Overcoming the three tiers of groupthink

In 1972, the social psychologist Irving Janis first documented the mechanics of groupthink. He described how the desire for harmony and cohesion in a group often overrides the necessity of considering alternative solutions to a problem. Janis’s original work focused on historical policy fiascos, such as the Bay of Pigs invasion and the Vietnam War. Poor decisions taken in these situations shared a critical root cause: structural flaws in decision-making that consistently led smart, well-intentioned people to take sub-optimal or irrational actions.

In B2B thought leadership, the stakes may be considerably lower than these geopolitical crises, but the problem is identical. I have seen this exact dynamic play out in countless planning meetings. When capable, well-meaning marketers, analysts and executives get together to produce thought leadership, the desire for internal harmony often overrides the need for original insight. But worse still, the problem of groupthink isn’t just limited to decision-making within small, internal groups. There is also a wider, macro groupthink at play across the broader business landscape, which is now being made more acute by a third layer: the consensus building machines of large language models. These three tiers of groupthink create a strong drag towards herd-like behaviour and present powerful barriers to creating distinctive, original insight.

Tier 1: The micro filter

When Irving Janis examined how groups make decisions, he identified several distinct symptoms of groupthink, three of which appear just as frequently in B2B marketing committees as they did in government policy teams.

First, Janis observed that certain group members appoint themselves as mindguards, protecting leadership from negative or contradictory information that might shake their confidence or disrupt unity. In B2B thought leadership, we see a similar phenomenon when communications managers, compliance teams and risk managers quietly soften provocative findings before executive leadership or the market sees them. Commercial guardrails have their place, of course, and nobody is advocating a reckless approach to messaging, but there is a vast difference between managing genuine legal risk and watering down an argument simply because it might spark debate. Too often in thought leadership, sharp conclusions are blunted and defensive caveats are added. Anyone who has waited weeks for a legal or compliance sign-off knows how easily a clear, directional finding can be transformed from a bold hypothesis into a list of cautious qualifications. By the time the draft reaches the desk of senior stakeholders for review, it has been quietly watered down and stripped of the arguments that made it compelling in the first place.

The second symptom is collective rationalisation: the process by which groups invent elaborate justifications to explain away anomalies and evidence that challenge a chosen narrative. In thought leadership, this occurs whenever empirical data contradicts a pre-approved story. The pressure to make the data fit the brief in these projects is immense. Suppose a consultancy commissions a survey intending to prove that companies are increasing investment in sustainability. When the data comes back showing the exact opposite, panic ensues. Instead of confronting that inconvenient truth, which would yield a groundbreaking, contrarian report, the committee buries the outlier or spins the numbers to align with a pre-agreed message.

Finally, when senior teams work together behind closed doors, they can easily develop an illusion of invulnerability, a subconscious belief that their past success and pedigree insulate them from getting it wrong. In corporate content, this creates an unintended blind spot. Established firms naturally lean on their hard-won credibility, sometimes assuming that a point of view will carry weight simply because of the brand behind it. In doing so, teams can inadvertently overlook the need to test or challenge a thesis because they are relying on truisms that served them well in the past. This makes it difficult to break new ground or create truly original thinking.

Tier 2: the industry herd

Looking outside the company walls, we find another tier of groupthink. When companies are planning a thought leadership campaign, one of the first steps they take is to look at what competing firms and others operating in a similar space are producing. This competitive benchmark is important. After all, no one wants to produce something that simply repeats what a close rival is doing. Yet rather than leading to a conclusion that the campaign ought to be radically different from everything else out there, the exact opposite often occurs. If all your competitors are publishing papers on “Responsible AI Governance,” it’s a natural response to consider that your firm needs one too. Failing to have a view on a trending topic feels like a gap in the portfolio and is viewed as ceding ground to the competition.

This is a natural response, yet it ultimately produces a herding phenomenon that sociologists call institutional isomorphism. This is the tendency for organisations in the same field to mimic each other’s structures and outputs in order to appear legitimate. In marketing, it leads to sameness rather than distinctiveness. Copying the market’s agenda feels safe: no board ever fired a marketing team for commissioning research on a topic that McKinsey and Deloitte had already covered. Yet this safety is an illusion. Of course, no one wants to be contrarian just for the sake of it. But adopting a consensus perspective doesn’t confer authority on your positioning. It simply blends into the noise that senior buyers already struggle to filter.

