Artificial intelligence could play a major role in unearthing key data for investors seeking sustainable investments, but discerning unreliable information will be a key challenge, and humans will not be replaced any time soon, ESG experts say.

Investment managers are coming under increasing pressure to measure environmental, social and governance criteria in their portfolios. In recent days, a large number of banks, pension funds and insurers have committed to promoting climate-friendly investment at the United Nations Climate Action Summit in New York.

But a lack of data is making it hard for banks to assess long-term risks and rewards, putting a brake on the market.

AI, whereby computers perform tasks traditionally done by humans, will act as a catalyst for sustainable investment because it will filter essential data that investors currently lack, according to the CEO of one ESG data and analytics firm.

"AI is really the catalyst here that allows us to find ... very fine-grain nuggets of information in massive unstructured database sources," Hendrik Bartel, the CEO of San Francisco-based TruValue Labs, told S&P Global Market Intelligence in an interview.

Data combing

TruValue Labs uses computing power to comb through data related to ESG information on 16,000 securities in real time.

Algorithms sift through the information, classify it and then make it available to investors much faster than a human analyst could, Bartel said. For example, TruValue conducted one year of analysis on the automotive sector using AI and found that it would have taken a human analyst six years.

His company is not the only one looking at AI and ESG.

Jeroen Bos, head of specialized equity and responsible investing at NN Investment Partners Holdings N.V., said his firm is increasingly using machine learning, big data analysis and natural language processing, through which computers analyze speech, in ESG investing.

His company — the asset management arm of Dutch insurer NN Group NV — uses software to screen company conference calls to look for key words used by executives, which might provide pointers as to what ESG approach a fund manager should take. Words like "but" or "however" might suggest doubt or controversies, and if used too often, a "sentiment score" for the company in question decreases, he said.

While there is evidence that this strategy works, Bos said it is still at the testing stage.

TruValue Labs, when it builds data on a company, looks at local, national and international news. It also uses information from nongovernmental organizations, analyst research, blogs and social media, to research papers and analysis by academics.

By doing so it aims to give a full overview of a company's ESG standing, looking at the risks and the opportunities, Bartel said. It might take changes in maternity leave policies into account. Or the data might demonstrate that an energy company is restructuring its business model into renewables, or that a biochemical company has developed plastic utensils that are fully biodegradable.

"All of those are positive things, but most ESG data sets completely discount [them]," he said. Most data take into account potential ESG risks at a company, but that is "a very skewed, one dimensional picture," he added.

'Greenwashing'

TruValue Labs data also excludes company statements or filings to avoid green washing, whereby a firm makes itself appear more ESG-friendly than it actually is.

"A lot of the company self-provided data is highly greenwashed, highly biased and by and large not material," he said.

A company can say that it planted 10,000 trees and did community service, but if it is not material to its business, it has no interest to investors, Bartel said.

"There's no standardization about ESG data and how you have to report it," he said. "Companies can basically report whatever they want."

In this context, AI can be the "ideal playground" for ESG investment, said Lara Kesterton, an ESG analyst at Vontobel Asset Management. Algorithms can reduce human bias and identify more granular ESG data, she said, adding that AI may also be able to flag warning signs at an earlier stage.

But she said AI needed a framework to work with, and that part of the challenge is that there are no clear parameters on risk definitions. Some investors may look more at social justice, while for others it may be the environment.

Unreliable data would also be challenging for the technology, she said.

"While AI can help with human bias, it can't overcome what it can't see," Kesterton said. If data is unavailable or not reliable, ESG analysts speak to the company directly or to other analysts in the local market who know the stock well, she said.

"[That] makes humans in ESG investing far from redundant yet," she said.