To AI or Not to AI
Why the real bottleneck on enterprise AI is leadership, not the workforce
The problem has less to do with the workforce adopting AI. It has far more to do with unfair expectations and siloed communication from leaders and stakeholders.
That claim has data behind it now. MIT’s 2025 "State of AI in Business" report found that 95% of enterprise generative AI pilots deliver no measurable P&L impact - and the reason isn't the models. The same report found a thriving “shadow AI economy” inside these companies: roughly 90% of employees are already using personal AI tools at work, while only 40% have been given official, sanctioned access. Employees are adopting AI just fine on their own time. It's leadership that hasn't caught up.
So here's the real question: how many enterprise leaders - even in the tech industry - are willing to set their pride aside and actually learn AI? Isn't that where the problem starts?
Let me explain. Leaders talk a lot about how "AI-native" their organization already is. Meanwhile, Harvard Business Review has argued that employees won't trust AI if they don't trust their leaders in the first place - and trust isn't built by a slide claiming AI-native status. It's built by what leadership actually does next.
Communication is not optional
In large enterprises, communicating all the way down to the last person - and giving people the courage and conviction that this AI wave will not be a tsunami, that this tide will also ebb and flow - isn't just a leadership responsibility. It's a prerogative.
Leaders are human too. They're also vulnerable. No one can predict exactly where this is going, or where humanity ends up as a result. It's okay to say that to your teams. Be honest. Be open. Lead the way by adapting, and by unlearning and relearning in public. Futurist Alvin Toffler was gesturing at exactly this, decades before any of us were talking about GenAI: in Future Shock, he argued that the literacy that would matter most going forward wasn't reading and writing, but the capacity to keep learning, discard what no longer applies, and learn again.
We've seen this movie before
When the Metaverse was at the peak of its hype cycle, I remember calling out one point of view clearly: there would be an ebb, once people realized they didn't actually need a metaverse yet - not in the consumer industry, and not in the enterprise. Heavy industries would benefit far more from those technologies than everyday business would.
That call held up. Meta's Reality Labs division alone has burned through close to $80 billion in cumulative losses since the 2020 rebrand bet the company's future on it - while the consumer and mainstream-enterprise use cases the hype promised largely failed to show up.
Now it's AI's turn. Traditional AI will stay and rule. Machine learning algorithms will keep thriving quietly in the background, the way they always have. GenAI and agentic AI have real, magic-like potential - but do enterprises actually need them, right now, for this use case? Leaders have to answer that question honestly, not rhetorically.
The responsibility leaders keep skipping
Taking their people along, investing in their knowledge curve, and ensuring their careers and livelihoods are secured - that's part of the job of leading through this, not an optional extra.
Before we ask AI systems for "Explainable AI" - a real research agenda, launched by DARPA back in 2016 to make black-box models legible to the humans depending on them - let's ask leaders to do the same thing for their own decisions. Let them explain why they need AI, what they're actually planning to achieve, and what the end goal of the business even is. A leader who can answer all of that clearly can't be knocked over by an AI tsunami. There won't be one to knock them over with.
What I see walking into these organizations
As a consultant, I get to see this from the outside - and what I witness in enterprise after enterprise is leaders drawing up KRAs and KPIs around “AI adoption” without pausing to understand what that adoption actually means: to their shareholders, to their customers, and to their teams. This isn't a Java rollout or a CRM implementation. AI is deep, and it's layered, in a way most enterprise technology decisions of the last two decades simply weren't.
One size will not fit all. Before you flatten the enterprise pyramid into a rhombus or a diamond, make a real, circular connection with everyone in the organization first - and take an informed, collective call.
The real question
To AI or not to AI is not the question. The real question is: who benefits from implementing AI? And why do we need that benefit?
Sources
Alvin Toffler, Future Shock (1970), on learning, unlearning, and relearning
Comments