Tier 3: The algorithmic layer

Committee-like groupthink and industry herding have been problems for decades, but now companies need to contend with a third, perhaps even more insidious, tier that has been added to the thought leadership process: generative AI. When marketing teams use Large Language Models (LLMs) to brainstorm themes, outline white papers or “polish” executive drafts, they are inadvertently watering down their original thinking and dragging it towards the middle ground.

By design, generative AI does not discover contrarian truths or experience the messy operational friction where real insight is born. An LLM predicts the next most statistically probable word based on vast bodies of existing human text. Its default output is, quite literally, the average or midpoint of conventional wisdom.

This introduces two new problems into thought leadership production:

• AI as the invisible mindguard: Prior to AI, human mindguards used to manually edit bold drafts to make them safe for corporate consumption. Today, writers and subject matter experts can bypass this process by feeding rough, provocative notes into an AI tool with a simple prompt: “Make this sound more polished and executive-ready.” In seconds, the model sands off every sharp edge, replacing a distinctive voice and uncomfortable truths with smooth, diplomatic prose.
• Editorial model collapse: In computer science, model collapse occurs when AI systems are trained repeatedly on AI-generated outputs, causing outliers and variance to vanish until the system converges on a bland, repetitive mean. B2B thought leadership is undergoing real-time editorial model collapse. Firm A uses AI to summarise existing industry thinking; Firm B uses AI to scan the web for trends and summarises Firm A; Firm C uses AI to draft its report and incorporates themes from Firms A and B. With every cycle, original perspectives are diluted, and the entire sector converges on a similar narrative.

Breaking the circuit: antidotes to groupthink

So how can thought leadership teams break out of these three tiers of groupthink? I would suggest four practical changes to how ideas are sourced, tested and approved:

1. Hunt for the anomaly

The research process in most thought leadership projects is structurally designed around consensus. A hypothesis is set based on past information or on a competitor review of what is already out there, and then a survey questionnaire is designed that explores a topic within a set, rigid framework. Respondents are asked questions but the choices of answers they can give are pre-determined. There is little room for outliers or surprise because that is not designed into the process. Instead of designing a quantitative survey around safe, familiar industry themes, start with open-ended qualitative conversations that probe where conventional wisdom is breaking down. Or use emerging methods such as AI-led qualitative research that enable a more open conversation and break free from overly structured questionnaires.

2. Appoint an official “devil’s advocate”

In his original recommendations for overcoming groupthink, Irving Janis argued that organisations must appoint a respected member to challenge the consensus. Relying on dissent to happen naturally fails because employees fear being seen as awkward. But by assigning a designated dissenter with an explicit mandate to attack the draft’s core hypothesis, hunt for industry truisms, and articulate counter-arguments, this problem is neutralised. At the same time, teams should make the process of editing collaborative and open. This forces reviewers to challenge arguments openly rather than quietly removing them behind the scenes.

3. Run a campaign premortem

Derived from decision theorist Gary Klein’s strategy work, the premortem is an antidote to the illusion of invulnerability. Before launching a thought leadership campaign, gather the authors and reviewers for a meeting with one question: “Assume it is 12 months from today, and this report generated zero client conversations and was ignored by the market. Why did it fail?” A premortem gives team members psychological cover to voice the doubts that they may have silenced to maintain harmony during planning meetings.

4. Flip AI from author to dissenter-in-chief

Resist the temptation to use generative AI for ideation, outlining or “polishing” the voice of your drafts. Instead, invert its role in the editorial workflow. Feed your human-derived thesis into the model and prompt LLMs to act as a sceptical reader, tear arguments apart and provide counter-arguments to challenge the thesis. When used as an author, AI drags your thinking toward the average; used as an automated dissenter, it forces your team to sharpen its edge before the market ever sees the piece.

Closing thoughts

In a business world dominated by risk mitigation and overly cautious brand management, herding and sameness are becoming almost guaranteed. The tension facing marketing leaders is clear: they can optimise their process for internal comfort and consensus, or they can optimise it for market distinction, but rarely both. The purpose of thought leadership is not to comfort the converted, it is to lead the market somewhere it has not yet dared to go.