The Boardroom Guide to Buying Better Research in the Age of AI
Index
Executive Summary
The crucial mistakes companies make when buying research in the age of AI
The false confidence problem
What AI can and cannot do
The five research products leaders should stop buying
The roadmap: insight, meaning, decision
How to brief research properly
How to spot weak research proposals
When to use which research approach
The boardroom research checklist
What we measure vs. What drives decisions
From research reports to decision systems
Contacts
TL;DR:
As a leader responsible for high-stakes decisions, you need this guide because research now carries a new kind of boardroom risk.
AI has made research faster, cheaper and more convincing. Used well, this is a major advantage. But when research is fueled by weak questions, generic assumptions, professional panellists instead of consumers or AI synthesis not grounded in real human data, what looks like a shortcut can become a costly illusion.
The danger is false confidence: research that appears rigorous enough to justify investment, but does not actually reduce uncertainty. For Chairs, Boards and senior leaders, this matters because the cost of weak research is rarely limited to the research budget. It shows up later as misdirected capital, weak adoption, reputational damage, internal misalignment and strategic decisions built on a misunderstanding of the people the business depends on.
The central argument of this guide is simple: every high-stakes decision must be grounded in the lived reality of consumers, audiences, staff and stakeholders. Not just what people say they want in a survey. Not just what a dashboard tracks. Not just what an AI model can plausibly infer. But how people actually live, decide, trust, resist, spend, avoid, adopt and attach meaning to products, brands and experiences.
In an environment where many organisations are giving in to shallow, AI-produced research because it is fast and cheap, thorough human-centred research becomes a competitive asset. The companies that invest in deeper understanding will be better equipped to see risks earlier, identify unmet needs before competitors, avoid strategic missteps, and make decisions that are not only opinions-informed but reality-tested.
This guide was created to help boards buy better research, not more research. It explains the mistakes that create false confidence, the limits of AI when it is not grounded in fieldwork, the research products leaders should treat with caution, and the difference between monitoring what is easy to measure and understanding what actually drives decisions.
Most importantly, it gives leaders an actionable framework they can use immediately. The Boardroom Checklist helps you assess whether the research you buy is clearly connected to a decision, grounded in real human complexity, able to challenge assumptions, and designed to translate into action. It also asks whether the work leaves behind reusable intelligence, rather than another one-off report.
The future of research will not be more static decks. It will be living decision systems: intelligence assets grounded in real human data, enriched over time, and made more usable by AI. Organisations that build these systems well will move faster with greater confidence. Those that build them on shallow foundations will simply scale expensive fiction.
This guide is for leaders who cannot afford to mistake polished research for reliable evidence, and who want research to do what it should do: reduce risk, protect reputation, sharpen strategic judgment and help the organisation decide, align and act in time to matter.
The crucial mistakes companies make when buying research in the age of AI
Research is now operating in a fundamentally new environment, and the chance of getting it wrong and losing serious money has increased. AI has made it faster, cheaper and easier to gather information, summarise evidence, generate hypotheses and produce outputs that look impressively complete. Used well, this is a powerful advantage: AI can help organisations investigate higher-value strategic questions with greater speed and depth. But used carelessly, it can also scale weak thinking, unreliable sources and generic assumptions at extraordinary speed - without anybody in the C-suites and board rooms realising it.
Imagine a Chair signing off a £20 million product expansion after reviewing a research deck built around AI-generated insights, a survey of thousands of professional panellists and a set of generic questions asking people what they “would buy” in a future scenario. The deck is polished. The charts are clear. The numbers look solid. The recommendation feels commercially attractive. But the research has never really tested the human reality underneath the decision: whether the actual audience trusts the category, whether the product fits into their lives, whether the stated interest survives social pressure, price tension, habit or emotional resistance. Six months after launch, adoption is weak, internal teams are misaligned, marketing spend is defending a proposition people do not quite believe in, and the board is asking why the research did not catch the risk earlier.
That is the danger. When the research questions are average or wrong, and the research is not based on lived reality, the organisation does not become more informed. It becomes more confidently misled. For boards and senior decision-makers, the real risk is no longer just bad research. It is bad research that looks good enough to justify costly decisions. To state it plainly: bad research wastes budget, misdirect investment and puts millions in future growth at risk.
The false confidence problem
This is the central issue. False confidence occurs when research appears rigorous, even sophisticated, but does not meaningfully reduce uncertainty.
We see this happening across industries and boards: large samples answering shallow questions, synthetic personas built on generic or outdated assumptions, social listening presented as cultural understanding, dashboards that monitor signals without interpreting them, and AI-generated summaries with no grounding in fieldwork.
