Why most board packs fail, and why AI won't fix it on its own
I ran a live AI training session for GrowCFO recently, working with finance leaders on how to use AI properly in board reporting. It's a topic I know from both sides. Over 20 years I've sat in the room reviewing board packs as an adviser, and I've written more than a few myself. I've also spent the last two years building and teaching AI workflows for finance teams. This post sits at the join of those two things.
The four reasons board packs actually fail
Ask most finance teams why a board pack didn't land well and you'll get a shrug, or a story about a difficult NED. In practice, it's almost always one of four things:
Length. It gets read in the taxi on the way in, and nothing is retained.
Data without a story. The board is handed numbers and left to build its own narrative, usually the wrong one.
Buried decisions. The three things that actually need a decision are on page 14, 39 and 62.
Unexplained movement. Numbers shift from the last pack or the budget with no explanation, and the board notices before anyone addresses it.
None of these are writing problems. That matters, because it's exactly where most people go wrong when they bring AI into board reporting.
Board reporting is a context and judgement problem, not a writing problem
The instinct is to open ChatGPT or Claude, paste in some numbers, and ask for a board update. What comes back reads like it was written by a stranger, because it was. Generic AI has no idea what your board actually cares about, what's contentious right now, or what got asked last quarter that never got answered properly.
That's not a prompting problem. It's a briefing problem. Think of AI as a very capable new employee on day one. You wouldn't hand a new joiner a spreadsheet and ask them to write the board update without ever telling them who's on the board, what the business is trying to do this year, or what's currently uncomfortable. Yet that's how most people use AI for this.
The fix isn't a cleverer prompt. It's building the context once, properly, and reusing it every cycle. Done well, the groundwork that used to take hours before every board meeting takes minutes, because the thinking is already captured.
Protecting your credibility
There's a line I'm firm on in every session, and it matters more the more AI gets used in finance: AI can restructure, summarise and interrogate the figures you give it. It cannot calculate them, and it should never be trusted to. Every number that reaches your board has to originate in your finance system, your model or your consolidation, and a human needs to be able to trace it back. If a number reaches the board via an AI tool without anyone having checked it against source, that's not a productivity gain. It's a governance failure waiting to be found.
The part most people skip
Most board reporting failures aren't errors. They're messages that landed badly with a specific person, for a reason that was entirely predictable in hindsight. The chair who wanted the headline first. The NED who always focuses on cash. The one who reads inconsistency as evidence something's being hidden, even when nothing is.
That's the second half of the job, and it's the part generic AI use skips almost completely, because it requires knowing your board as individuals, not just your business as a set of numbers.
Where this goes
None of this is difficult once you've done it properly once. It's a discipline, not a trick, and it compounds. The teams I work with build it into a repeatable routine: get the context right, use AI to specify and pressure test the pack rather than just write it, and update the context after every meeting with what actually happened.
If you're responsible for board reporting and want to see this built out properly for your own team, that's exactly what I run sessions on.
