The Value of AI in Writing Board Packs

Given the speed with which organizations are working to add artificial intelligence (AI) to their products and internal capabilities, one might expect the corporate directors overseeing them to have embraced the technology in their own duties. Yet, AI remains conspicuously absent from boardrooms.

It’s not that board members aren’t interested in the technology. The Board Intelligence and the Corporate Governance Institute report Adapt or Perish—The Technology Challenges For Boards found that more than 80 percent of directors are keen to use technology that could enhance their boards’ performance, and we’ve stopped counting how many times we have heard directors ponder whether OpenAI’s ChatGPT could summarize their board decks.

Instead, board members are weary of the inclusion of AI in the boardroom as they don’t trust their own judgment or understanding of the technology. Adapt or Perish found that one-third of board members worry about their lack of technology expertise, while nearly half can think of at least one fellow director who should be replaced by someone more tech-savvy.

With that in mind, what should directors know about AI and how to use it in the boardroom?

How does AI work?

AI is the consumer term for machine learning, and machine learning turns programming on its head: rather than tell a computer what steps to follow to reach an end goal, we tell the computer what end goal to reach and let it find the steps to get there.

To build a tool such as ChatGPT, those who train the machine feed it millions of books and other text sources, then hide words at random in that training data and ask the computer to play fill-in-the-blank. Through trial and error, the program uncovers patterns and finds ways to guess the hidden words based on the surrounding text. It can then simulate a conversation by filling in the “blank” at the end of sentences or prompts.

Simply put, text-generating AI provides the words that are most likely to follow the ones already used. It does not truly answer the question asked but rather autocompletes it.

What should directors make of it?

The inner workings of AI matter because they show where the technology could fall short in the boardroom—and where it could shine.

One potential application in the boardroom is using AI to summarize board decks. A Board Intelligence and NACD report, Board Packs: The Elephant in the Boardroom, found that nearly two-thirds of directors see at least three areas of concern within their board materials, including “too much detail,” “too much jargon,” and “poor structure.”

Yet, using AI to write summaries of board packs can be dangerous. While AI can work wonders, it doesn’t live in a world of accurate or inaccurate facts but rather more or less statistically likely sentences. It can therefore incorrectly summarize key points, miss what is important, or even completely fabricate information. The only way to know whether the AI did a good job would be to read the whole report, which defeats the purpose for which it is being used.

Using AI this way also amounts to addressing the symptoms rather than the cause of the issue: the people writing board reports aren’t given the right tools for the job. Management teams are often poorly briefed; many don’t know how to deliver the focused, well-communicated critical thinking that directors need; and the templates or training they receive are easy to misuse and forget.

The better move for directors, then, is to equip their authors with AI tools that can sharpen reports as they are being written, instead of trying to fix problems afterward.

To start, AI can guide the writers’ thinking, steering them around common mistakes. For instance, AI’s text analysis capabilities can flag whether the content of a report neglects to link to the organization’s goals, is too backward-looking, does not address risks, or isn’t sufficiently candid. Board pack authors can then decide how to act on that feedback because, as the subject matter expert, they’ll know best. What matters is that the AI challenged the thought process and pointed out potential gaps.

Great thinking won’t do much if the text isn’t easy to parse and process. Here, too, AI can help by guiding the writing process and analyzing the following: Has an executive summary been included? Are sentences concise or rambling? Is the vocabulary simple or full of jargon? Is there simply too much text for the average time-poor director to consume?

Since AI excels at modifying text, it can fix these issues with the click of a button by simplifying or shortening paragraphs or extracting the report’s key points and putting them in an executive summary. Those writing the board pack remain in charge of the content, ensuring no hallucinations slip through while saving significant amounts of time by letting AI do the heavy lifting.

To guide the use of AI in their writing board materials, directors should consider the following questions:

• Is the AI tool used by those writing the board reports stimulating critical thinking and guiding writing, or is it doing the thinking and writing for them?
• Is a human always in the loop? More generally, is AI being used to augment and improve people’s work, rather than automate it entirely?
• How safely is it doing so? Is the AI tool forgetting the content it sees after use, or is it using it to self-improve at the risk of sharing the information with others at a later time?
• When deciding on an AI tool to augment this process, should a public or private instance of the tool be used?
• If a public tool is used, how can confidential information be protected and not used to further train the tool?

AI will not—and should not—replace board pack authors or directors. But it can become a trusted ally when used to improve writing in the long-term rather than to automate it in the short-term.

Niamh Corbett is head of Americas at Board Intelligence. She spent several years at Morgan Stanley advising FTSE boards and management teams on their capital market interactions. Corbett speaks and writes regularly on topics related to financial regulation and has coauthored a fortnightly column in the Financial Times.