Research as Organizational Intelligence : A playbook for research leaders in the AI era

Chapter 1: The Shift

From service delivery to strategic intelligence

The new operating model for research, and why most teams are still running the old one.

Why the old service model is breaking down

In the service model, research operates as a vendor to the rest of the organization. A product manager has a question, commissions a study, and receives findings on a timeline that runs parallel to but separate from the decision the findings were meant to inform. The researcher is skilled, the study is rigorous, and the deliverable is thorough. The product manager decides whether to use it.

This model produces excellent work. However, it also produces three predictable challenges. 

1. The team has already moved on by the time the research arrives.
2. Findings are delivered in “user language" in rooms that speak business language. 
3. The insights generated by each study disappear once the project ends, leaving the next team to relearn what the last one already knew.

None of these challenges is new. What has changed is how much they now cost. AI has compressed product cycles, democratized prototyping, and raised the number of decisions a team can make in a quarter. In that environment, a research function that waits to be asked loses influence faster than it can build it.

Consider the most common experience in a service-model team. A planning meeting happens in Q3. The product roadmap for the next year is being shaped. The research leader is not in the room. A few weeks later, the research team is asked to validate a direction that is already mostly decided. They run a careful study, surface a real concern, and deliver the findings. The concern is acknowledged and filed. The roadmap ships on schedule because the momentum is already too strong to redirect.

The cost is not the individual study. It is the compounding loss of influence that happens when research is consistently too late to change the decision. Over time, the team that could have prevented a bad bet becomes the team that documents why it failed.

What strategic intelligence looks like instead

A research function operating as strategic intelligence does three things the service model does not.

It shapes questions before decisions are locked. When a planning conversation starts, the research leader is either in the room or has already made sure the people in the room have what they need. The research that matters most is not the study that validates a direction; it is the synthesis that informs which directions are worth pursuing in the first place.

It translates findings into business consequences. A user insight is the beginning of the work, not the end of it. The researcher who can say “this friction is costing us conversion in the segment that matters most” operates in a different category than the researcher who stops at “users find this confusing.”

It makes knowledge durable and accessibleA body of research compounds in value only if it is easily accessed. Teams that invest in making past studies findable and reusable can answer new questions in minutes that would otherwise take weeks to commission. That speed is itself a strategic capability.

“The moment research becomes strategic is usually one specific conversation with one specific person who suddenly gets it. You are not trying to convert the whole organization. You are trying to find that person and give them something they can use.”

— Danika Patrick, Design Research and Strategy Lead, Carta

How the rest of this playbook is organized

The next three chapters cover the capabilities that make strategic intelligence possible: processes, tools, and people. Each chapter follows the same structure: what the capability is, why it matters now, what excellence looks like in practice, the plays that get you there, and the pitfalls to avoid. The final chapter operationalizes everything into a ninety-day plan.

Chapter 2: Processes

What modern research workflows look like

Research at decision speed, without losing the judgment that makes findings trustworthy.

For most of history, the research process ran in a predictable straight line. A question was formed, a study was designed, participants were recruited, sessions were run, data was analyzed, findings were written up, and a report was delivered. Then the next question began the cycle again.

AI has not eliminated any of those steps, but it has compressed every one of them. What used to be a six-week sequence can now be a one-week iterative loop, with several rounds of learning happening inside a single product sprint. The research function that takes advantage of that compression can move at the same speed as design and engineering. The one that does not falls further behind with each cycle.

Speed creates new research risks

Fast tools make it easy to confuse motion with progress. A team can now run a study in a day, iterate on a prototype between sessions, and produce a synthesis overnight. That velocity is a real asset. It is also a new source of risk, because the judgment once built into a slower process now has to be applied deliberately at speed. 

Markers of excellence in modern research workflows

The processes described in this chapter are the ones we saw consistently at teams that are running at AI-era velocity without sacrificing the quality of the decisions their research informs.

Research happens before decisions are locked. Planning conversations include a researcher who can bring existing knowledge to the discussion without commissioning a new study. Research is an input to how priorities are set, not an audit of priorities that are already set.

Research and design iterate together. A concept tested on Monday is modified on Tuesday and retested by Wednesday. The prototype is a conversation tool for a research session, not a finished artifact requiring a separate design cycle before it can be tested. Design and research are working on the same question at the same time.