None of these is inherently flawed, and many can be even useful in the right context. However, they tend to produce confidence and alignment instead of depth and understanding, so the problem emerges when they are treated as sufficient.Take Gen Z, for example. There is often a clear disconnect between how they are described, and therefore how LLMs describe them, and how they actually behave.
They are frequently positioned as the “new consumers.” But many are economically constrained, institutionally sceptical, and culturally moving away from traditional ideas of consumption and status. They are building identity in public, forming trust in smaller networks, and engaging with brands very differently from previous generations.
The result is a generation that is revising consumption as we know it. And yet, much of the research still treats them as a conventional consumer segment. This is exactly how false confidence begins: when research repeats the category language around an audience, instead of intercepting the lived reality that is changing underneath it.
Even strong, institutional metrics can be misleading. We worked with an enlightened CEO who challenged a stagnant industry and its positive NPS, asking: “Will our future consumers still want what we are offering today?” He was right: taking his lead, we discovered a whole world of unmet needs and business opportunities hiding behind a simplified metric. Meanwhile, in trust-critical categories, the real barrier is often the emotional defence that lies behind rational understanding. People may say they understand the product or the service while still avoiding the decision of buying or using it because it inherently forces them to confront emotionally loaded themes such as illness, loss, financial vulnerability or future dependence. Research that only captures declared attitudes can therefore look reassuring while missing the deeper reasons people postpone, distrust or disengage. A prime example of this dynamic can be observed in parenting categories, where judgement often becomes a market force. Parents may understand the value of a product and still avoid it because using it exposes them to criticism, guilt or social disapproval. In this context, the real adoption barrier is not product quality or pricing alone, but the fear of what choosing that product says about them as a parent. Research that only measures preference or purchase intent can therefore look positive while missing the stigma that quietly blocks adoption, loyalty and advocacy.False confidence also happens when organisations confuse wealth with spendability. An affluent audience may have money, but that does not mean they are easy to move. In one study, we saw that people with significant resources were not simply “luxury consumers”; they were highly selective curators. They mixed everyday value buys with purposeful splurges, rejected performative branding, and expected every purchase to justify itself through meaning, quality, utility or legacy.
Another form of false confidence comes from mistaking market presence for cultural presence. Brands can be visible in the market, present in distribution, and supported by internal strategy, while still being culturally absent or lukewarmly perceived where their target audiences are building meaning, which is one of the strongest early indicators of future loss of awareness and consideration. In one project the real issue was that brands were largely absent from the conversations, rituals, creator ecosystems and cultural spaces where those young audiences’ behaviours were taking shape. For leaders, this matters because if a brand is not part of the environments where meaning is being made, it becomes harder to shape relevance, trust or adoption later.
For boards, this disconnect is seldom visible early. It typically emerges later, after decisions have been taken and capital deployed, through signals such as underperformance, weak adoption, or backlash. At that stage, the impact of false confidence gets to strategic positioning and, ultimately, financial loss.
What AI can and cannot do
Used well, AI is an extraordinary tool. We are big fans of it and pioneered AI in consumer research ten years ago.
AI gives research greater range, speed and usability. But it becomes risky when it is used to replace contact with real people, or to generate answers that are not clearly grounded in lived reality. It can flatten nuance, over-index on what is easy to measure, and present outputs with more confidence than their foundations justify.
The danger is subtle: answers that are not necessarily wrong, but incomplete, and presented as sufficient.
A funny example is Napoleon. During a client project, we asked leading AI tools to identify who is influencing affluent consumers in France today. The answer came back quickly and confidently, and it had Napoleon at the top of the list.
As humorous as this output may sound, it is a revealing one. If we stretch the concept, it might even be right in a historical sense, but it still clearly exposes the difference between what is plausible and what is useful, current, and decision-relevant.
That’s why AI can’t and shouldn't replace human research, but it can and should be trained by it.In one recent project, this became the central principle: rather than enhancing research with AI, we enhanced AI with research. We built a qualitative system grounded in ethnography, psychology and human-AI expertise to capture real human complexity, then translated that into structured intelligence that could inform synthetic personas and future AI tools for the client. The point was simple: AI is only as good as the human truth it is trained on. If the underlying research is shallow, the model will simply scale shallowness. If the underlying research is ethnographic, psychologically rigorous and grounded in lived reality, AI becomes far more useful for business decisions.
The five research products leaders should stop buying
Seen through this lens, certain types of research begin to show their limits.
Data dashboards are valuable for monitoring but rarely sufficient for finding meaning.
Surveys often ask people to rationalise decisions they do not fully understand.
Panels can deliver clean answers quickly, but from the wrong people (people that answer as a job as opposed to actual consumers).
Trend reports describe what is happening, but often fall short of explaining why it matters in a specific context.
And increasingly, AI-generated personas, when not grounded in real human data, create a compelling illusion of understanding without the substance behind it.
