We’re being sold two competing futures with artificial intelligence. Two sides of the same coin. In one, AI is the utopian cure-all: productivity skyrockets, profits follow, decisions sharpen, work becomes lighter, and organisations finally operate like the sleek, integrated machines they always aspired to be. In the other, AI is the villain: jobs vanish, inequality deepens, wealth concentrates in the hands of a small and powerful few, and the machines quietly displace the working middle class. Both futures dominate the narrative. Both are emotionally charged. And both skip over the most inconvenient truth: neither future is possible unless organisations do the hard, unglamorous work of getting ready, starting with their people.
What AI adoption actually means in practice
Let’s be direct about something the vendor demos won’t tell you: AI adoption is not a technology project. It is a human project, wrapped in technology. In practice, AI adoption means a finance analyst’s weekly reporting cycle gets compressed from two days to two hours but only if that analyst understands how to interrogate the output, validate its assumptions, and apply judgement to what the model can’t see.
It means a customer service team handles more complex queries because AI is triaging the routine ones, but only if that team has been retrained in exception handling, empathy at the edge case, and knowing when to override the machine.
It means a procurement manager uses AI to surface contract anomalies, but only if the underlying data is clean, classified, and governed well enough for the model to trust.
This is what AI adoption actually looks like on the ground. Not a grand transformation moment, but a steady, sometimes uncomfortable shift in how people do their jobs, make decisions, and understand their own value. The technology is rarely the bottleneck; it’s the human system around it. The mess is us. Data is siloed, duplicated, or outdated. Metadata is an afterthought. Knowledge lives in people’s heads and is hoarded because in many workplaces, knowledge still equals power. AI doesn’t resolve that dynamic, it amplifies it. Drop a powerful AI model into a fragmented organisation and you don’t get transformation. You get faster fragmentation.
Building capability: preparing your workforce to adapt
The utopian promise requires a workforce that can work with AI, not just alongside it. The dystopian risk accelerates when organisations treat AI as a replacement strategy rather than a capability-building opportunity. The difference between those two futures is largely determined by one decision: whether you invest in your people before, during, and after implementation. Workforce capability for AI is not about making everyone a data scientist. It is about three distinct layers of literacy.
Foundational AI literacy means every employee understands at a basic level what AI can and cannot do, how outputs should be interpreted, and what responsible use looks like in their role. This is not optional. An organisation where only the technology team understands AI is an organisation where AI will be misused, over-trusted, or quietly ignored by everyone else.
Role-specific capability means identifying, honestly and specifically, how each function will change, what new skills are required, and where current skills transfer. A data-informed approach to workforce planning maps current capabilities against what AI-augmented roles demand, identifies the gap, and closes it through targeted learning pathways, not generic e-learning modules.
Adaptive capability is the hardest and most overlooked layer. It is the ability of your workforce to keep learning as the technology evolves to update their mental models, question their assumptions, and stay effective as the landscape shifts underneath them. Organisations that build adaptive capability invest in psychological safety, encourage experimentation, and treat mistakes as data rather than failures.
This last layer is where traditional training programs fall short. You cannot train adaptive capability into people through a course. You build it through culture.
Building digital resilience
Building digital resilience through AI means treating AI adoption as a continuous change journey, not a go-live event. Start with honest diagnosis. Before designing any change intervention, understand the current state with precision. Where is resistance likely to come from, and why? Which teams are early adopters who will pull others forward? Where does AI feel threatening, and is that fear based on misunderstanding or on a legitimate reading of the organisation’s intentions? Change that isn’t grounded in honest diagnosis will miss its target.
Engage, don’t announce. The instinct in most organisations is to communicate AI adoption top-down: strategy presentations, town halls, intranet articles. This creates awareness, not engagement. Genuine engagement means involving people in designing the change by asking frontline teams where AI would help, piloting with people who will use the tools, and iterating based on what you learn. People support what they help build.
Name the losses. Every significant change involves loss. It can be the loss of familiar processes, of established expertise, of role identities that people have built careers around. Change that only talks about the gains will be experienced as dishonest, naming what is genuinely changing, and creating space for people to process that, is not weakness. It is the precondition for moving forward.
Build change capability, not just change compliance. The goal is not a workforce that tolerates AI because they were told to. It is a workforce that actively builds and improves AI-enabled ways of working because they understand why it matters and have the skills to do it well. That requires sustained investment in leaders who model adaptive behaviour, managers who coach rather than direct, and structures that reward learning over performing certainty.