Fidelity matches the question. Low fidelity prototypes are used for discovery, where the goal is a reaction to the concept. Medium or high fidelity is used for refinement, where participants already have a mental model of the product and the question is about improvement. Teams do not default to high fidelity just because AI makes it easy.

Recruitment is precise. The team can describe the target participant specifically enough to recruit against that description unambiguously. For example, not “enterprise software users,” but “mid-market ops managers at companies with 200 to 500 employees who currently use Excel for workflow management.” A well-recruited sample of five produces more actionable findings than a loosely recruited sample of fifteen.

AI scales execution only after human calibration. On a study running thirty or more participants, the researcher runs the first several sessions themselves before handing off to AI-assisted moderation. Those researcher-led sessions build the reference point needed to recognize when AI output is trustworthy and when it is not.

Research is embedded in the documents where decisions get made. Findings are referenced inside product change requests, strategic planning memos, and roadmap documents. Research is not a parallel deliverable to be consulted; it is an input to the document where the decision is written.

Chapter 3: Tools

What excellent insight infrastructure looks like

Research infrastructure that is alive, not archival.

By tools we mean the infrastructure that makes research useful after the study ends: the repository where findings live, the tagging and search systems that make them retrievable, the connections between qualitative insight and quantitative signal, the channels through which findings reach the people who can act on them, and the AI capabilities that make all of the above work at scale.

Most research teams have a knowledge problem they misdiagnose as a volume problem. They assume that running more studies will give the organization more insight. In practice, the bottleneck is almost never the amount of research. It is the accessibility of the research already on hand.

Why it matters now

AI has changed what is possible here in a way that is hard to overstate. A team with five years of interview transcripts and a hundred studies on file can now answer questions across the full archive in minutes. Patterns that would have taken a team a week of manual review to find can be surfaced on demand. The ceiling on what a research team can offer the rest of the organization has risen sharply.

That capability matters because the alternative is still, at most companies, what Jess Holbrook describes as institutional amnesia: research that costs real time and money to generate, gets used once, and then effectively disappears. The same users are recruited six months later to answer a question the earlier study had already answered, and nobody knows the first study exists.

“The most valuable thing I can do in the next hour is not run another study. It is to make sure the study we ran six months ago is answering the question being asked in the product planning meeting right now.”

— Jess Holbrook, Partner Head of UX Research, Microsoft AI

Markers of excellence in insight infrastructure

Insights are searchable and reusable. A product manager preparing for a strategy session can answer the question “what do we already know about this segment?” in minutes, not days. Past research is a first stop, not a last resort.

Findings are stored as reusable knowledge. The repository holds insights as discrete entities separate from the reports they came from, along with the study materials, raw video clips, and quotes that let a future reader understand how the insight was generated and replicate the method.

Research is accessible in the flow of work. People who need a research finding can get it in the tools they already use — Slack, the product change request template, the planning doc — without learning a new system or filing a request with the research team.

Qualitative and quantitative signals are connected. An NPS dip triggers a qualitative investigation rather than standing alone as a conclusion. Researchers can pull behavioral data directly, without routing every question through a data science partner, and combine it with qualitative findings in a single synthesis.

Knowledge is maintained, not just archived. Someone is responsible for the repository. Outdated findings are flagged or retired. Contradictions between studies get surfaced and reconciled. Research knowledge that is not actively maintained depreciates, and untended archives become unreliable.

Chapter 4: PEOPLE

What excellent research leadership looks like

Leadership practices that turn a research function into an organizational habit.

By people we mean the leadership practices and individual capabilities that make a research function influential over time. This includes what the head of research does to build the function’s standing in the organization, and what individual researchers do to translate findings into decisions and maintain quality judgment in AI-assisted workflows.

These practices are the most transferable part of the playbook. Tools change. Processes evolve. The underlying work of building relationships, speaking business language, and protecting the quality of research judgment is durable, and it is where the most impactful changes tend to sit.

Why it matters now

As AI takes over more of the execution, the relative value of the work it cannot do rises. Running a study is increasingly commoditized. Knowing which study to run, which findings will change the decision, and how to make those findings impossible to ignore is not. Leadership capability becomes a larger share of what research contributes, not a smaller one.

The research functions we saw thriving were led by people who treated influence as a craft. They were deliberate about where they spent their time, careful about how they framed findings for different audiences, and clear-eyed about the difference between producing good research and being present when it mattered.

Markers of excellence for research leaders

Leaders understand how decisions actually get made. They know the real flow of authority in the organization, not just the org chart. They know when options close and who shapes them before they do. They attend planning meetings, ask sponsors what keeps them up at night, and time their work to land when decisions are still open.

