Three plays to fix the broken research process
A familiar scene plays out in product organizations every day.
A team is wrestling with an important decision. Perhaps they're debating a new feature, trying to understand a sudden drop in engagement, or exploring a promising new direction for the product.
The researcher in the room faces a choice. They can insist on following the process: scope the study, write the discussion guide, recruit participants, run sessions, synthesize findings, and present recommendations several weeks later. Or they can acknowledge what everyone already knows: by the time the research is complete, the decision will almost certainly have been made.
For years, we've accepted that good research takes time. But that assumption is becoming harder to defend.
AI has compressed nearly every stage of product development. Teams can move from idea to prototype in a matter of hours. Concepts that once took weeks to visualize can now be generated in a single afternoon. Product organizations have become dramatically faster.
But speed alone isn't the goal. As Design Better co-founder Aarron Walter put it in our on-demand webinar, Design confidence under pressure, AI can help teams “build products faster, get them to market quicker.” But if what they're building isn't right for the market, he cautioned, “that's still a fundamental problem.”
The challenge, then, isn't simply making research faster. It's making sure human insight can keep pace with product decisions.
This doesn't mean research quality has declined. In many cases, the opposite is true. Researchers are producing thoughtful work, asking better questions, and generating richer insights than ever before.
Yet many are still finding themselves further from the decisions that matter most. But we don’t think they need better research.
They need better workflows to integrate with the rest of the organization.
For our new Research as Organizational Intelligence playbook, created in collaboration with Design Better's Aarron Walter and Eli Woolery, we spoke with research leaders at Microsoft AI, Instacart, Carta, and Google Ventures. A different pattern emerged.
The teams gaining influence weren't producing dramatically more research than everyone else, nor were they sacrificing rigor in pursuit of speed. They had simply redesigned their workflows around a reality most organizations are still struggling to accept: product development has accelerated, and research has to accelerate with it. Three practices surfaced repeatedly.
PLAYBOOK
Research as organizational intelligence
We partnered with the team at Design Better to create a comprehensive playbook for how to transform your research team to thrive in the age of AI.
The two-user gut check
Researchers spend a lot of their careers explaining why small sample sizes can be misleading. Which makes it somewhat uncomfortable to advocate for a method built around talking to only two people.
Yet the research leaders we interviewed kept returning to the same observation: in fast-moving organizations, the comparison is rarely between a two-person study and a thirty-person study.
More often, it's between a two-person study and no study at all.
Imagine a product team facing a decision that needs to be made by Friday. Everyone agrees user input would be valuable. There simply isn't time to recruit a large sample, conduct a full study, and synthesize the results. As a result research disappears from the conversation at that point and the decision proceeds on instinct.
The two-user gut check offers another option.
Recruit two participants who closely match the target audience. Run focused sessions around the question at hand. Be explicit about what the output represents: not validated findings, but directional input.
That distinction matters.The goal is not certainty. The goal is to expose assumptions before they harden into decisions.
Two conversations can reveal surprising blind spots. They can surface obvious points of confusion. They can challenge a team's confidence in an idea that looked promising in a planning meeting but falls apart when confronted with real users.
That direct contact with users matters even more as AI accelerates everything around it. “You can, again, create prototype things very quickly,” said Design Better co-founder Eli Woolery. “But is it something that humans need? Is it something humans want?”
For Eli, maintaining that connection to customers becomes increasingly important precisely because so much else can now be accelerated or automated.
Nobody should mistake those conversations for a comprehensive study. But in many situations, they are enough to change the quality of the discussion.Research leaders who use this approach well aren't lowering the bar for evidence. They're recognizing that influence often depends on being present before certainty is possible.
The bullseye customer sprint
If the two-user gut check challenges our assumptions about sample size, the bullseye customer sprint challenges our assumptions about recruiting.
Most teams spend considerable energy debating how many participants they need. Far fewer spend the same energy defining exactly who those participants should be. As a result, recruitment criteria often remain surprisingly broad.
"Enterprise software users."
"Small business owners."
"Knowledge workers."
The categories sound precise until you start recruiting against them.
One of the strongest ideas we encountered during our interviews came from a simple observation: participant quality matters far more than most teams realize.
Before recruiting begins, define the target participant with enough specificity that two different recruiters would find essentially the same people.
Not just “enterprise software users,” but “Operations managers at companies with 200 to 500 employees who currently manage workflows in spreadsheets.”
Not “online shoppers” but “Parents of children under ten who purchase specialty dietary products at least twice a month.”
This level of specificity changes the entire study because recruiting becomes easier and findings become clearer. Stakeholders spend less time arguing about whether participants were representative and more time discussing what was learned.
Several leaders described situations where five carefully recruited participants generated more useful insight than fifteen loosely matched ones. The competitive advantage isn't running more sessions. It's talking to the right people.
The hybrid study
If there was one theme that surfaced in every conversation, it was caution about automation.
The leaders we spoke with were enthusiastic about AI's ability to accelerate research workflows. Many were actively experimenting with AI-assisted moderation, synthesis, and analysis.
But they also worried about a subtle risk. As AI becomes more capable, researchers may gradually lose the hands-on experience required to evaluate its output. The danger isn't that AI produces obviously bad work. It's that it produces work that sounds convincing.
The danger isn't that AI produces obviously bad work. It's that it produces work that sounds convincing.
Aarron sees that tension firsthand. He uses AI to sift through large datasets and identify patterns that would be difficult to find manually. “Any sort of pattern recognition and discovery is super helpful,” he said. But he added an important caveat: “I would still verify because I also watch AI regularly hallucinate.”
His approach is telling: use AI to find where to look, then preserve direct contact with the source material and the people behind it.
The strongest teams have developed a simple response. Before scaling a study with AI, they conduct the first sessions themselves. Five sessions came up frequently as a useful benchmark.
These early conversations accomplish several things at once. They reveal weaknesses in the discussion guide. They surface unexpected themes. Most importantly, they help researchers develop an intuitive understanding of what the data actually looks and feels like.
Only then do they begin introducing automation.
Eli describes the skill researchers are protecting as a combination of “craft,” “taste,” and “judgment”—the accumulated experience that gives someone a “spidey sense” when an AI-generated interpretation doesn't quite ring true.
One research leader described those initial sessions as “quality infrastructure.” Without that foundation, researchers risk outsourcing not just execution but judgment.
With it, they can scale confidently, knowing they have a reference point against which to evaluate AI-generated outputs.
The future of research is unlikely to be fully manual. It's equally unlikely to be fully automated.
The more useful distinction may be between offloading work and offloading judgment.
Aarron describes AI at its best as a thought partner—something that can broaden your thinking without removing you from the chain of reasoning. That's also what connects these three research practices.
The two-user gut check keeps human evidence in decisions that might otherwise proceed without it. The bullseye customer sprint makes sure that evidence comes from the right people. And the hybrid study uses AI to expand the reach of research without outsourcing the judgment that makes the findings trustworthy.
That's the larger shift explored in Research as Organizational Intelligence: research doesn't become more strategic simply by producing more studies. It becomes more strategic when human insight is available at the speed, place, and moment decisions are actually being made.
ON-DEMAND WEBINAR
Designing the insight system of tomorrow: how UXR Leaders can shape the role of research in the age of AI
Watch this on-demand webinar to learn from how research leaders at Slack, Salesforce, Adobe and Siemens are transforming the way their teams operate in the age of AI.
