In the mid-1990s I was invited to a meeting with senior executives from a few leading US banks. My message was simple: going forward, digital channels will drive sales growth. They responded by asking me to leave the room, convinced that customers would never purchase banking products online. Over the following years, these same banks created much ‘brochure-ware’ - a pale imitation of a truly digital experience. When a technology is too complex to ignore but too daunting to truly integrate, leadership defaults to theatre: activity that looks like progress, without the decisions that create it.
In 2026, that theatre has two main acts: Pilot Proliferation—fifty experiments, none of which ever reach the core business—and the LLM License Binge—ten thousand licenses, and a quiet hope that productivity will simply appear.
The honeymoon is over. Markets are no longer rewarding “AI potential.” They are rewarding EBITDA impact and strategic defensibility.
And this is where the dangerous vacuum opens up. Senior executives can see the potential, but struggle to convert it into decisive action. Board members sense the scale of the opportunity, but can’t quite picture what it means in practice. Meanwhile, middle managers are stuck between signals & safety—confused, nervous, and quietly defending the status quo because they don’t know whether anyone has their back when the trade-offs get real.
To bridge this, we must stop treating AI like a software update and start treating it as what it actually is: a structural mandate. That means the board stops asking for “use cases” and starts governing value creation, operating model change, and risk ownership—with clear division of labour across the Main Board and its committees.
Strategy: Valuation and Viability (Main Board)
The Main Board’s job is not to sponsor a portfolio of disconnected pilots. It is to decide—explicitly—how AI changes the economics and defensibility of the firm. The question is not “where can we deploy AI?” but “where does AI collapse our costs, compress our cycle times, or shift our ability to differentiate—across the value chain?”
We have already seen sectors like software, education, and law meaningfully revalued because AI stripped away traditional moats. Your sector is not immune—only lagged. Which is why the competitor set must be redrawn. Your rivals are no longer just companies that look like you. They are AI-native attackers who can collapse margins because they don’t carry your legacy overhead, your complex governance, or your multi-year process debt. If you define “competition” too narrowly, you will measure performance against the wrong threats until it’s too late.
In industrial sectors—logistics, manufacturing, energy, pharma—AI is also worthless if the last mile is broken. The strategic battleground is now decision-to-action latency: how quickly can your physical business move when an algorithm spots an anomaly—inventory risk, quality drift, fraud patterns, demand spikes, safety issues? If an AI system can see a problem in seconds but your organisation takes weeks to act, the advantage belongs to someone else.
Once the strategic path is clear, the board must do something many organisations avoid: stop things. This is where theatre dies. If an initiative does not have a credible path to industrial-scale deployment in 180 days, it is a hobby, not a strategy. And “path to scale” cannot mean a demo: it means a funded backlog, a named owner, secured data access, security sign-off, operating model integration, and a route into business-as-usual. The board’s anti-pattern to avoid is “use-case bingo”—fifty pilots, ten vendors, no wedge into the core business.
Succession & Profiles: The Expert Balance (Nominations Committee)
NomCo’s job is not to “add an AI expert” and declare the board covered. It is to create cognitive diversity that improves decisions without breaking board functionality. That means new voices—leaders and directors who have navigated disruption elsewhere, who understand pace, product, data, risk, and talent—without becoming lone-wolf disruptors who reduce governance to ideology.
The competency test is simple: can the board challenge the CEO on AI without defaulting to nodding at slides it doesn’t fully understand? If the board cannot probe an AI strategy—its economics, its operating model implications, its controls and dependencies—then it is not providing governance. It is providing comfort. The anti-pattern here is appointing one “AI-savvy” non-exec as a talisman while the rest of the board remains unable to interrogate decisions.
People & Incentives: Rewarding Bravery (Remuneration Committee)
RemCo holds one of the most underused levers in transformation: incentives. If leaders are only rewarded for maintaining the core, they will cling to the old—quietly, rationally, and relentlessly—because that’s how bonuses are protected and kingdoms are preserved.
In 2026, governance must reward dual performance: running the core with discipline while reshaping it with intent. That means explicitly rewarding the behaviours most organisations avoid: decommissioning legacy work, reducing process debt, redesigning teams around new workflows, and turning AI-enabled productivity into measurable margin—not just “activity.” If nobody gets rewarded for stopping outdated work, it will never stop. The anti-pattern is setting “efficiency targets” while quietly rewarding leaders for keeping headcount, scope, and legacy processes intact.
Risk: Own the Black Box (Audit Committee)
Fiduciary duty in 2026 includes algorithmic traceability. If an AI-driven system changes pricing, shifts supply chain decisions, alters credit outcomes, or impacts customer experiences—can the Audit Committee explain why to a regulator, a court, or a front page?
AI is not a tool bolted onto the business; it becomes part of the firm’s decision-making nervous system. That nervous system must be auditable with the same seriousness as financial controls: what data shaped the outcome, what model behaved how and why, who owns monitoring and thresholds, what happens when it goes wrong, and whether decisions can be reproduced under scrutiny. The anti-pattern to avoid is treating model risk as a privacy add-on rather than a core control system.
The Narrative: The Stakeholder Bridge (The Chair)
The CEO must tell the story. But the Chair must ensure it’s true, stewarding credibility, shareholder trust, and the board’s intent. Your narrative cannot be “we’re adopting AI.” That’s a press release, not a strategy. The narrative must answer how AI makes this company a winner in the new landscape—and why shareholders, employees, and customers should believe you.
This is also where capability, cadence, and trust stop being slogans and become execution assets. Real reskilling isn’t an “ Generic AI literacy module”; it is role-specific redesign with visible outcomes—what work is automated, what decisions move closer to the front line, how teams are reshaped, and how careers are protected through the transition. Likewise, the organisation won’t change if the top doesn’t change. If the C-suite and Board don’t alter how they run the business—weekly performance dialogues, investment decisions, risk reviews—AI will remain a side project. Leaders don’t need to be power users, but they must be fluent enough to demand measurable impact and to notice when theatre is creeping back in. And none of this scales without trust. Speed requires consent. The board needs to communicate with radical clarity about what AI will change, what it won’t, how decisions will be governed, and how people will be protected. Trust isn’t a comms exercise; it is the only currency that buys the pace required to win.
The Board’s Test
By the next two board cycles, you should be able to answer—plainly: where will AI expand revenue or margin, what must be true to scale it, what are we stopping to fund it, and how will we prove it is safe and accountable?
If you can’t answer those questions, the organisation isn’t transforming. It’s rehearsing.
The lack of alignment between the corporate strategy and the AI initiatives is for me a huge challenge (just like with digital initiatives in a not-so-distant past by the way).
I see quite often people say "a list of use cases doesn't make a strategy". Unfortunately most books on AI don't define what a strategy is and implicitly equates it to a list of business priorities/strategic goals (including “Rewired” from the Firm), while a strategy is a set of integrated choices about where and how to compete (whether you use McKinsey's definition or Roger Martin's).