Leaders build allies and visible wins. They spend early-tenure energy on the teams and individuals who already believe in research, deliver for them memorably, and document the business impact in language that travels. The PM who championed a research project tells the story of how it changed the roadmap. That story reaches rooms the research leader will never enter.

Leaders set quality standards as AI absorbs more execution. They decide which tasks humans continue to do. They sample AI output, require verification on high-stakes findings, and keep the team doing enough hands-on work that the calibration needed to review AI synthesis does not erode. This is quality guardianship at the level of the function, not the individual study.

Leaders build organizational habits, not just project outputs. They invest in recurring practices — a standing Watch Party, a regular insight digest, a query channel — that compound in influence over time. A one-off study is an event. A habit is an infrastructure.

Markers of excellence for individual researchers

Researchers translate findings into business implications. For every user insight, they write the corresponding business sentence. “Users struggle to find X” becomes “this friction is likely contributing to the drop-off we see at this step.” Findings that connect to revenue, retention, or risk get acted on. Findings that stop at user pain do not.

Senior researchers maintain quality judgment in AI-assisted workflows. They continue to conduct sessions, tag their own data, and write their own synthesis often enough to stay calibrated. They do not approve AI outputs they have not stress-tested against source material. Quality guardianship is a perishable skill, and the practice required to maintain it is part of the job, not a distraction from it.

Researchers synthesize across multiple signal types. They bring behavioral data, qualitative findings, market research, and experiment history into a single point of view. The synthesis that used to require a week of cross-team coordination now takes a morning when the researcher has the fluency to direct it. That integrated perspective is what makes a researcher the most valuable person in a strategic planning conversation.

“Researchers can now tap into behavioral data, market research, what competitors are doing, even what experiments we’ve run inside the company and what we learned. It gives them a multitude of signals to come with a very sharp point of view. The researchers who can do that are the ones who will succeed in the next two years.”

— Prakriti Parijat, Head of Research, Instacart

Researchers offer the reframe. The most irreplaceable thing a senior researcher does is help someone see a problem differently: “here is the bigger picture you are not seeing.” This is not pattern-matching across transcripts. It requires understanding the organizational context, the unstated assumptions, and the history of how the problem was framed in the first place.

“The most important thing — the best advice a senior researcher gives — is the reframe. It is the ‘here is the bigger picture you are not seeing.’ I do not get that from AI yet.”

— Jess Holbrook, Partner Head of UX Research, Microsoft AI

“Before you build anything, you really need to go observe someone — an allergy parent shopping for a gluten-free product in the store. I don’t think there is a way to bypass that quickly. You have to go, spend the time, get that nuanced insight, come back, and then build something.”

— Prakriti Parijat, Head of Research, Instacart

Chapter 5: YOUR FIRST 90 DAYS

A plan to strengthen strategic research capability

A practical plan for research and design leaders ready to move from service provider to strategic intelligence. This chapter does not introduce new concepts. It operationalizes the playbook.

How to use this plan

The plan is organized by the same three capabilities as the rest of the playbook: processes, tools, and people. Each section lists the first moves we would recommend a research leader make. Three caveats are worth naming upfront.

The pace is aggressive by design. Most teams will not complete every action in ninety days, and that is fine. The point is to pick a starting direction in all three capabilities, rather than sequencing them and losing the early momentum.

Teams should adjust the order based on which capability is weakest in their current environment.

First moves in processes

• Run a Two-User Gut Check in the first thirty days. Pick a live product question, recruit two participants against a precise profile, and frame the findings explicitly as directional input. The goal is to make research visible in a conversation that would otherwise happen without any user signal.
• Run a Bullseye Customer Sprint in the first sixty days. Define the target participant in precise detail. Recruit five matched participants. Test two or three concepts in a single day with the team watching. Debrief together and document what you learned in your repository.
• Run a Hybrid Study by day sixty. Pick a topic that warrants a larger sample. Run the first five sessions yourself to calibrate on the protocol. Scale to thirty or more participants using AI-assisted moderation. Review the scaled sessions against the reference point built in the first five.
• Match fidelity to purpose from day one. On every prototype-based study you scope in the first ninety days, write down explicitly whether the question is discovery or refinement, and choose fidelity accordingly.

