities involves handling sensitive user information. Each creates potential exposure points where data could be misused, breached, or processed in ways users didn't anticipate. The question isn't whether you should use AI but rather how to use it responsibly.

‍

‍

‍

Building privacy into your AI research practice

‍

‍

Privacy can't be an afterthought. It needs to be foundational to how you approach AI-powered research. Collect only the data you actually need. This seems obvious, but AI's hunger for information can encourage overcollection. Before implementing any AI tool, ask: What's the minimum data required to achieve our research goals? Just because you can collect comprehensive behavioral data doesn't mean you should. Be intentional about what you gather and why.

‍

Data security basics also become even more critical when AI is involved. Encryption, secure storage, access controls, these aren't optional. But security goes beyond technology. It includes policies around who can access data, how long it's retained, and what happens when a project concludes. AI systems often retain data to improve their algorithms. Make sure you understand your tools' data retention policies and ensure they align with your privacy commitments. A good example of this is how some tools, like Optimal, offer PII redaction on user interviews to ensure data security and privacy. 

‍

‍

‍

Be transparent with users

‍

‍

Users deserve to know how their data is being used. This goes beyond the standard privacy policy checkbox. When conducting research with AI-powered tools, you need to clearly communicate:

‍

• What data you're collecting
• How AI will process that data
• What insights you're hoping to gain
• How long you'll retain the information
• Who else might have access to it

‍

Give users meaningful control. If they're uncomfortable with AI analysis, offer alternatives. If they want their data deleted, make that process straightforward. Transparency builds trust. And trust is the foundation of good research.

‍

‍

‍

The bias problem

‍

‍

Something that all teams who incorporate AI into their research practices need to be aware of is that AI systems can perpetuate and amplify bias. Machine learning algorithms learn from training data. If that data contains biased patterns, and most data does, the AI will replicate those biases in its analysis. This can lead to research insights that systematically overlook certain user groups or misinterpret their needs. For researchers, this creates a serious challenge. You're using AI to understand users, but the AI itself might have blind spots that skew your understanding. Eliminating bias entirely is probably impossible. But you can take concrete steps to minimize its impact.

‍

1. Diversify your training data. If you're building custom AI models, ensure your training data represents the full diversity of your user base. This includes obvious factors like demographics, but also less visible ones like technical proficiency, language preferences, and usage contexts.
2. Use multiple analytical approaches. Don't rely solely on AI-generated insights. Combine algorithmic analysis with traditional qualitative methods. When AI flags a pattern, validate it through direct user research. When you see a trend in the data, talk to actual users to understand the context.
3. Interrogate unexpected findings. When AI produces surprising insights, don't accept them at face value. This skepticism isn't about distrusting AI. It's about using it thoughtfully.
4. Ensure diverse perspectives on your research team. Bias is easier to spot when you have people from different backgrounds reviewing the work. Build research teams that bring varied perspectives and life experiences. They'll be more likely to notice when AI-generated insights don't ring true for certain user segments.

‍

‍

‍

Navigating third-party AI tools

‍

‍

Most research teams don't build their own AI systems. They use third-party tools that come with built-in AI capabilities. This creates an additional layer of privacy and ethical considerations. Before adopting any AI-powered research tool you need to understand the vendor's data practices. Not all vendors handle data the same way. Choose partners who take privacy seriously.

‍

‍

‍

Stay current with regulations

‍

‍

Data privacy regulations are evolving rapidly. GDPR, CCPA, and emerging laws around AI governance create complex compliance requirements.nEnsure your AI-powered research practices align with relevant regulations in the jurisdictions where you operate. This isn't just about legal compliance, it's about respecting user rights.

‍

‍

‍

The most Important Ethical AI Component: Human judgment 

‍

‍

Here's what ties all of these considerations together: Human judgment must remain central to AI-powered research. AI can process data faster than any human, but it can't recognize when an algorithm is producing biased results or understand the ethical implications of a particular insight. These responsibilities fall to human researchers. And they can't be automated.

‍

At Optimal, we believe AI should enhance research capabilities while respecting user privacy and maintaining ethical standards. That's why we're committed to transparent data practices, secure infrastructure, and tools that put researchers in control. Because the goal isn't just better insights. It's better insights achieved responsibly.

The traditional product development process is a linear one. Research discovers insights, passes the baton to design, who creates solutions and hands off to product management, who delivers requirements to engineering. Clean. Orderly. Completely unrealistic in today’s modern product development lifecycle.

