Board-level AI conversations are happening in every enterprise. Most of them are producing the wrong outcomes, not because directors lack intelligence, but because the questions being asked are fundamentally misdirected.
What You Need to Know
- Most boards are asking "What's our AI strategy?" when they should be asking "What are our highest-value knowledge problems, and which ones can AI solve?"
- The five most common board mistakes: focusing on models instead of outcomes, treating AI as IT's problem, expecting ROI before investing, demanding certainty in an uncertain market, and confusing AI activity with AI value.
- Boards that frame AI as a business transformation initiative (not a technology initiative) get better outcomes. AI investment should sit alongside digital strategy, not underneath IT.
- The most useful board metric isn't "how many AI tools have we deployed." It's "how much faster and cheaper is each new AI capability compared to the last?" That measures compound value.
- Boards should be asking about governance as seriously as they ask about ROI. The EU AI Act (approved May 2024) makes this a regulatory requirement, not just a best practice.
92%
of companies plan to increase AI investment, but most can't estimate AI's value
Source: McKinsey & Company, The State of AI in 2024, May 2024
The Five Mistakes
Mistake 1: "Which AI model should we use?"
This is like asking "which database vendor?" when you haven't defined your application. The model is the least differentiated part of the enterprise AI stack. GPT-4o, Claude 3, Gemini. They're increasingly commoditised and interchangeable.
Better question: "What are the three most expensive knowledge bottlenecks in our business, and do we have the data to address them?"
Mistake 2: Treating AI as an IT problem
AI is not infrastructure. It's a business capability that changes how work gets done, how decisions get made, and how value gets created. When AI sits solely in IT, it gets scoped as a technology project instead of a business redesign initiative.
Better question: "Who in operations is sponsoring our AI capabilities, and are they accountable for business outcomes?"
Mistake 3: Demanding ROI before investment
"Show me the ROI before we invest" creates a chicken-and-egg problem. AI ROI depends on data quality, integration depth, and adoption, none of which exist before you invest. You need an AI discovery sprint to develop informed projections.
Better question: "What's the minimum investment needed to develop a credible business case for our top AI opportunity?"
Mistake 4: Confusing activity with value
"We have five AI initiatives running" sounds impressive. But if none of them have reached production, none are integrated into workflows, and none are measuring business outcomes, you have activity, not value.
Better question: "How many AI capabilities are in production, integrated into daily operations, and measuring business outcomes?"
Mistake 5: Ignoring governance
Boards spend 90% of AI discussion time on opportunity and 10% on risk. Post-EU AI Act, this ratio needs to invert, or at least balance. AI governance isn't a constraint on value; it's the framework that enables responsible value creation at scale.
Better question: "Do we have a governance framework that allows AI deployment without ad-hoc risk assessment for every initiative?"
The boards getting the best AI outcomes aren't the ones with the most technical knowledge. They're the ones asking the right business questions, and holding management accountable for business answers, not technology answers.
Isaac Rolfe
Managing Director
- Should we have AI expertise on our board?
- Helpful but not essential. What you need is board members who understand digital business transformation and can ask the right questions. An AI-specific board advisory role can provide technical perspective when needed, without requiring a permanent board seat.
- How should boards measure AI progress?
- Three metrics: (1) Number of AI capabilities in production (not pilots). (2) Business outcome improvement per capability. (3) Time and cost trend for each new capability, which measures whether you're building a compound foundation or repeating standalone projects.