Now the Business Is Next | Part 3

Individual adoption is the foundation. Organisational transformation is where the real value lives. But most companies get it the wrong way round.

Karl George MBE · Founder, The Governance Forum & Governance AI

In Part 1 of this series, I set out the evidence that a senior leader can reclaim 6.5 hours per week through AI adoption. In Part 2, I explored what to do with that recovered time and demonstrated how reinvested hours can generate £600,000 per year across a 50-person organisation, or a tenfold return for a supported housing provider.

Those first two parts were deliberately focused on the individual. Your time. Your setup. Your templates. Your productivity. And that was intentional, because the individual is where responsible AI adoption must start.

But it cannot end there. The business is next.

The failure rate nobody talks about

Before we rush to the organisational opportunity, we need to confront an uncomfortable truth. Most enterprise AI projects fail. Not some. Not a few. Most.

95%

of enterprise AI pilots deliver zero measurable P&L impact

MIT NANDA, State of AI in Business 2025

80%+

of AI projects fail overall, twice the rate of non-AI technology projects

RAND Corporation

42%

of companies abandoned most AI initiatives in 2025, up from 17% in 2024

S&P Global Market Intelligence

15%

of US employees say their workplace has communicated a clear AI strategy

Gallup, 2024

Read those numbers again. Companies are pouring $30 to $40 billion into generative AI, and 95% of those enterprise pilots are delivering no measurable return. The S&P Global survey found that 42% of companies abandoned most of their AI initiatives in 2025, a dramatic spike from just 17% the year before. And only 15% of employees report that their organisation has even communicated a clear AI strategy.

This is not a technology failure. The models work. GPT-4, Claude, Copilot: these tools are genuinely powerful. The failure is in implementation. It is an execution failure, a culture failure, and, frankly, a governance failure.

The technology is not the problem. The problem is that organisations try to transform the business before they have transformed the people. They skip the foundation and wonder why the building falls down.

Why the foundation matters

The MIT research reveals something that should not surprise anyone who works in governance: the strongest enterprise AI deployments began with individuals. They started with power users, people who had already experimented with tools like ChatGPT or Claude for personal productivity. These people understood what AI could and could not do. They became the internal champions, the bridge between executive ambition and operational reality.

MIT calls this the “prosumer” effect. Rather than relying on a centralised AI function to identify use cases from the top down, successful organisations allowed budget holders and domain managers to surface problems, test tools and lead rollouts from the bottom up. This bottom-up learning, paired with executive accountability, accelerated adoption while preserving operational fit.

This is precisely why Parts 1 and 2 of this series focused on the individual. If your leaders have not used AI themselves, if they have not experienced the 6.5 hours, if they have not built their own templates and workflows, they are not equipped to lead an organisational transformation. They will make decisions based on vendor demos rather than lived experience. They will approve strategies they do not understand. And they will join the 95%.

The pattern of failure: 67% of organisations that attempted and failed to implement an AI project cited insufficient skills as the main barrier. 33% said they did not fully understand the problem they were addressing. 44% underestimated the data expertise needed. These are not technology problems. They are readiness problems. (Sources: OneAdvanced Housing Trends Report 2025; MIT NANDA 2025)

The AI-first business: what it actually means

An “AI-first business” is not an organisation that has replaced its people with algorithms. It is an organisation where AI is embedded into how people think, work and make decisions, at every level, with proper governance and human oversight at every stage.

It means that before a team starts a new process, someone asks: “How could AI support this?” It means that when a board paper is being prepared, AI assists with the drafting, the data analysis and the horizon scanning, while the human provides the judgement, the context and the accountability. It means that when a new member of staff joins, they are trained not just on systems and policies but on how to use AI responsibly and effectively in their role.

Getting there requires six things, in this order. Skip a step, and you join the majority who fail.

AI literacy at leadership level

Board members and senior leaders must understand what AI can and cannot do, not as technologists but as governors. They need to be able to ask the right questions, assess risk, and make informed decisions about AI adoption. This is not optional; it is a governance requirement.

Maturity assessment

Before you invest, you need to know where you are. What is the current level of AI understanding across the board, the leadership team and the workforce? Where are the capability gaps? Where are the data gaps? Where are the governance gaps? You cannot build a strategy on assumptions.

Risk assessment

AI introduces new categories of risk: data protection, hallucination, bias, intellectual property, reputational exposure and regulatory compliance. These need to be identified, assessed and mitigated before you scale, not after something goes wrong. The organisations in the 95% failure group typically skipped this step.

Culture and individual adoption

AI transformation is a people transformation. If your staff are not confident, trained and supported, no amount of technology investment will deliver returns. The NTT DATA research on AI failure makes this point forcefully: the definition of adoption is getting people to work in a different way. That requires trust, training and time.

