The headlines have been hard to miss: according to a new report from MIT’s NANDA initiative ‘The GenAI Divide: State of AI in Business 2025′, only about 5% of generative AI pilots deliver rapid revenue growth while the remaining 95% stall.
It’s a startling statistic, and one that has spooked investors and sent many marketing leaders scrambling. Yet behind the clickbait lies a deeper truth: AI isn’t failing us – it’s our approach to AI that is failing.
As communications professionals and board advisers, we’ve spent the last three years training, testing, measuring and learning how generative AI will work effectively and ethically; not just for our business, but for our clients.
What we’ve learned is that investing well and owning the process works. What the MIT findings highlight backs up our approach: technology alone doesn’t create value; strategy, governance and culture do.
What does the MIT report say?
The GenAI Divide: State of AI in Business 2025 surveyed 300 public AI deployments, interviewed 150 leaders and polled 350 employees. The team found that, despite a rush to adopt generative tools, only a handful of pilots deliver measurable profit. Several insights stand out:
- Fewer than one in ten pilots succeed. Roughly five percent of organisations in the dataset achieve rapid revenue acceleration, while the vast majority deliver little to no impact on the P&L.
- The issue isn’t model quality but enterprise integration. Generic tools like ChatGPT perform well for individuals, yet they stall in companies because they don’t learn from or adapt to internal workflows.
- Budgets are misallocated. More than half of generative AI spending goes to sales and marketing, yet MIT researchers found the biggest return in back‑office automation, from cutting outsourcing costs to streamlining operations.
- Partnerships outperform DIY builds. Purchasing specialised AI tools and partnering with vendors succeeds about 67% of the time, whereas internal builds succeed only one‑third as often.
These findings expose a simple reality: AI pilots fail when organisations treat them as experiments instead of strategic transformations. When adoption is driven by hype rather than a clear business case, and when leaders expect technology to “just work” without investing in people, process and governance, disappointment is inevitable.
Meanwhile, technology and opportunities are evolving at pace.
The real reasons AI pilots fail
We see three recurring themes in failed AI projects:
- Lack of purpose: Many pilots are launched to “see what happens” rather than to solve a defined problem. MIT found that successful start‑ups pick one pain point, execute well and partner smartly. By contrast, unfocused pilots dissipate energy and yield vague insights.
- Fragmented governance: AI is often delegated to specialist teams or innovation labs. Yet the report shows that empowering line managers to drive adoption and integrating AI into existing workflows are critical. Without board‑level oversight and cross‑functional collaboration, tools don’t evolve with the organisation’s needs.
- Skill and culture gaps: The “learning gap” identified by MIT is two‑fold: AI tools struggle to adapt to enterprise data, and employees lack the training to use them effectively. When people see AI as a threat or a black box, they resist adoption and starve pilots of the feedback loops they need to learn. AI is about change management first and technology execution second.
Doing nothing is the biggest risk
Some leaders respond to these findings by stepping back. But inaction is itself a risk. Our own research shows that around 45% of boards still don’t have AI on the agenda, and one in four UK businesses admits to having no formal AI plan. Meanwhile, early adopters are already seeing significant returns. AI isn’t going away; the companies that hesitate now will find themselves playing catch‑up later.
At TWI, we encourage boards to see AI as a governance challenge rather than a technology trial. That means putting AI on the risk register, engaging with regulators and aligning innovation with your organisation’s values and strategy. If AI isn’t on your risk register, your Board isn’t looking far enough ahead.
Building a strategic approach
How can organisations avoid becoming part of the 95%? Start by reframing AI adoption as a long‑term programme, not a pilot. In our AI Strategy Consultancy for Boards, we structure discussions around six elements:
- Managing risk: Embed AI into your risk register and governance processes, ensuring ethical, legal and reputational considerations are addressed.
- Creating an AI roadmap: Define clear business objectives and prioritise use cases that align with your purpose and deliver measurable value.
- Investing in the right tools: Avoid the hype and choose technologies that integrate with your workflows; vendor partnerships often deliver better results than bespoke builds. I still find it hard to believe that businesses are approaching such a seismic shift without clear investment.
- Developing skills and training: The skills gap will be huge. Equip your teams with the knowledge to work alongside AI and foster a culture of experimentation and feedback.
- Appointing AI champions: Identify internal advocates who can bridge technical and operational perspectives.
- Measuring ROI: Take a step back and establish metrics from the outset to track value, learning and impact. Embed a series of metrics, from productivity to quality, and even softer qualitative evidence – are your people spending less time on work they don’t enjoy?
These steps echo the MIT report’s call to integrate AI into core operations and to focus investment where the returns are highest, such as back‑office automation. They also align with our belief that trust, leadership and transparency are just as important as code and data.
Where to begin again
The MIT report should serve as a wake‑up call, not a deterrent. Yes, most AI pilots are failing, but they’re failing because organisations jump in without a plan. With the right strategy, governance and culture, AI can be transformational. We’ve seen clients streamline grant writing, improve knowledge sharing and generate policy drafts using generative tools. We’ve also seen how a clear roadmap and ethical framework build confidence in teams and boards, and honestly just make people’s lives better.
The future belongs to organisations that make AI work for them, not those who wait for a perfect model. If you’re a business leader or board member looking to translate AI hype into real value, I’d love to share what we’ve learned. Let’s turn the 95% failure rate into a 95% success story.
We’re here to help
Find out more about our AI Consultancy for Boards and Leadership Teams. We have also partnered with experts who can help you on your wider AI journey, so let’s talk AI.