‍

Beyond the Linear Workflow

‍

The old workflow assumed each team had distinct phases that happened in sequence. Research happens first (discover users problems), then design (create the solutions), then product (define the specifications), then engineering (build it). Unfortunately this linear approach added weeks to timelines and created information loss at every handoff.

Smart product teams are starting to approach this differently, collapsing these phases into integrated workflows:

‍

• Collaborative Discovery. Instead of researchers conducting studies alone, the product trio (PM, designer, researcher) participates together. When engineers join user interviews, they understand context that no requirement document could capture.

• Live Design Validation. Rather than waiting for research reports, designers test concepts weekly. Quick iterations based on immediate feedback replace month-long design cycles.

• Integrated Tooling. Teams use platforms where research data and insights across the product development lifecycle, from ideation to optimization, all live in the same place, eliminating information silos and making sure information is shared across teams.

‍

What Collaborative Workflows Look Like in Practice 

‍

• Discovery Happens Weekly. Instead of quarterly research projects, teams run continuous user conversations where the whole team participates.

• Design Evolves Daily. There are no waterfall designs handed off to developers, but iterative prototypes tested immediately with users.

• Products Ship Incrementally. Instead of big-bang releases after months of development, product releases small iterations validated every sprint.

• Insights Flow Constantly. Teams don’t wait for learnings at the end of projects, but access real-time feedback loops that give insights immediately.

‍

In leading organizations, these collaborative workflows are already the norm and we’re seeing this more and more across our customer base. The teams managing it the best, are focusing on make these changes intentional, rather than letting them happen chaotically.

As product development accelerates, the teams winning aren't those with the best researchers, designers, or product managers in isolation. They're organizations where these teams work together, where expertise is shared, and where the entire team owns the user experience.

‍

Product, design and research teams today are drowning in user data while starving for user understanding. Never before have teams had such access to user information, analytics dashboards, heatmaps, session recordings, survey responses, social media sentiment, support tickets, and endless behavioral data points. Yet despite this volume of data, teams consistently build features users don't want and miss needs hiding in plain sight.

‍

It’s a true paradox for product, design and research teams: more information has made genuine understanding more elusive. 

‍

Because with  all this data, teams feel informed. They can say with confidence: "Users spend 3.2 minutes on this page," "42% abandon at this step," "Power users click here." But what this data doesn't tell you is Why. 

‍

The Difference between Data and Insight

‍

Data tells you what happened. Understanding tells you why it matters.

‍

Here’s a good example of this: Your analytics show that 60% of users abandon a new feature after first use. You know they're leaving. You can see where they click before they go. You have their demographic data and behavioral patterns.

‍

But you don't know:

‍

• Were they confused or simply uninterested?
• Did it solve their problem too slowly or not at all?
• Would they return if one thing changed, or is the entire approach wrong?
• Are they your target users or the wrong segment entirely?

‍

One team sees "60% abandonment" and adds onboarding tooltips. Another talks to users and discovers the feature solves the wrong problem entirely. Same data, completely different understanding.

‍

Modern tools make it dangerously easy to mistake observation for comprehension, but some aspects of user experience exist beyond measurement:

‍

• Emotional context, like the frustration of trying to complete a task while handling a crying baby, or the anxiety of making a financial decision without confidence.
• The unspoken needs of users which can only be demonstrated through real interactions. Users develop workarounds without reporting bugs. They live with friction because they don't know better solutions exist.
• Cultural nuances that numbers don't capture, like how language choice resonates differently across cultures, or how trust signals vary by context.
• Data shows what users do within your current product. It doesn't reveal what they'd do if you solved their problems differently to help you identify new opportunities. 

‍

Why Human Empathy is More Important than Ever 

‍

The teams building truly user-centered products haven't abandoned data but they've learned to combine quantitative and qualitative insights. 

‍

• Combine analytics (what happens), user interviews (why it happens), and observation (context in which it happens).
• Understanding builds over time. A single study provides a snapshot; continuous engagement reveals the movie.
• Use data to form theories, research to validate them, and real-world live testing to confirm understanding.
• Different team members see different aspects. Engineers notice system issues, designers spot usability gaps, PMs identify market fit, researchers uncover needs.

‍

Adding AI into the mix also emphasizes the need for human validation. While AI can help significantly speed up workflows and can augment human expertise, it still requires oversight and review from real people. 

