To borrow the title of a recent Microsoft report… AI at Work Is Here. Now Comes the Hard Part.

In other words, if you’re a business leader, you need to be thinking about AI. But I suspect this isn’t news to you – we’re awash in headlines painting it as either the impending downfall of humanity or a technological saviour.

I’d like to inject a bit of nuance into that polarised conversation. My next few blog posts will focus on AI, or more specifically, on what you as a business leader need to know about it. What are the easy wins here? What are the biggest pitfalls?

To start with, let’s cover the basics – what exactly is AI?

You will have heard a huge spike in the number of references to AI recently. Most of them are about generative AI specifically, but the term encompasses a lot more.

As IBM puts it:

“Artificial intelligence (AI) is technology that enables computers and machines to simulate human learning, comprehension, problem solving, decision making, creativity and autonomy.”

It has a long history – the famous Turing test, which states that machines that exhibit human-like intelligence have “artificial intelligence”, was developed back in 1949. In a process called “machine learning”, AI models are trained on specific data sets, which they can then analyse or extrapolate from.

Generative AI takes things one step further. It’s trained on “deep learning” (very complex machine learning) models that mimic the human brain’s learning and decision-making processes. After that training and fine-tuning, it can then respond to prompts, creating something new – text, code, images, audio, video – based on the data that fed it.

You’re already using AI

Though gen AI is creating the current buzz, you’re almost certainly already using AI more broadly.

Do you use navigation apps like Google Maps or Citymapper? A voice assistant like Alexa, Bixby or Siri? Search engines? Streaming services? Then you’ve interacted with AI.

Even in 2020, which feels like an age ago in terms of AI adoption, McKinsey reported that half of respondents said their organizations had “adopted AI in at least one function”, with service operations, product or service development, and marketing and sales the top reported business functions.

But of course, those examples are ‘traditional’ AI rather than generative AI.

Gen AI is great for some things – but not everything

I suspect some of the hype around generative AI is just the usual excitement about a new technology. Remember when apparently every movie was going to be shot in 3D? Or when we were all going to be wearing glasses with integrated cameras?

But the potential impacts in this case are a little more far-reaching, and the technology can certainly help you in some situations.

For example, if you need to present information to various stakeholders, you could ask a generative AI model to reframe your presentation for each different audience. You could also ask the AI model to respond to your work as someone from a specific group, helping you understand different perspectives and prepare for possible challenges.

It could also have applications in leadership training and advancement, from data-based personalisation of materials to creation of simulations for scenario-based training content. Though the efficacy of these tools is still variable, there’s little doubt that they’ll continue to improve, and as RISEUP Global reports, major companies like Google and IBM are already using them.

But there are many cases where generative AI is unhelpful. It’s not always accurate, and it’s certainly not unbiased, which can be extremely dangerous if you don’t know how to spot that. It can also weaken your critical thinking and research skills if you over-rely on it.

Ethan Mollick’s One Useful Thing has a lot of sensible things to say on this topic. I particularly appreciate:

“Knowing when to use AI turns out to be a form of wisdom, not just technical knowledge.”

You need to consider safety and security

When you do or don’t use these tools should also be informed by safety concerns. For example:

The information you provide to an AI tool may be accessible to the provider of that service, and could be used to train AIs, so be careful about sharing anything internal or confidential.

That same information, stored by a third-party service provider, could be open to cybersecurity threats.

AI often confidently gives incorrect answers, an issue known as AI hallucination. This could have a negative impact on decisions you make based on it.

Similarly, the output is not always high quality – one often-used example is the fact that AI-generated images tend to get people’s hands disastrously wrong…

As generative AI is a new tool, the full range of threats is not yet understood. You’ll need to stay informed of recent developments like prompt injection and data poisoning

Gen AI has major environmental impacts

I’ve spent a lot of my career advocating for more environmental responsibility in business. And unfortunately, there’s no getting around it: generative AI is an ecological nightmare right now. In fact, almost half of businesses surveyed have weakened their sustainability goals specifically because of generative AI.

The water use, electricity use and carbon emissions can be “staggering”, to quote MIT. Asking a generative AI chatbot like ChatGPT a question rather than using an online search engine is far, far more wasteful. Estimates vary, but to use one dataset:

One ChatGPT query is 10 times more energy intensive than a Google search

One ChatGPT query causes 340 times the carbon emissions of a Google search

AI data centres are projected to account for 3–4% of global energy use by 2030, versus 1–2% in 2024

This is just another reason to carefully consider how you use generative AI. I understand it’s an exciting new technology, and business leaders simply have to reckon with it or risk being left behind. But personally, I think it’s best to limit it to use cases where the benefit is clear and the safety risks are low. Because unless it’s really elevating your business, it’s just not worth the environmental impact.