Governance framework and guardrails

Policies, acceptable use guidance, data classification, oversight structures, training programmes and accountability. These are not bureaucracy; they are the foundation that allows responsible scaling. Without them, you get shadow AI, data breaches and reputational damage.

Strategic use cases and workflow redesign

Only now, with literacy, maturity, risk, culture and governance in place, are you ready to identify and implement strategic AI use cases that redesign end-to-end workflows. McKinsey’s 2025 AI survey confirms it: organisations reporting significant financial returns are twice as likely to have redesigned workflows before selecting tools.

Notice the order. The technology comes last. Literacy, maturity, risk, culture and governance come first. This is the opposite of how most organisations approach it, which is why most organisations fail.

AI literacy: from the boardroom to the front line

The single most important predictor of whether an organisation will succeed with AI is the literacy of its leadership. Not their technical expertise. Their literacy. Can the board ask intelligent questions about AI risk? Can the CEO distinguish between vendor hype and genuine capability? Can department heads identify which of their team’s tasks are suitable for AI assistance and which are not?

The Gallup finding that only 15% of employees report a clear AI strategy from their employer is not primarily a strategy failure. It is a literacy failure. Leaders who do not understand AI cannot articulate a strategy for it. They cannot communicate expectations. They cannot set guardrails. And they cannot lead by example.

This is why AI literacy must cascade from the top. The board goes first. Then the executive team. Then department heads. Then the wider workforce. Each level needs a different depth of understanding, but everyone needs a baseline. A board member needs to understand enough to govern AI responsibly. A housing officer needs to understand enough to use it safely and effectively in their daily work. A finance director needs to understand enough to assess the investment case and the risk profile.

The organisations that succeed are not the ones with the best technology. They are the ones where leadership understood AI well enough to make good decisions about it, and then cascaded that understanding to every level of the organisation.

Maturity recognition: know where you are before you decide where to go

One of the most common mistakes I see is organisations investing in AI tools before they understand their own readiness. They buy platforms, commission pilots and hire consultants without first answering fundamental questions. How AI-literate is our board? Which departments have the data quality to support AI? Where are we already using AI informally, and with what governance? What risks are we already exposed to through shadow AI?

A maturity assessment answers these questions with evidence rather than assumption. It benchmarks where the organisation sits today, identifies the gaps, and produces a clear roadmap for building capability. Without it, you are investing blind.

The social housing sector illustrates this vividly. The Leeds University research found that 31% of staff are already using AI, but only 22% are aware of any organisational provision. Less than 25% of staff feel confident using AI, and 30% rate their confidence as poor or very poor. And 44% of housing associations have no AI policy at all. These are maturity gaps that no amount of technology procurement will fix. You have to diagnose before you prescribe.

Risk assessment: the step most organisations skip

AI risk is not hypothetical. It is real, measurable, and growing. The 2025 research landscape makes this clear: 77% of businesses express concern about AI hallucinations, and 47% of enterprise AI users made at least one major decision based on hallucinated content in 2024. Shadow AI is now structural, with 78% of AI users bringing their own tools to work. And the regulatory environment is tightening, with the EU AI Act creating binding requirements and UK regulators increasingly focused on AI governance.

For regulated sectors like social housing, the risk calculus is particularly sharp. An organisation under regulatory scrutiny that adopts AI without a proper risk assessment, without data classification, without policies on what can and cannot be shared with AI tools, is not being innovative. It is being reckless.

A proper risk assessment does not prevent AI adoption. It enables it. By identifying and mitigating risks upfront, you create the confidence to move faster, not slower. The red/amber/green data classification framework I outlined in Part 1’s companion report is a practical example: it tells staff exactly what they can do, on which platforms, with which data, so they can act with confidence rather than hesitation.

Culture eats strategy for breakfast. And AI for lunch.

Peter Drucker’s famous observation applies with particular force to AI adoption. You can have the best AI strategy in the sector. If your culture does not support it, it will fail. NTT DATA’s research on AI implementation failure makes a point that deserves far more attention: why are so few AI specialists talking about the people?

The answer, I think, is that technology feels more tractable than culture. You can buy a platform in a week. Changing how 50 or 500 people think about their work takes months. But the evidence is unambiguous: organisations that invest 70% of their AI resources in people and processes (not just technology) are the ones that succeed. The technology is the easy part. The culture is the hard part. And the culture is what separates the 5% from the 95%.

