A recent survey done by MIT found that despite $30–40 billion in investments, 95% of generative AI pilots deliver no measurable P&L impact, with only 5% achieving multi-million-dollar value.
The primary reason that they found was that AI systems failed to adapt to workflows, or evolve over time, unlike flexible consumer tools. They also found that there exists a “shadow AI economy”, which thrives as 90% of workers use tools like ChatGPT for personal tasks, bypassing stalled enterprise efforts.
So, the question we need to ask is the following –
While employees are using AI tools in their personal lives, why are they not adopting the same tools when deployed by their organizations?
And what can we do as leaders to help them adopt these tools at scale in our organizations?
Here is where I believe that there is a lot that we can learn from a different paradigm – from the field of memetics.
In 1976, Richard Dawkins proposed one of the most provocative ideas in modern science. In his book, The Selfish Gene, Dawkins coined the term “meme” and defined it as a unit of cultural transmission that replicates, mutates, and competes for survival, much like a gene does in a biological ecosystem.
Memetics is the study of how memes (ideas) spread through populations.
And when we look at Enterprise AI adoption challenge from this point of view, we can see that this is not a technology deployment problem but is an epidemiological one.
Epidemiology examines the distribution (who, where, when) and determinants (risk factors, causes) of spread of diseases (or in our case, ideas or adoption) in a population. We can apply the learning from this field to the spread of ideas and adoption of technology as well.
What this means is to pivot our frame of reference from – “How do we implement and scale AI” to – “How do we make the idea of AI usage so compelling, so easy, and such that it spreads on its own?”
Once we see adoption through this frame, several patterns come into sharp focus.
The Three Conditions for Spread
Susan Blackmore, in her book, The Meme Machine, identified three properties that determine whether a meme thrives or dies: fidelity, fecundity, and longevity.
Fidelity:
The teams that adopt AI successfully are not the ones with the best technology. They are the ones with the clearest, simplest and the most replicable use case.
“We use AI to draft the first version of every customer email” copies intact across departments. “Integrate AI everywhere” mutates into confusion before it reaches the second floor.
Fecundity:
In the context of memes, fecundity refers to a meme’s reproductive potential—specifically, how quickly and prolifically it spreads or replicates from person to person.
So, the question we need to ask is how easy and simple is it for our employees to try and use AI in their workflows.
This explains why “shadow AI”, employees quietly using ChatGPT on their personal devices (simpler and easier) — outpaces formal enterprise rollouts (long drawn, confusing boundary conditions, susceptible to power dynamics).
Longevity:
AI initiatives that stick are those embedded into existing workflows, not bolted on as separate tools.
Just as memes survive by attaching to established cultural practices, AI tools persist when they become part of how work already gets done, attaching to existing cultural contexts, behaviors and organizational habits.
Memeplexes Beat Isolated Memes
Individual memes are fragile. But memes that travel in clusters, what Dawkins called memeplexes, are remarkably resilient. Religions, political ideologies, and scientific paradigms all persist not because any single idea within them is unassailable, but because the ideas reinforce each other. Remove one, and the others pull it back.
The parallel to enterprise technolgoy adoption is striking. AI adoption fails when it is pitched as a single tool or a single project. It succeeds when it arrives as a mutually reinforcing bundle: new practices, new incentives, new narratives, and crucially, a new identity, all rooted in and connected to existing practices, expectations and outcomes.
“We are an AI-first team” is a memeplex. “We need to start using AI” is an isolated meme. One survives organizational antibodies. The other does not.
The Host Matters More Than the Meme
Memetics teaches us that susceptibility varies. The same meme lands differently depending on the beliefs, needs, and social context of the host. This explains one of the most frustrating patterns in enterprise AI: the same pilot, with the same technology, succeeds brilliantly in one business unit and fails completely in another.
We keep diagnosing this as a technology problem or a training problem. It is neither. It is a host-environment problem. Team culture, leadership, openness to experimentation, psychological safety around failure, existing skill levels — these are the variables that determine whether the AI meme takes root. The technology is the seed. The culture is the soil.
This has a practical implication that most transformation leaders miss – instead of rolling AI out uniformly, we need to identify the most receptive hosts first and enable them to become super-spreaders. The meme will spread outward from there far more effectively than any top-down mandate.
The Immune System You Are Fighting
Organisms develop immune systems against parasitic memes. Enterprises do too — they just call them different names. Middle management skepticism. IT security gatekeeping. Learned helplessness from the last three failed transformation programs. The collective memory of the CRM rollout that ate eighteen months and delivered nothing.
These are not irrational responses. They are adaptive defenses, honed by experience. And the insight from memetics is that you do not fight an immune system head-on.
You work with it and around it. Frame AI adoption in terms that the existing immune system recognizes as safe. Connect it to goals the organization already values, behaviors that is already recognized and rewarded. Present it not as disruption, which triggers every antibody in the system, but as amplification of what people already do well, as a multiplier.
Prestige Bias and the Carrier Problem
Henrich and Boyd’s research in cultural evolution reveals two powerful transmission biases. Prestige bias – we preferentially copy high-status individuals and Conformist bias – we copy the majority. Both have direct implications for AI adoption.
Prestige bias means that AI adoption accelerates dramatically when senior leaders visibly use AI themselves — not just endorse it from a stage, but use it in meetings, in emails, in decisions that others can observe. The meme needs visible, high-status carriers. An executive who says “I asked Claude to help me think through this decision” does more for adoption than a hundred training sessions.
Conformist bias means there is a tipping point. Before critical mass, adoption feels like swimming upstream. After it, adoption becomes the default. The strategic implication: concentrate early efforts on reaching that tipping point in a contained population, rather than spreading resources thinly across the entire organization.
Let the Meme Mutate
Here is where most enterprise leaders get it exactly wrong. They want pristine replication. Best practices. Standardized playbooks. Uniform adoption.
But memetics tells us that mutation is not a bug. It is how ideas adapt to local environments. The teams that personalize the AI tool in creative, unexpected ways are often the ones discovering its highest-value applications.
Controlled mutation, giving teams the toolkit and the guardrails, then letting them adapt usage to their context and expectations, produces stronger, more resilient adoption than any centralized mandate.
Conclusion:
The organizations winning at AI adoption are not the ones with the biggest budgets or the most sophisticated models.
Organizations do not change because new technologies are introduced. They change because new behaviors are created, spread and sustained over time.