‍

AI can spot trends humans miss, processing millions of data points instantly but it can't understand human emotion, cultural context, or unspoken needs. It can summarize what users say but humans must interpret what they mean.

‍

Understanding users has never been easier from a data perspective. We have tools our predecessors could only dream of.  But understanding users has never been harder from an empathy perspective. The sheer volume of data available to us creates an illusion of knowledge that's more dangerous than ignorance.

‍

The teams succeeding aren't choosing between data and empathy, they're investing equally in both. They use analytics to spot patterns and conversations to understand meaning. They measure behavior and observe context. They quantify outcomes and qualify experiences.

‍

Because at the end of the day, you can track every click, measure every metric, and analyze every behavior, but until you understand why, you're just collecting data, not creating understanding.

‍

AI is transforming how businesses approach product development. From AI-powered chatbots and recommendation engines to predictive analytics and generative models, AI-first products are reshaping user interactions with technology, which in turn impacts every phase of the product development lifecycle.

‍

Whether you're skeptical about AI or enthusiastic about its potential, the fundamental truth about product development in an AI-driven future remains unchanged: a product is only as good as its user experience.

‍

No matter how powerful the underlying AI, if users don't trust it, can't understand it, or struggle to use it, the product will fail. Good UX isn't simply an add-on for AI-first products, it's a fundamental requirement.

‍

‍

Why UX Is More Critical Than Ever

‍

Unlike traditional software, where users typically follow structured, planned workflows, AI-first products introduce dynamic, unpredictable experiences. This creates several unique UX challenges:

‍

• Users struggle to understand AI's decisions – Why did the AI generate this particular response? Can they trust it?
  ‍
• AI doesn't always get it right – How does the product handle mistakes, errors, or bias?
  ‍
• Users expect AI to "just work" like magic – If interactions feel confusing, people will abandon the product.
  ‍

AI only succeeds when it's intuitive, accessible, and easy-to-use: the fundamental components of good user experience. That's why product teams need to embed strong UX research and design into AI development, right from the start.

‍

‍

Key UX Focus Areas for AI-First Products

‍

To Trust Your AI, Users Need to Understand It

‍

AI can feel like a black box, users often don't know how it works or why it's making certain decisions or recommendations. If people don't understand or trust your AI, they simply won't use it. The user experiences you need to build for an AI-first product must be grounded in transparency.

‍

What does a transparent experience look like?

‍

• Show users why AI makes certain decisions (e.g., "Recommended for you because…")
  ‍
• Allow users to adjust AI settings to customize their experience
  ‍
• Enable users to provide feedback when AI gets something wrong—and offer ways to correct it
  ‍

A strong example: Spotify's AI recommendations explain why a song was suggested, helping users understand the reasoning behind specific song recommendations.

‍

‍

AI Should Augment Human Expertise Not Replace It

‍

AI often goes hand-in-hand with automation, but this approach ignores one of AI's biggest limitations: incorporating human nuance and intuition into recommendations or answers. While AI products strive to feel seamless and automated, users need clarity on what's happening when AI makes mistakes.

‍

How can you address this? Design for AI-Human Collaboration:

‍

• Guide users on the best ways to interact with and extract value from your AI
  ‍
• Provide the ability to refine results so users feel in control of the end output
  ‍
• Offer a hybrid approach: allow users to combine AI-driven automation with manual/human inputs
  ‍

Consider Google's Gemini AI, which lets users edit generated responses rather than forcing them to accept AI's output as final, a thoughtful approach to human-AI collaboration.

‍

‍

Validate and Test AI UX Early and Often

‍

Because AI-first products offer dynamic experiences that can behave unpredictably, traditional usability testing isn't sufficient. Product teams need to test AI interactions across multiple real-world scenarios before launch to ensure their product functions properly.

‍

Run UX Research to Validate AI Models Throughout Development:

‍

• Implement First Click Testing to verify users understand where to interact with AI
  ‍
• Use Tree Testing to refine chatbot flows and decision trees
  ‍
• Conduct longitudinal studies to observe how users interact with AI over time
  ‍

One notable example: A leading tech company used Optimal to test their new AI product with 2,400 global participants, helping them refine navigation and conversion points, ultimately leading to improved engagement and retention.

