AI and ESG go together like love and marriage

Patterns. AI is about finding patterns in data. And thanks to the internet, the internet of things and the digital transformation of the global economy, there is a lot of data containing patterns that AI can identify.  

At a recent ESG conference, DTE London, organised by GRC World Forums, Sue Turner, OBE and Chief Executive of AI Governance Limited, said: “There are two sides to AI in ESG - how it is being used to improve ESG investing and how you can use it in your business to improve your ESG performance.” 

So, let’s drill down 

AI and ESG investing 

AI is important to investors wishing to have high EEG exposure for multiple reasons.

According to Sue Turner, two of these reasons are: 

• Research has used AI to look at 1,400 companies from 34 stock markets globally and found that those with better ESG scores tended to have better financial performance — although, of course, this does bring up the age-old chestnut, correlation does not necessarily mean causation. 
• AI can also be used as a kind of friend of the investor ascertaining ESG authenticity. So, for example, ESG ratings often rely on the information a company discloses about itself. But AI can find patterns in data, and data can take many forms, including audio, video, images and text. Various AI tools, including natural language processing (NLP), can analyse this data. So, for example, it can read annual reports and press releases, look at advertising, including recruitment advertising and social media posts and come up with analysis of diversity and sustainability practices and how well they are adhered to within an organisation. Sue Turner says: “This means we don’t have to rely on what a company says in its annual report about its ESG activities & impact.” 
• But companies can get wise to how AI can be used in the ways described above. Sue Turner said: “The corporations themselves are testing whether their public disclosures about climate change will pass muster and how AI will interpret the words and phrases used in their quarterly earnings calls. And many CEOs  have been coached to change the language they use - avoiding some words and phrases, using others more often - to fit the narrative they want us to consume.” 

Then we turn to AI as a tool for ESG within an organisation. Examples include:

• Smart building — AI could be applied to a large building such as a hospital, factory, warehouse, or office and analyse historical data on energy and water usage and combine this with data relating to weather forecasts to manage energy and water usage optimally.
• AI used in forecasting and stock control to reduce waste fuel used in distribution and to cut plastic in packing.  
• AI applied in aircraft to improve fuel efficiency, such as monitoring wind speed and data on temperature to recommend optimal altitude at any one time. Sue Turner cited an example of Norwegian Air: “Once the data is collected, it is automatically sent to the aircraft’s systems, AI can optimise the flight path.” And the conclusion is that if AI was used across Norwegian Air’s 160-aircraft fleet, carbon emissions could be reduced by 16,000 tonnes yearly.”  
• AI and drones, for example, a test in Bristol was able to use AI to study location of litter so that strategically placed ads can be placed to dissuade people from leaving litter. 
• AI used in bike and scooter rentals to ensure data in supply match and minimise unnecessary use of resources. 

Applying ESG to AI

So AI can support ESG, but there are dangers with AI too, which takes us to ethical AI:

• AI analysis of data can pinpoint potential customers, by analysing geographic specific factors but thus can discriminate against certain ethnic populations that are particularly common in a certain area.  
• AI in recruitment can build inferences based on historical data but can potentially work against building a diverse team.
• Also, in recruitment, AI can interview candidates but, in the process, discriminate against less tech-savvy people. 
• AI chatbots offer many advantages, but by learning from data, but this can turn up undesirable results; for example, Microsoft’s Tay chatbot was supposed to be a digital female teenager that would learn from interacting with people on Twitter.  But it was taken down after less than a day when it learned from interacting to be a Hitler-loving, feminist-bashing troll. Sue Turner said: “The social impact of AI being ‘led astray by humans’ needs to be considered.” 
• Be aware of the energy cost of AI. Running an AI algorithm is not necessarily that energy-intensive, but the cost of training the AI tool can be. Deep learning is can have an especially high energy cost.

How to manage AI and mitigate risk

So what can companies do to harness AI and overcome the negative dangers?

They can:

1. All board members must have a basic knowledge of AI. A survey showed that 65 per cent of executives in businesses using AI couldn’t explain how their AI models make decisions.  
2. Related to this, don’t abdicate responsibility for AI to say IT or marketing; it is too important for that. Again, this is something for which the board must take responsibility. 
3. Then there are the risks from data. Sue Turner says: “Establish data & AI governance that fits your culture and enhances wider society.”
4. Sue also says you need to get creative and think about the business problems AI could help serve. 
5. A key area is risk analysis — you need to understand AI risks, including the risk of new regulations and the possibility that your AI tools won’t adhere to new rules.
6. Building upon risk, you need to understand your supply chain and consider third parties in the supply chain which use AI.
7. Governance: ensure good governance practices are embedded into AI from the beginning. This can build upon existing ESG strategies and ensure diversity is embedded into your ethics committee or equivalent. She cites the example of Zillow, which applied AI, speculating with property. The move cost the company “$300m in the space of a few months, and the company’s share price crashed, wiping $9bn off its value.” 
8. Staff and culture make another important point. There is, of course, a skill shortage in this area; companies can’t just buy in talent; they have to spend money on in-house training staff and building expertise.