Your research repository is probably a graveyard. Here’s how to change that with AI
Every research leader has lived through some version of this conversation.
A product manager asks a perfectly reasonable question during planning.
"Have we ever looked at this before?"
The room goes quiet. Someone remembers a study from last year, though nobody is quite sure what it concluded. Someone else recalls seeing a few relevant interview clips in a quarterly presentation. A researcher volunteers to dig through old folders after the meeting.
Two weeks later, the team commissions another study.Not because the organization lacked the answer, but because it had forgotten where the answer lived.
It's an oddly common pattern. Companies invest enormous amounts of time and money understanding their customers, yet much of that understanding has an astonishingly short shelf life. A study is completed, the findings are presented, decisions are made, and the work slowly disappears into a repository that becomes a little harder to navigate with every passing quarter.
The irony, of course, is that research teams have become incredibly good at generating knowledge. They know more about their customers than ever before. The problem is that very little of that knowledge compounds. Every new project begins with the assumption that the question is new, when in reality the organization may already know part of the answer.
For a long time, this felt like an unavoidable limitation of research. Repositories were built to store documents, not answer questions. Finding a useful study often depended on remembering that it existed in the first place, and knowing exactly what it had been called. Even well-maintained libraries eventually became archives—valuable, respected, and rarely visited.
AI changes that equation in a surprisingly fundamental way.
During our conversations with research leaders at Microsoft AI, Instacart, Carta, and Google Ventures, we expected to spend a lot of time talking about AI-generated discussion guides, automated moderation, and faster synthesis. Those topics certainly came up. But the most interesting conversations were about something else entirely.
Memory.
The leaders we spoke with weren't asking how AI could help them conduct more research. They were asking how it could help their organizations remember what they had already learned.
That shift in thinking changes the role of research infrastructure. The goal is no longer to build a better repository. It's to build an organizational memory that grows more valuable with every study instead of quietly resetting after each one.
Three ideas surfaced again and again.
DOWNLOAD
Get the playbook
We partnered with the team at Design Better to create a Research as Organizational Intelligence, a comprehensive playbook for how to transform your research team to thrive in the age of AI
Build for the question, not the project
Research teams naturally organize their work around studies. Every project has a beginning and an end, a report, a folder, and a presentation.
The rest of the organization doesn't think that way.
Product managers don't remember the names of studies. They don't remember whether the onboarding work happened last spring or last fall. They simply want to know what the company has already learned about onboarding.
That's why leading research teams are beginning to optimize their repositories around questions instead of projects. AI makes this increasingly practical. Rather than asking people to navigate folders and taxonomies, they can ask a plain-language question and retrieve insights drawn from multiple studies, complete with the evidence behind them.
The repository stops being a place where research is stored and becomes a place where questions are answered.
Get out of the repository and into the conversation
One story from Instacart illustrates this shift particularly well.
After the research organization shrank from forty researchers to seventeen, demand for customer insight didn't shrink with it. If anything, more teams across the company wanted access to research than ever before.
The obvious response would have been to improve the repository.
Instead, the team asked a different question.
Why are we expecting people to come to the repository at all?
They built a research assistant into Slack, where product managers, marketers, operations teams, and executives were already spending their day. Anyone could ask a question in plain language and receive an answer grounded in previous research, along with links back to the original studies and the researcher who conducted them.
This is a fundamental change in how we think about research portals.
For years we've treated repositories as destinations. If someone needed customer knowledge, they were expected to stop what they were doing, search for it, interpret it, and then return to the work at hand.
The best teams are beginning to reverse that relationship. Research comes to the meeting instead of waiting in the repository, bringing the insights into the environment where conversations (and decisions) are already being made.
WEBINAR
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.
Connect the dots before running another study
One of the underlooked advantages of using AI is its ability to recognize patterns across old insights rather than just generating new ones.
Customer interviews, NPS results, product analytics, support tickets, sales calls, and usability studies have traditionally lived in different systems, owned by different teams, and answered different questions. Bringing them together required considerable manual effort, which meant it happened less often than it should have.
Several research leaders described moving toward a different workflow.
A drop in NPS no longer triggers an immediate research request. It first triggers a search across everything the organization already knows. Existing interviews are revisited. Support conversations are examined. Previous studies are surfaced. Product data provides additional context.
Only then does the team decide whether a new study is actually necessary.
Instead of treating every project as the beginning of a conversation, organizations start each new question with the benefit of everything they've already learned.
Use AI for durability, not just speed
When people talk about AI transforming research, the conversation usually revolves around speed. Faster recruiting. Faster analysis. Faster reporting.
Those gains are real.
But after spending time with research leaders who are living through this transition, they no longer feel like the most interesting part of the story.
The more profound change may be that customer knowledge is finally becoming durable.
For the first time, organizations have a realistic chance of building systems that remember what they've learned, connect insights across time, and make years of research available at the moment decisions are being made. That’s what gets us to true organizational intelligence.