‍

‍

The Future of AI Products Relies on UX

‍

The bottom line is that AI isn't replacing UX, it's making good UX even more essential. The more sophisticated the product, the more product teams need to invest in regular research, transparency, and usability testing to ensure they're building products people genuinely value and enjoy using.

‍

Want to improve your AI product's UX? Start testing with Optimal today.

‍

Be honest. Are you guilty of being a gatekeeper? 

‍

For years, UX teams have treated research as a specialized skill that requires extensive training, advanced degrees, and membership in the researcher club. We’re guilty of it too! We've insisted that only "real researchers" can talk to users, conduct studies, or generate insights.

‍

But the problem with this is, this gatekeeping is holding back product development, limiting insights, and ironically, making research less effective.  As a result,  product and design teams are starting to do their own research, bypassing UX because they want to just get things done. 

‍

This shift is happening, and while we could view this as the downfall of traditional UX, we see it more as an evolution. And when done right, with support from UX, this democratization actually leads to better products, more research-informed organizations, and yes, more valuable research roles.

‍

The Problem with Gatekeeping 

‍

Product teams need insights constantly, making decisions daily about features, designs, and priorities. Yet dedicated researchers are outnumbered, often supporting 15-20 product team members each. The math just doesn't work. No matter how talented or efficient researchers are, they can't be everywhere at once, answering every question in real-time. This mismatch between insight demand and research capacity forces teams into an impossible choice: wait for formal research and miss critical decision windows or move forward without insights and risk building the wrong thing.

‍

Since product teams often don’t have the time to wait, teams make decisions anyway, without research. A Forrester study found that 73% of product decisions happen without any user input, not because teams don't value research, but because they can't wait weeks for formal research cycles.

‍

In organizations where this is already happening (it’s most of them!) teams have two choices, accept that their research to insight to development workflow is broken, or accept that things need to change and embrace the new era of research democratization. 

‍

In Support of  Research Democratization

‍

The most research-informed organizations aren't those with the most researchers, they're those where research skills are distributed throughout the team. When Product Managers and Designers talk directly to users, with researchers providing frameworks and quality control they make more research-informed decisions which result in better product performance and lower business risk. 

‍

When PMs and designers conduct their own research, context doesn't get lost in translation. They hear the user's words, see their frustrations, and understand nuances that don't survive summarization. But there is a right way to democratize, which not all organizations are doing. 

‍

Democratization as a consequence instead of as an intentional strategy, is chaos. Without frameworks and support from experienced researchers, it just won’t work. The goal isn't to turn everyone into researchers, it's to empower more teams to do their own research, while maintaining quality and rigor. In this model, the researcher becomes an advisor instead of a gatekeeper and the researcher's role evolves from conducting all studies to enabling teams to conduct their own. 

‍

Not all questions need expert researchers. Intercom uses a three-tier model:

‍

• Tier 1 (70% of questions): Teams handle with proven templates
• Tier 2 (20% of questions): Researcher-supported team execution
• Tier 3 (10% of questions): Researcher-led complex studies

‍

This model increased research output by 300% while improving quality scores by 25%.

‍

In a model like this, the researcher becomes more important than ever because democratization needs quality assurance. 

‍

Elevating the Role of Researchers 

‍

Democratization requires researchers to shift from "protectors of methodology" to "enablers of insight." It means:

‍

• Not seeking perfection because an imperfect study done today beats a perfect study done never.
• Acknowledging that 80% confidence on 100% of decisions beats 100% confidence on 20% of decisions.
• Measuring success by the "number of research-informed decisions made” instea dof the "number of studies conducted" 
• Deciding that more research happening is good, even if researchers aren't doing it all.

‍

By enabling teams to handle routine research, professional researchers focus on:

‍

• Complex, strategic research that requires deep expertise
• Building research capabilities across the organization
• Ensuring research quality and methodology standards
• Connecting insights across teams and products
• Driving research-informed culture change

‍

In truly research-informed organizations, everyone has user conversations. PMs do quick validation calls. Designers run lightweight usability tests. Engineers observe user sessions. Customer success shares user feedback.

‍

And researchers? They design the systems, ensure quality, tackle complex questions, and turn this distributed insight into strategic direction.

‍

Research democratization isn't about devaluing research expertise, it's about scaling research impact. It's recognizing that in today's product development pace, the choice isn't between formal research and democratized research. It's between democratized research and no research at all.

‍

Done right, democratization isn't the end of UX research as a profession. It's the beginning of research as a competitive advantage.

‍