The AI question researchers should be asking

The speed trap

For decades, researchers have been rewarded for rigor. Now they're increasingly being asked for speed.

Harsha Vemulapalli, Director of Experience Design and Research at Siemens, described the pressure succinctly. The expectation from stakeholders, he explained, is that AI should make everything faster.

"You launched the survey yesterday. Why can't you just throw it all into Copilot and have it summarize and give me the answers?"

It's a fair question. AI can summarize interviews, synthesize survey responses, generate discussion guides, and surface themes in minutes. What once took days can now take hours. What once took hours can sometimes take seconds.

But speed creates its own illusion. Faster access to information can make organizations believe they're getting insight when they're really getting something else.

Harsha drew a distinction that may become one of the defining frameworks for the AI era: signal versus insight.

"If it's signal they're looking for, that's an amazing opportunity," he said. "If it's intent that they're looking for, wow, that's a lot harder."

A signal tells you what happened. Insight helps explain why.

AI excels at the first. Humans are still better at the second.

The democratization dilemma

The other major shift isn't technological. It's organizational.

Research is no longer confined to researchers.

At Salesforce, David Gardner sees a future where product managers, designers, and engineers increasingly conduct their own research, supported by AI-powered tools and workflows. Rather than resisting that reality, he argued that research leaders should embrace it.

"Research is being democratized in many ways," David said.

The instinctive response might be defensive. If everyone can conduct research, what happens to researchers?

But David framed the opportunity differently. Researchers should become the architects of quality rather than the sole producers of research. Their role is less about owning every study and more about establishing standards, creating systems, and ensuring that evidence remains trustworthy.

The analogy is less newsroom editor and more air traffic controller. Researchers may not be flying every plane, but they're responsible for making sure they don't collide.

That responsibility matters because AI has lowered the barriers to gathering feedback while doing little to lower the risks of misinterpreting it.

The danger of false confidence

Lucas Puente, Vice President of Research at Slack, highlighted a growing challenge: AI can make weak evidence feel persuasive.

Today, anyone can ask a chatbot what customers think about a product and receive an answer that sounds authoritative. The model might pull comments from Reddit, social media, or online forums and present them as a coherent narrative.

The problem is that coherence is not the same thing as truth.

"It's really easy for PMs or non-researchers to get false confidence," Lucas warned.

Researchers have always worried about sample bias, representativeness, and methodological rigor. Those concerns haven't disappeared. If anything, they've become more important because AI can package incomplete information so convincingly.

As Lucas put it, "The only thing worse than no data is bad data."

That's not an argument against AI. It's an argument for maintaining healthy skepticism toward its outputs.

Leadership's new responsibility

If there was one point of consensus among the panelists, it was that leaders can no longer outsource AI literacy.

David offered perhaps the clearest advice of the session.

"You cannot just delegate AI to your team and hope for the best."

That sentiment was echoed by Lucas, who shared a favorite metaphor. Leaders, he said, shouldn't be "watching the dance from the balcony." They need to be on the dance floor.

The implication is straightforward. AI adoption isn't a project to assign. It's a capability leaders must develop themselves.

That doesn't mean becoming a prompt engineer or software developer. It means understanding enough to recognize possibilities, limitations, and risks.

Organizations often underestimate how much cultural change accompanies technological change. New tools alter workflows. New workflows alter expectations. New expectations alter careers.

Ignoring that process won't stop it.

What remains uniquely human

Perhaps the most revealing moment came near the end of the discussion.

As the conversation turned to synthetic users, AI avatars, and digital twins, Adobe's Pert Eilers redirected attention to a more fundamental question.

What is the real superpower of research?

"It was never capital-R Research," she said. "It was always critical thinking."

That observation cuts through much of the anxiety surrounding AI.

The value of researchers was never their ability to manually code interviews or build spreadsheets. Those activities were merely vehicles. The deeper skill was always interpretation—asking hard questions, identifying blind spots, detecting biases, challenging assumptions, and understanding people in context.

Technology may automate more of the mechanics of research. It may even automate portions of analysis. But it doesn't eliminate the need for judgment.

And judgment remains stubbornly human.

As organizations race to adopt AI, the winners may not be the teams that automate the most work. They may be the teams that become better at understanding what their customers need most—and using AI to act on those insights faster.

Or, as Pert said, "It was never capital-R Research. It was always critical thinking."