None of these should necessarily be discarded. But they should be recognised for what they are: partial tools, not decision frameworks.
The roadmap: insight, meaning, decision
What is required now is a different standard for what research delivers. At a minimum, it must move across three layers.
First, insight: what is happening.
Second, meaning: why it matters culturally, psychologically, emotionally.
And finally, decision: what should be done as a result.
Most research comfortably reaches the first layer; some reach the second, and very little consistently reach the third. And yet, it is only at that point that research becomes truly valuable.We worked with global luxury brands sitting on a staggering amount of data, and still, they were almost paralysed when asked to create an app, a campaign or an experience. The turning point was grounding their know-how into the lived experiences of both their consumers and staff/stakeholders. They needed to understand what to do next and bring everyone on board to make it happen. In other words, research only starts to create value when it stops being a description of the audience and becomes a shared decision-making tool for the organisation.This is especially true when a company is building a new product, platform or experience that has to be delivered across different parts of the organisation. In one luxury app project, the research could not stop at what customers and prospects wanted from digital luxury; it also had to understand how store teams worked, how the selling ceremony happened, how after-sale support was experienced, and how the physical brand world connected to digital identity. That changed the role of research: it was not simply informing an app, but helping design an ecosystem that staff could deliver, customers could recognise, and the brand could credibly own.
Also, let’s walk in the shoes of your company staff and stakeholders: a persona, segmentation or insight platform does not hold the same value for brand, product, sales, fundraising, operations or communications in the same way. Each department interprets it through the lens of its own priorities and constraints. The real question is: what does this insight change for each team’s decisions? Only when insight is translated into decisions that different parts of the organisation can use and act upon does it move from being informative to becoming operational.
How to brief research properly
The quality of research is largely determined before it begins. Many briefs still start from a general ambition: “We want to understand our audience.” This tends to produce broad, descriptive outputs that are difficult to act on.
Stronger briefs start with the decision that needs to be made. What is at stake if it is wrong, what the organisation currently believes, and where uncertainty lies.A useful brief should also define the internal users of the research. Who needs to believe it? Who needs to act on it? Which teams will translate it into communication, product, experience, service or operational change? Research is more valuable when it is designed not only around the audience being studied, but around the decisions it must unlock inside the organisation.When framed this way, research becomes sharper, more purposeful. It also becomes easier to assess whether it is answering the right question, rather than just producing an answer.
How to spot weak research proposals
For experienced leaders, weak research proposals are often recognisable early. They tend to lead with methodology rather than the decision at hand. They promise speed without clarifying trade-offs. They rely heavily on demographics or treat AI as a substitute for understanding rather than a tool within it.
More importantly, they often fail to explain how findings will translate into decisions. There is no clear “what now” or “so what” layer. No mechanism for dealing with contradiction, no recognition that different stakeholders shape outcomes.Another warning sign is when a proposal studies only the obvious audience. In many categories, the buyer is not the only person shaping the decision. Advisors, brokers, frontline staff, family members, creators, peers or internal teams may all mediate trust and influence behaviour. Strong research asks who really shapes the decision, formally or informally.In high-consideration categories, the buyer is rarely acting alone. In one destination wedding project, the decision was not shaped only by the couple, but by a wider trust ecosystem: vendors, photographers, planners, bridal media, family members, creators and the visual codes circulating across the category. Mapping that ecosystem showed where trust, aspiration and authority were actually being built, allowing the brand to move from generic visibility to targeted influence. This is the difference between knowing who the customer is and knowing who makes the customer feel safe enough to choose.
When to use which research approach
Every research methodology has its place. The problem is when organisations reach for the familiar one rather than the right one. The choice should always be driven by the company's goals and the decision at hand, ensuring the research is tailored to real needs from the outset and not retrofitted to them. Here are some examples from our work where situation and decision shaped the choice of methodology:
Entering a new market requires an analysis of cultural context, local meaning systems and trust dynamics. Semiotics and lingo analysis are the MVP, and a full ethnography/netnography the ideal methodology. The point is not only to understand how a category is perceived today, but to see how it is already being remixed by people, creators and communities, so leaders can build around what the market is ready for next, creating a real advantage over competitors.
Launching a product demands clarity on what might stop people from using it, what would make it feel relevant in their lives, and how likely they are to actually adopt it, moving from interest to real-world use. Observing behaviour through ethnography and netnography is then recommended - followed by quantitative validation through surveys.
