Written for the NED.
A non-executive can’t oversee well what they’ve never seen. Before the next strategy discussion, a board benefits from watching AI work on real material — not to operate it, but to ask sharper questions and give better assurance.
Boards are increasingly expected to have a view on AI — its opportunities, its risks, management’s plans. Yet most non-executives are forming that view from the same sources as everyone else: headlines, hype, and the occasional alarming anecdote. That’s a thin basis for the two things a board owes the business on any material topic: informed encouragement and informed challenge.
The gap isn’t technical knowledge. A NED doesn’t need to operate AI any more than they need to run the finance system. What helps is having seen what it genuinely does — because oversight of something you’ve only read about tends to be either credulous or unduly fearful, and neither serves the company.
Why seeing changes the quality of oversight
There’s a specific reason abstract AI briefings leave boards no wiser: they ask non-executives to imagine the leap from “clever tool” to “consequential in our business,” and that leap is exactly where judgement is needed. Watching AI work on recognisable material closes the gap. The board stops debating a concept and starts assessing a capability — which is the proper posture for oversight.
What a board benefits from seeing
A short, board-grade demonstration uses the kind of material non-executives actually handle:
A financial model, stress-tested live. An assumption changed in plain language, with upside, base and downside moving together — showing how management could interrogate numbers faster, and prompting the question of how that changes the quality of what reaches the board.
A long report reduced to its risks. The genuine obligations, exposures and deadlines pulled from a dense document in seconds — and the immediate governance question of who validates that extraction.
Two documents compared. Changes and risks between contract or policy versions surfaced instantly.
A decision pressure-tested by a panel. This is the one that resonates most with a governance mind. AI can convene named expert personas — a risk reviewer, a commercial sceptic, a red-teamer whose only job is to find the flaw — to challenge a proposal from several angles. Used this way, AI is a challenge instrument, not a chatbot.
The move a NED should notice
The most important thing to observe isn’t the speed. It’s that how you frame a request to AI determines whether you get an honest answer or a flattering one. “Show me why this plan is right” produces a confident echo; “argue the strongest case against this plan and tell me what we’re not seeing” produces something useful. That distinction — comfort versus challenge — is a governance instinct rendered in software, and it reframes AI from a productivity toy into something a board should care about how management uses.
The oversight questions that follow
Once you’ve seen it, the right questions become obvious: Where is management already using AI, and do they know where? Who owns the outputs, and who is accountable when they’re wrong? What data must never go near these tools? And are we accelerating a process that works — or one that’s quietly broken? That last question matters, because AI applied to a flawed process doesn’t fix it; it reaches the failure faster.
A short board prompt
For a private, non-confidential trial of the challenge instinct:
“Act as a sceptical non-executive director. Here is a strategic proposal management has put forward: [describe it, no sensitive detail]. Argue the strongest case against it, identify the assumptions that haven’t been tested, and list the three questions a board should ask before approving it. Do not reassure me.”
It’s a small demonstration of AI as an aid to assurance rather than a threat to it.
What to do next
Ask for a short, board-level AI session before the next strategy discussion — one that shows what AI does on real material and equips the board with the questions to put to management. Assurance improves markedly when the board has seen the thing it’s being asked to oversee.
In closing
A board that has watched AI work gives better assurance and asks sharper questions than one working from headlines. The point isn’t to make non-executives operators; it’s to make their oversight informed.
If your board would value a session pitched at exactly this level — what to see, and what to ask — that’s something Savant and Axulu provide for boards. Where deeper assurance is needed, Savant can also connect boards to experienced technology and security leaders who can advise on AI oversight.