Measure adoption, not deployment. An AI tool deployed is not an AI tool adopted. Digital resilience is measured by whether people are using AI effectively in their work, whether quality and outcomes are improving, and whether the organisation’s capacity to absorb the next wave of change is stronger than before. These are the metrics that matter.
The real transformation
The utopian version of AI is possible. So is the dystopian one. The distance between them is not determined by the sophistication of the technology. It is determined by whether organisations are willing to invest, genuinely and consistently, in the human system that technology depends on.
That means building capability before demanding performance. It means designing change with people, not at them. It means treating digital resilience as an organisational muscle that gets stronger with use, not a project milestone to be checked off.
AI isn’t the transformation, we are. And the organisations that will realise the promise, not just talk about it, are the ones building that capability now, one honest conversation, one well-designed learning pathway, and one courageous change decision at a time.
Are you building AI capability or chasing AI hype?
AI accountability is the best strategy
There is a familiar pattern playing out in boardrooms and executive suites right now across the business landscape. A CEO or senior Executive presents an impressive AI strategy. It is ambitious, well-designed, and full of the right language and projected figures and graphs, transformation, efficiency gains, productivity increases, competitive advantage, future-readiness. Questions are asked about investment size and expected return on investment, completion dates. The board nods, the initiative is approved.
Six months later, the pilot hasn’t scaled. Twelve months later, the business case is being quietly revised. Eighteen months later, a new vendor is being evaluated.
The technology wasn’t the problem, it never is, and yet the questions that would have surfaced the real barriers, about data, about people, about governance, about organisational readiness, were never asked in the room where it mattered most.
This is a governance failure, at the highest level. Not a technology failure. Not a management failure.
The accountability gap at the top
Boards are accustomed to holding executives accountable for financial performance, regulatory compliance, and strategic delivery. AI readiness requires the same rigour and is not yet receiving it.
Part of the problem is literacy. Many board members and senior executives came of age in an era where technology was a support function, not a strategic differentiator. AI feels technical, which makes it easy to delegate entirely to the CTO or CDO and consider the matter handled. This is a category error. AI is not an IT initiative. It is a business transformation with technology as the enabler. Delegating it entirely to the technology function is like delegating your customer strategy entirely to the marketing department. Oversight without understanding is not governance.
Part of the problem is incentive. Boards and executives are rewarded for decisive action and forward momentum. Asking hard questions about data quality, change management maturity, and workforce capability doesn’t generate a press release. It generates homework. In a culture that rewards announcements, the unglamorous readiness work is systematically underinvested and under-scrutinised.
And part of the problem is simply that nobody has told boards and executives clearly enough what questions they should be asking and why asking them is their job. This article can start the right conversation.
Boards set the conditions for how organisations approach risk, investment, and long-term capability. On AI, there are four things every board should be actively promoting, not just permitting.
Honest readiness assessment over headline strategy. Every organisation should have a clear, evidence-based view of its current AI readiness across data, people, governance, and technology. Not a vendor-commissioned maturity assessment that produces a flattering benchmark. An honest internal audit that surfaces the real gaps. Boards should be asking whether that assessment exists, who commissioned it, and whether the findings informed the AI investment plan, or were quietly set aside because they were inconvenient.
Long-term capability investment over short-term use cases. The pressure to show AI ROI quickly is understandable and, in many cases, entirely appropriate. But it creates a systematic bias toward narrow, visible use cases at the expense of the foundational investments, data governance, workforce capability, change management infrastructure, that determine whether AI can scale across the organisation. Boards need to protect and actively promote the unglamorous investment, not just the headlines. If the only AI spend being approved is tool procurement, the strategy will fail.
A culture of responsible experimentation. The dystopian risk of AI is not primarily a technology risk. It is a governance and culture risk, the risk that AI is deployed without adequate oversight, that errors are not surfaced and corrected, that the organisation optimises for speed at the expense of safety. Boards should actively signal that responsible AI practice is a non-negotiable expectation, not a constraint to be managed around. That signal must come from the top, and it has to be consistent.
Workforce investment as a strategic priority. The boards that will look back proudly on their AI leadership are the ones that treated workforce capability, not just headcount efficiency, as the central AI question. How are we building the skills, the adaptability, and the psychological safety that allow our people to work effectively with AI over time? This is a board-level question. It belongs on the agenda alongside capital allocation and risk appetite.
Prompting sets the conditions; inquiry creates the accountability.