In practice, culture change for AI means three things. First, psychological safety: staff need to feel comfortable experimenting, making mistakes and asking for help without fear of judgement. Second, visible leadership: if the CEO is not using AI themselves, the organisation receives a clear signal that this is not a real priority. Third, practical support: training, templates, prompt libraries, a named AI champion, and regular opportunities to share what is working. Culture is not built through memos. It is built through daily practice.

The Governance AI Journey: designed for exactly this sequence

I designed The Governance AI Journey specifically because I watched organisations getting this wrong. They were buying tools before building literacy. Scaling before assessing maturity. Adopting before governing. And failing, at rates entirely consistent with the MIT and RAND research.

The seven-module programme follows the sequence that the evidence says works.

Phase One: Foundation

Build literacy, assess readiness, lay governance groundwork

Module 1: AI Wake-Up Call Workshop builds shared AI literacy at board and senior leadership level. Directors complete a literacy review, the board conducts a maturity assessment, and a gamified risk workshop produces the first version of an AI risk register. This is where leadership stops relying on vendor presentations and starts understanding AI from direct experience.

Module 2: GovernIQ™ Diagnostic benchmarks AI readiness across three layers: individual director capability, departmental maturity and organisational structures. This is the maturity assessment that replaces assumption with evidence.

Module 3: Departmental Use Case Taster pilots a live AI use case in one department, generating early wins while the AI steering committee practises its oversight role. This is where culture change begins, with a real, supervised experiment.

Module 4: Governance AI Playbook co-creates the governance framework: AI usage policy, oversight committee terms of reference, and alignment with global standards including ISO 42001 and the EU AI Act. This is the guardrail step that enables responsible scaling.

Phase Two: Development

Deepen capability, secure infrastructure, unlock strategic use cases

Module 5: Thematic Workshops & Implementation Discovery brings leadership and operational teams together to identify high-value AI use cases through hackathons and idea sprints. This is where the organisation moves from experimentation to strategic implementation.

Module 6: Data & Security Diagnostic assesses the data infrastructure and cyber resilience needed for responsible AI deployment. This is the step that ensures the invisible infrastructure around AI, the bit that MIT says most organisations get wrong, is solid before you scale.

Phase Three: Accreditation

Demonstrate maturity, gain external recognition

Module 7: Governance AI Quality Mark prepares the organisation for formal, independent recognition of its AI governance maturity. This validates responsible oversight, strengthens trust with stakeholders and regulators, and positions the organisation as a leader in governing AI ethically.

Notice the architecture. Literacy first. Then maturity assessment. Then a supervised pilot. Then governance. Then strategic use cases. Then data and security. Then external validation. Every step builds on the one before. Every step addresses one of the reasons the research says organisations fail.

The AI-first business is not a destination. It is a discipline.

Becoming an AI-first organisation is not something you achieve on a particular date and then move on. It is a continuous discipline of learning, adapting, governing and improving. The technology will keep evolving. Agentic AI, which we explored in Part 1, will change the nature of what AI can do. New regulations will change what AI must do. New risks will emerge. New opportunities will appear.

The organisations that thrive will be the ones that built the muscle in the right order: literacy, maturity, risk, culture, governance, and then technology. They will have leaders who understand AI well enough to govern it. Staff who are confident enough to use it. Frameworks robust enough to manage it. And a culture curious enough to keep evolving with it.

Irrelevance does not come from AI advancing. It comes when we stop evolving with it. The AI-first business is one that never stops learning, never stops governing, and never stops asking: how can we do this better?

Where this series leaves you

Across these three parts, we have covered the full journey from individual to organisation.

Part 1 gave you the evidence: 6.5 hours per week recoverable through personal AI adoption, with a research-backed breakdown by task and practical guidance on GPTs, Skills and the agentic AI that is coming next.

Part 2 showed you what to do with the time: reinvest half into wellbeing, half into higher-value work, and watch the returns compound, from £600,000 per year in amplified output to a tenfold return for a housing provider.

Part 3 has made the case that the business is next, but only if you get the sequence right. Culture before technology. People before platforms. Literacy before strategy. Governance before scaling. The 95% failure rate is not inevitable. It is the consequence of skipping steps.

The Governance AI Journey exists because these steps need to be taken in order, with expert guidance, proper governance and measurable outcomes. It is not a technology programme. It is a leadership programme. And it starts with the people at the top.

Ready to take the business from individual adoption to organisational transformation?

The Governance AI Journey is a seven-module programme that takes boards and senior leaders from AI literacy through to accreditation. Phase One begins with the AI Wake-Up Call Workshop and the GovernIQ™ Diagnostic.

Contact Karl George MBE at The Governance Forum

Part 1

Can AI Really Save You a Day a Week? The Evidence Says Yes.

Part 2

You’ve Found the Time. Now What?

Part 3

Now the Business Is Next.