First moves in tools

• Audit your research archive in the first ten days. How much of it is findable? How much is current? What has been used versus filed and forgotten? This diagnostic is the starting point for everything else in the tools capability.
• Stand up the Knowledge Repository by day forty. Five tags, not fifty. Store video clips and quotes alongside reports. The test is whether someone who was not in the room can answer a reasonable question about users using only what is in the system.
• Build the Analysis Skills Library by day sixty. Create the first ten structured AI prompts for your team. Encode your standards for pattern recognition, insight framing, and recommendation structure. Require every researcher to fork and personalize their copy.
• Connect one quantitative source by day sixty. NPS, analytics, or support tickets. The goal is a workflow where a quantitative signal triggers qualitative investigation. Start with one source you can actually integrate, not a comprehensive data layer.
• Launch the Research Query Channel by day eighty. Pre-index the archive, connect it to an AI assistant in Slack, tag source researchers in every response. Announce it first to the allies you built early in the quarter. Let it grow by word of mouth.

First moves in people

• Map the decision-making terrain in the first five days. Who makes decisions, when they close, who shapes them before they do. This is observational work: attend planning meetings, ask sponsors what is keeping them up at night, listen more than you contribute.
• Identify three to five allies and book time with them in the first two weeks. Teams and individuals already open to research. Deliver something useful before the month ends, even if it is a Two-User Gut Check on their most pressing question.
• Schedule a standing check-in with the Head of Product by day forty-five. Monthly at minimum. Bring a summary of what research has been learning and a list of open questions about the company’s highest-priority bets. Frame it as research intake, not a status update.
• Run your first Watch Party by day seventy. Invite your new allies to observe a live research session together, with structured note-taking roles and a debrief. The goal is a shared experience, not a shared deliverable.
• Start the Insight Digest by day seventy-five. A brief, regular summary of findings calibrated to audience. Push it to one cross-functional partner first. Regularity matters more than length.
• Schedule the first C-Suite Listening Tour conversation by day ninety. One senior leader. Listening, not reporting. The output is a better sense of where the company is headed and a relationship that earns research a seat at the planning table.

How to measure early progress

At day ninety, look back at four numbers.

1. How many product decisions this quarter referenced a specific research finding? This is the clearest signal of whether research is being used as an input to decisions, or only as a debrief.
2. How quickly can the team answer a new question using existing knowledge, versus commissioning a new study? A team that can answer in minutes what used to take weeks has moved from service delivery to intelligence.
3. How many teams outside of product and design are actively using research? Marketing, policy, operations, sales. Research reaches them when it is accessible in the tools they already use.
4. What was the cost of an insight this quarter, compared to last? Not a comprehensive metric, but a directional one. A healthy insight system drives the cost per insight down over time.

None of these numbers are easy to track in the first quarter. Asking the questions, and building the systems that make them answerable, is itself a signal of a research function operating strategically.

A closing note

The research teams that navigate this moment well will not be the ones that hired the most researchers or bought the best tools. They will be the ones that made human insight useful faster than the organization around them could stop listening. That is a craft, not a science. It is built one conversation, one study, one converted skeptic at a time.

About Design Better

Design Better is a podcast and Substack by Aarron Walter and Eli Woolery that explores the intersection of design, technology, and the creative process through in-depth conversations with inspiring guests across the creative fields. Whether you are design curious or a seasoned design pro, Design Better is guaranteed to inspire and inform. The show has been recognized by the Webbys, W3 Awards, Vanity Fair, and Architectural Digest.

Aarron Walter

Aarron Walter is a design and technology leader. He started the user experience design practice at Mailchimp and helped grow the company from a few thousand customers to tens of millions. At InVision, he studied design teams at some of the most admired tech companies to identify the traits that influence success. He later joined former CDC Director Dr. Tom Frieden at Resolve to Save Lives, applying design and technology to emergency public health response for the Africa CDC and the WHO. Aarron’s design guidance has helped the White House, the US Department of State, and dozens of major corporations, startups, and venture capital firms. He is the author of several books, including the second edition of Designing for Emotion.

Eli Woolery

Eli Woolery trained in the Product Design program at Stanford University, where he now teaches as a lecturer. He brings a rare breadth of background: photography, filmmaking, product design, and industrial design. He is the former Director of Design Education at InVision, and founded Out of the Deep Blue, a design consultancy whose clients have included Genentech and Kaiser Permanente. A lifelong devotee of the ocean, Eli loves to surf, dive, and kayak.

Subscribe to Design Better at designbetterpodcast.com.