Generative AI has transformed the way we work over the past two years, and perhaps unexpectedly, also in the boardroom.
Discussing technology at board meetings is nothing new, but the advent of Generative AI (ChatGPT/CoPilot, Gemini, et al) has prompted a shift in process.
Typically, a CTO is tasked with a desirable business objective to be met or might bring their own technological improvements to the table. With public access to these AI systems readily-available, board members now are trying AI for themselves, like what they see, and it’s shaping their thinking.
There are a number of factors at play here:
The perception of competitive advantage
Companies often want to apply AI to systems simply because their competitors have it (or are talking about it). Keeping up with the AI Joneses can be a factor in adopting AI, but not a valid one. AI should be adopted when there’s a clear, tangible benefit to your organisation.
Demonstrating modernity
If you’re not discussing AI adoption in your annual report, you risk being seen as behind-the-times. This can be a valid concern if you work for a large organisation – perception is 100% of reality for those not involved in the detail. No-one is saying you can’t have a plan, but consider how you’ll execute your vision, and what makes your offering different (if not unique), and more importantly, worthwhile.
It looks easy
The interface is simple. Making a request of these natural language processors (NLP) is no more complex than using a search engine. You type something in, it understands you and replies – and usually accurately. What’s not to love? How hard can it be to get right? Actually, very hard. There have been numerous cases around large language models (LLMs) “hallucinating” – creating fictional “facts” from its training data:
● A lawyer who used AI to prepare his case, but which generated fake cases as precedent
● An airline’s AI chatbot assured a customer he could claim a bereavement discount for a flight to a funeral after the flight had commenced, against their policy
● McDonald’s drive-thru voice recognition system ordering – amongst other things – bacon added to ice-cream and adding $200+ worth of chicken nuggets to an order
Making something simple to use can be extremely challenging. How are you ensuring your chatbot isn’t going off-the-rails?
The risks aren’t obvious
Using Generative AI to get holiday destination ideas or to help plan-out a wedding is a great proving ground, but that isn’t the same as using your corporate data for its AI model training.
As with any new technology, there are a whole host of potential exploits for LLMs, and mitigations that can be implemented. Essentially, if you have a chatbot trained on sensitive company data, you don’t want the system exposing that to your competitors, journalists, or hackers.
AI is the next step in your regular IT journey, and so must be paired with cybersecurity to keep it ringfenced, just as you would any other company digital asset.
Spare a thought, then, and plenty of patience for your CTO. The phenomenal breakthrough pace of Generative AI has captured the modern world and rapidly become the “new normal”. As with everything in business, a cautious rollout is always best – see the numerous examples from Google and their “AI overviews” feature. They acknowledged that they didn’t get it entirely right straight away, and you probably don’t have their budget. When thinking about AI, consider Apple’s general approach to new markets – never the first, but aspiring to be the best.
Oh, and if you were wondering – no, AI did not write this article.
Jaymie is a co-founder of AI Wales (http://www.aiwales.org/), which holds free-to-attend events explaining the principles, ethics, and mechanics of AI. With a software development background, Jaymie advocates for the sensible, safe introduction of AI into businesses and wider society.
