I was sitting in a room of global business leaders at the YPO Global Marketing Summit a few weeks ago when the slide went up: ninety-five per cent of corporate AI pilots deliver no measurable financial return. The room did that thing rooms do when a statistic lands a little too close to home. We nodded. We exhaled. And then Stephanie Hills said the line that pinned me to my chair: "Most companies aren't failing at AI. They're failing at leadership."
I looked around and did the maths. Every leader in that room, myself included, had approved at least one AI pilot in the last twelve months. We had budgets, vendors, slide decks, and steering committees. What we did not have, in most cases, was a clear answer to the question of who actually owned the outcome. The discomfort was the point. We were caught between FOMO, the fear of missing out on AI, and what Charlene Li calls FOGI, the fear of getting it wrong. Both fears produce the same paralysis dressed up as activity.
This article is what I took home, what I am doing differently, and what I think every Australian founder and senior leader should consider in 2026.
The real problem isn't the technology — it's the accountability gap
The tools work. The models are extraordinary. The fact that your pilot did not move the needle is almost never about the technology. It is about who owns it, what it is meant to achieve, and whether anyone in the business is genuinely accountable for the result.
I have sat on enough boards and watched enough pitches across four seasons of Shark Tank to recognise the pattern. When everyone is responsible, no one is. AI in most organisations has been treated as a technology project rather than a leadership project, and the consequences are showing up in the numbers.
One leader, one mandate — and why committees don't cut it
According to IBM Institute for Business Value research published in 2025, organisations with a dedicated AI leader achieve roughly ten per cent higher ROI on their AI investment and are twenty-four per cent more likely to outperform peers on innovation. Only twenty-six per cent of organisations actually have someone in that role. The rest have a committee, which is to say they have no one.
This is not a job description you can paste into your CIO's existing remit. The profile is genuinely rare. From what I observed at the summit and in businesses I have evaluated, four attributes matter:
- A proven change agent who has actually shifted culture before, not just attended workshops about it.
- Technically credible enough to ask the right questions of vendors and engineers without being dazzled.
- Business-fluent — they can connect a model to a margin and a workflow to a customer outcome.
- Enterprise-wide authority to make decisions that cross functional boundaries.
Finding or developing this person is a deliberate act of intentional leadership in a fast-moving world. It will not happen by accident, and it will not happen by committee.
Strategy first, tools second — building an AI roadmap that actually works
The most common trap I see is leaders asking each department for a list of AI use cases and calling the resulting spreadsheet a strategy. It is not a strategy. It is a wish list with logos on it.
Charlene Li shared a framework she calls the Double-S Matrix, which scores potential AI initiatives on strategic value and speed to value. Pair that with what she calls a "six quarter walk" — an eighteen-month rolling roadmap reviewed every quarter. Her phrase has stayed with me: strategy in ink, roadmap in pencil.
When David Anderson and I sit down to think through what is next at Big Red Group, the discipline is always the same. Before we talk about tools, channels, or vendors, we ask what the top three things this business must achieve in the next eighteen months are. Everything else either serves those three things or it does not earn the room. AI is no different. If your roadmap cannot survive that question, it is a list, not a plan. The same principle underpins the Big Red Group story — growth through purpose.
The four building blocks — and the one everyone skips
Charlene Li frames AI readiness in four parts: mindset, skillset, toolset, and decision-set. They are not sequential. You develop them in parallel, and most organisations skip mindset because it is the hardest to measure.
Mindset is the leader's job. While many teams still treat using AI as a kind of cheating, the leader's role is to model it openly. Start your next meeting by saying, "Here is how I prepared for this using AI. How did you?" That single question changes the room.
Skillset is about learning, not training. Training is something done to you in a windowless room. Learning is something you experience, usually by giving your team protected time to experiment with real work. AI fluency has four components worth naming: knowing what AI can and cannot do, using it responsibly and ethically, applying it to your specific role, and being able to teach someone else.
Toolset matters less than the conversation suggests, provided you have governance in place. Decision-set is the quiet one. Agentic AI is moving us toward systems that act, not just answer. Agents hate silos. Siloed data and siloed teams will become a structural liability the moment your business tries to deploy anything autonomous. This is also why building a team that can scale matters more, not less, in the age of AI.
Governance that accelerates rather than strangles
There is a Goldilocks principle to AI governance. Too little, and you have chaos. Too much, and you have bureaucracy dressed up as caution. Both kill momentum.
Charlene Li's AI Trust Pyramid is a useful frame. Safety, security, and privacy sit at the base. Above that, fairness, quality, and accuracy. Then accountability. Transparency sits at the top. You build it from the bottom up, and you do not pretend you have skipped a layer.
A real example landed for me. Ally Bank deployed AI in its call centre to remove administrative burden, and what changed was not just productivity. People started telling each other stories about the meaningful conversations they could now have with customers, and the culture shifted. Governance was the thing that let them go faster, not slower. You need good brakes to go fast.
In 2026, Australia's regulatory landscape is still evolving. The Department of Industry, Science and Resources continues to develop voluntary and mandatory guardrails for high-risk AI applications, and the leaders who build trust infrastructure now will move with far more confidence when the rules harden. This connects directly to how customer obsession drives growth — trust is the through-line.
.jpeg)
What I took home — and what I'm doing differently
The organisations winning with AI in 2026 are not the ones with the most pilots or the biggest budgets. They are the ones whose leaders have stopped trying to have all the answers and have become genuinely skilled at asking the right questions.
So here is my question back to you. Of the four foundations — one accountable leader, a strategy-first roadmap, the four building blocks, and governance that accelerates — which is weakest in your organisation right now? Start there. If it is to be, it is up to me. And it is up to you. If you want to bring this conversation into your next leadership offsite or conference, you can find my keynote topics for 2026 on the speaking page.
References worth your time: the IBM Institute for Business Value research on AI leadership, Charlene Li's published work, and the Department of Industry, Science and Resources guidance on responsible AI.
Frequently asked questions
Why do most AI pilots fail in large organisations?
They fail because no one is genuinely accountable for the outcome, the pilot is not tied to a top-three business priority, and the organisation treats AI as a technology project rather than a leadership one. The model is rarely the problem; the ownership gap is.
What does a dedicated AI leader actually do differently from a CIO?
A CIO runs technology. A dedicated AI leader owns enterprise-wide AI outcomes — strategy, capability, governance, and culture — with the authority to make decisions across functions. They are change agents first and technologists second.
How do I build an AI roadmap without being overwhelmed by tool choices?
Start with the three things your business must achieve in the next eighteen months, then map potential initiatives to value and speed. Strategy in ink, roadmap in pencil — review it every quarter and cut anything that has slid into the dead-ends quadrant.
What is the Double-S Matrix and how do I use it in my business?
It is a simple grid that scores AI initiatives on strategic value and speed to value, producing four quadrants: Momentum Makers, Strategic Bets, Quick Wins, and Dead Ends. Use it as a quarterly triage tool with your leadership team so investment follows priority, not noise.
How should leaders approach AI governance without slowing everything down?
Treat governance as the brakes that let you go faster, not a compliance hurdle. Build it in layers — safety and privacy first, then fairness and accuracy, then accountability and transparency — and keep it proportionate to the risk of each use case.