When a brand begins to lose relevance, the issue is rarely visibility alone and often lies in a shift in meaning, perception or cultural position. In these moments, existing data can show that something is going wrong, but not necessarily why the audience is drifting away or what would make them care again. The strongest approach often starts by digesting what the organisation already knows, then entering the audience’s cultural fabric through netnography, Cultural Opinion Leaders and co-creation, so that strategy, creative and commercial teams can rebuild relevance around what people actually value now. Similarly, in moments of reputational risk, what matters most is often what is not being openly said: underlying fears, tensions and narratives within communities. In these cases, intercepting Cultural Opinion Leaders of the target audience and engaging with them in In Depth Interviews delivers the best ROI for research.
In categories that are emotionally sensitive, such as health, money, ageing, eating and drinking, the gap between what people say and what they feel becomes even more pronounced. In these contexts, the right approach is rarely a standard survey alone. You may need cultural analysis to understand how the topic is discussed in the wild, Taboo Topics Protocols and ethnographic interviews to access what people struggle to say directly, observation to see how decisions actually unfold, and validation to understand which patterns scale. The method should follow the emotional complexity of the decision.
When the organisation is about to make a costly, difficult-to-reverse decision, the research standard should be even higher. In those moments, research should validate which ideas are strong enough to carry investment. The strongest approach is often layered: qualitative work to reveal what people genuinely need, cultural foresight to understand what will still matter in the future, multi-audience research to identify who must believe in the change and bring it forward, and quantitative validation to stress-test the most promising concepts before capital is committed. This is how research moves from inspiration to investment confidence.
In all these cases, the question is how close the research gets to lived reality, and the method should be decided accordingly.
The boardroom research checklist
As a practical framework, we have outlined a set of questions to help you evaluate the strength and relevance of research:
Does the research clearly connect to a decision?
Does it articulate what is at risk?
Does it capture real human complexity, beyond stated behaviour?
Does it study the full decision ecosystem, not just the obvious customer?
Does it combine different types of evidence (behavioural, cultural, psychological)?
Does it challenge assumptions, or just reinforce them?
Can it translate into action and what are the staff/stakeholders implications?
Does it include a plan for adoption inside the organisation?
Does it leave behind something reusable, such as an asset, rather than a one-off output?
Research that meets these criteria is far more likely to support confident, high-stakes decisions. When it does not, it risks informing without truly influencing what gets done.
What we measure vs. What drives decisions
Across sectors, the pattern is consistent: research often focuses on what is easiest to measure, rather than what actually drives behaviour and decisions. This pattern becomes particularly visible when looking at different industries. We observed it extensively while working on projects deeply grounded in specific fields:
In luxury and fashion, the challenge is whether we are decoding identity and aspiration, or simply tracking preferences.
In retail, the focus should move from monitoring transactions to understanding lived experience.
In hospitality, the question shifts from the service delivered to the transformation people expect to feel.
In financial services and insurance, we should go from studying declared rational preferences to considering the deeper emotional system around fear, control, avoidance, trust and institutional scepticism.
In healthcare, what is said openly rarely reflects the full emotional reality.
In technology, adoption depends as much on trust and cognitive load as on functionality.
Each sector has its own version of this gap. But the underlying issue is the same: the distance between observable behaviour and underlying meaning.
From research reports to decision systems
As AI continues to evolve, the organisations that benefit most will go beyond producing more research and into building decision systems.
These systems should not be static reports that capture a moment in time and slowly lose relevance. They should be living intelligence assets: grounded in real human data, updated as behaviours and culture shift, enriched over time, and designed to support multiple decisions across the organisation.
AI will play an important role in making these systems more powerful and more usable. Think of an agents’ swarms that pick the right insights for the right team at the right moment, or an internal intelligence layer that allows leaders to interrogate research directly: asking how a specific audience might respond to a product, message, experience, or market move, and receiving answers grounded in real ethnographic, cultural and behavioural evidence.
But their value will depend entirely on the quality of the human reality they are built on. If the underlying intelligence is generic, stale or poorly grounded, AI will simply scale errors, accelerate poor decisions and turn weak understanding into wasted investment. If it is built from real people, cultural context, behavioural evidence and psychological depth, AI can help turn research into a living layer of organisational judgment, creating a significant competitive asset for the organisation.Ultimately, what matters is not the amount of data available, but whether it changes what happens next.
Research should move beyond describing people and into helping organisations decide, align and act with confidence, in time to matter.
If this guide has raised questions about the quality, usefulness or decision value of the research currently informing your organisation, I would be happy to offer a private 45-minute consultation.
The purpose is simple: to look at one live decision, challenge or strategic uncertainty you are facing, and help you assess whether research could genuinely reduce risk, sharpen direction or support implementation.
It will be a confidential conversation with me, designed to help you clarify what you may need, what you may not need, and what kind of evidence would actually make the next decision easier.
If useful, I can then translate that conversation into a lean recommendation for the most appropriate next step.
To arrange a private conversation with Alessia Clusini, CEO of Trybes Agency, please contact: martinafaralli at trybesagency.com