Here are the questions every board should be directing at their CEO and executive team and should not accept a vague answer to.
What problem are we actually solving with AI, and how do we know AI is the right solution?
The correct answer is specific. A genuine answer names a business problem, explains why alternative approaches were considered and rejected, and articulates what success looks like in measurable terms. An answer that begins with “AI will help us remain competitive in a rapidly evolving landscape” is not an answer. It is a deflection.
What is our theory of change, how exactly does AI investment translate to business value?
This question separates organisations with a genuine AI strategy from those chasing the technology because their competitors appear to be. The logic chain from investment to outcome should be explicit, testable, and regularly reviewed.
Has the organisation conducted an honest audit of its data quality, and what did it find?
If the answer is no, the board should require one before approving significant AI investment. If the answer is yes, the board should ask for the findings, including the uncomfortable ones, and ask how the remediation plan is being resourced and governed.
Who is accountable for data quality, and what authority do they have to enforce standards?
Accountability without authority is theatre. If the honest answer is that data quality sits with the IT team but the business units own the data and aren’t being held to account, that is a governance gap that will undermine every AI initiative the organisation attempts.
How confident are we in the data feeding our AI systems, and what is our exposure if that data is wrong?
AI systems will generate outputs with confidence regardless of the quality of the inputs. Boards need to understand what the organisation’s risk exposure is when AI operates on poor data, and whether there are adequate human review processes and escalation pathways in place.
What is our workforce capability plan, and how does it address both current and future skill requirements?
A credible answer describes specific capability gaps identified through analysis, targeted investment in closing those gaps, and a way of measuring whether the investment is working. A generic reference to “upskilling programmes” or “our learning and development team is working on it” should prompt further scrutiny.
How are we managing the human impact of AI-driven change, and who owns that?
Organisational change management is consistently the most underinvested element of AI programmes and the most frequently cited reason for adoption failure. Boards should ask who specifically owns the change management function for AI initiatives, what their mandate is, and what resources they have.
Are we building a culture where people feel safe raising concerns about AI outputs?
The most dangerous AI environments are those where people feel pressure to trust the model and reluctant to flag errors. Psychological safety on AI is not soft; it is a critical risk management mechanism. Boards should be asking how the organisation is measuring and building it.
Do we have a clear AI ethics and responsible use framework, and is it being applied?
Policies are not governance. The question is whether the framework is operationalised, built into approval processes, procurement decisions, and deployment practices, and whether there is a meaningful accountability structure for ensuring it is followed.
What AI decisions are currently being made without adequate human oversight, and do we know where those are?
Most organisations do not have a complete inventory of their AI-assisted decision-making. This is a material governance risk. Boards should be asking for that inventory and for clarity on which decisions carry the highest risk if the AI is wrong.
What are our biggest AI-related risks over the next three years, and how are we managing them?
This question should produce a substantive risk register, not a generic reference to the AI regulatory environment. The risks should include capability risk, data risk, workforce risk, adoption risk, and reputational risk, not just the cybersecurity and compliance risks that technology teams typically surface first.
How are we measuring AI adoption, and are we measuring the right things?
Deployment is not adoption. The number of AI tools procured is not a meaningful metric. Boards should be asking for evidence of actual behaviour change: whether people are using AI effectively in their work, whether quality and efficiency outcomes are improving, and whether the organisation’s capacity to absorb change is growing.
What have we learned from our AI pilots that has changed our approach?
The right answer demonstrates that the organisation is genuinely learning and iterating, not defending its original assumptions. An executive team that cannot articulate what the pilots taught them, including what went wrong, is not running a learning programme. It is running a performance.
The Board’s role is not to have the answers. Let’s be clear about what this is not asking. It is not asking board members to become AI experts. It is not asking them to second-guess technical decisions or micromanage implementation. That is not governance; that is interference.
What it is asking is that boards hold executives accountable for the quality of their thinking, the honesty of their assessment, and the rigour of their planning. That they ask the questions that create the conditions for genuine AI capability, not just AI activity. That they treat AI readiness with the same seriousness they treat financial controls, risk management, and regulatory compliance. Because the organisations that will realise the genuine value of AI, the utopian possibility that is real and achievable, are not the ones that moved fastest. They are the ones where the people at the top demanded honest answers, protected foundational investment, and understood that technology without governance is not an asset. It is a liability waiting to surface.
The questions are not complicated. The discipline to ask them, consistently and without accepting a polished non-answer, is harder. That is the job of a board, and it always has been.
