Why Your Insights Team Already Has the Answers (And Can't Find Them)

The best market researchers have always known how to play a role that almost no one gives them credit for: librarian. Not the clerical kind. The strategic kind. The person who can look across years of accumulated organizational knowledge and say, "We already have the answer." Last April, I hosted a roundtable as part of an event with 40-plus brand-side and agency research leaders in New York. The most sophisticated teams in the room were talking about giving their librarian the infrastructure she has always needed.

What Is the "Librarian" Role in Insights Teams, and Why Does It Matter Now?

A senior insights leader at a major consumer goods company offered the clearest framing I’ve heard. A good researcher, she said, "is a good analyst, a good critical thinker, a good social scientist, and also a good librarian." By librarian, she meant the person who can look across the organization's accumulated knowledge and say: "We already studied this. We don't have to do it again. Let me connect the dots for you."

What she said stuck with me because it captures something the industry has talked around for years. The researcher who plays the librarian role well shouldn’t be mistaken for doing clerical work. She’s strategic. She's the reason the organization doesn't re-fund research it already paid for. She's the reason the new CMO's question about customer price sensitivity gets answered in an afternoon instead of a quarter.

Researchers understand the value of this role. The problem is that, until recently, the available tools and human limitations made it impossible to sustain. Research died in slide decks and SharePoint folders not because no one wanted to surface it, but because there was no practical way to make it queryable at the speed the business operated. The librarian's institutional memory could only scale as far as one person's ability to remember.

How Are Leading Insights Teams Using AI to Scale Institutional Knowledge?

Here is what I observed at the roundtable.

A brand insights leader at a consumer goods company described their path to building an organizational memory. They started with a problem every insights function knows: years of legacy research, locked in formats and folders that no one could find and fewer people read. Their solution to getting it uploaded was blunt: $100 gift cards. In six months, 3,000 legacy reports made it into the system. From there, they deployed an internal LLM to index 15,000 documents. The result: brand managers and senior vice presidents can now self-serve answers to strategic questions that used to require a two-week research cycle to produce.

That's the librarian role, operationalized by a modern, AI-enabled internal team with the resources to make it happen. The researcher who used to carry that knowledge in her head, and surfaced it when she happened to be in the right meeting at the right time, now has infrastructure that can surface it for anyone in the organization, on demand. The pattern holds across enterprise insights teams: research itself is rarely the constraint. The infrastructure for finding it is.

What I noticed about this team was their order of operations. Instead of trying to replace the researcher or automate a study from the start, they started by capturing what they already owned.

How Does AI Change What the Research Function Is Actually For?

The objection I've heard from insights professionals is that a queryable archive sounds useful in theory but less relevant in practice, because the business questions are new and hard to anticipate.

That's only partially true.

Some questions are genuinely new. They require fresh research, skilled moderation, and careful analytical framing. Those are the questions the researcher should be spending her time on. The problem is that in most organizations, the research function is spending significant capacity on questions that aren't new at all. Questions that were studied two years ago, or four years ago, by a team that no longer works there, in a format nobody can find. The insights existed. The organization just couldn't access them at the speed the business demanded, so someone requested a new study instead.

When the librarian function can scale, it changes what the research team is actually for. The researcher's expertise gets applied to questions that actually require her: new questions, hard problems, interpretive work that accumulated data can't answer on its own. The budget that was going to redundant research can go to harder questions.

How Much of Your Research Budget Is Going to Questions You've Already Answered?

The question I'd put to any insights leader reading this: what percentage of your research spend this year went to questions your organization had already paid to answer?

Most people I ask don't know. The number is hard to measure, which is part of why the problem persists. The studies get commissioned, the reports get filed, the platform changes or the team turns over, and the knowledge disappears from circulation while the budget line stays intact.

The teams that are furthest along on this problem have done something simple but underestimated: they've made it someone's job to know what the organization knows, and invested in the resources and infrastructure to make it accessible. Not just to commission new studies, but to surface the ones that already exist before a new project gets funded. The librarian role, formally recognized and properly resourced.

What AI adds is scale. One person's institutional memory can cover a few years of work, if she's exceptional and if she's stayed long enough. A properly indexed archive of 15,000 documents covers decades. It doesn't leave when the person does. And its value compounds over time.

The most sophisticated insights leaders I met in New York had arrived at the same conclusion: AI, for them, was the infrastructure that lets the research function do what the best researchers have always known how to do. The librarian role has always been strategic. What's new is that with AI, it's becoming scalable.