The key to sustainable investing lies in trustworthy AI
This article is the first in a series exploring the role of trustworthy AI in mainstreaming sustainable investing.
The Mismatch
In the sustainable investing world, there's long been a structural mismatch between the What? and the How? Not for professional asset managers with their expensive datasets, teams of junior analysts, access to senior executives at their portfolio companies. They spend hours and hours researching, assessing, optimising portfolios that seek long-term financial returns while actively promoting positive environmental, social, and governance (ESG) outcomes. They are integrating ESG issues into traditional portfolio analysis to manage risks and drive systemic change. It's an expensive process for which these professionals charge a handsome premium.
But for the retail investors - ordinary people with a workplace pension, an ISA, a small savings pot - sustainable investing is too often out-of-reach. According to a study by Morgan Stanley, 85% of younger investors are interested in sustainable investing, yet many continue to allocate significant portions of their portfolios to traditional, profit-driven assets. It is possible to move your money into more sustainable funds but how do you even do that? Which fund should you chose? Is the performance comparable or will you lose money investing sustainably? It's an expensive process (in time and money) for which these people cannot afford a handsome premium.
And, just to avoid the usual noise on this subject, we should be clear that sustainable investing isn't a "woke fad" that will be going away soon. Europe and Asia-Pacific are the leading regions for sustainable investing, driven by strong regulatory frameworks, active energy transition strategies, and high corporate engagement. The Middle East is seeing increased sustainable investment, heavily driven by sovereign and corporate bond issuances. Despite facing more political volatility and regulatory headwinds compared to Europe, North America remains a significant player with a focus on specific sectors.
Trustworthy AI
So how do we make progress? AI is part of the solution here. One of the obvious advantages of AI is that it can radically democratise access to knowledge. For example, a well-designed generative AI can do the heavy lifting required to steer you towards a sustainable investing strategy. A disciplined agentic AI can organise the timely execution of your strategy. And a relatively simple machine learning AI can continuously optimise your investment portfolio.Building this AI stack isn't the problem though. The real question is whether people will trust the AI with their hard-earned money. Here we run into a category error. If you ask the software design industry, they will tell you that trust is a feature and distrust a bug. More specifically, they will point to one of several frameworks for trustworthy AI. For example, the IBM version lists:
Accountability.
Explainability.
Fairness.
Interpretability and transparency.
Privacy.
Reliability.
Robustness and security.
Safety.
Unsurprisingly, these are all technical features that can be designed into a product or system with greater or lesser degrees of sophistication. They are also not how real people think about the trustworthiness of another.
As politicians, business leaders, NGOs and brand managers have long known, trustworthiness is in the eye of the beholder. In order to act together, I need to find you trustworthy enough for the activity. That level will be low (but not immaterial) for low-risk activities like buying a brand of washing powder. But for high-risk activities like investing your pension, trustworthiness is essential.
For a more productive model for trustworthy AI, we should look to the broader academic trust literature. There we find a consensus around a three-part model of trustworthiness:
Ability trust - does it do what it says on the tin?
Beliefs trust - do I care?
Consistency trust - does it work over time?
These dimensions need to be designed into the AI tools for sustainable investing if it is to be democratised and mainstreamed. This requires thinking less like a software engineer building features and more like someone who needs to be trustworthy enough to get elected, lead a company, run a charity, or build a brand.
Start with the end user
What would an end user need to see to find the AI trustworthy? Is that trustworthiness enough for this high-risk use case? How can these levels be maintained over time?
Some of this work has already begun. One example is Ask_Cameron - Uswitch for sustainable pensions. This is a conversational AI that builds your sustainability profile, assigns you to a "tribe" of like-minded pensions members, and suggests funds that best match your profile. The app specifically explains how and why your profile is built (Ability trust), the branding speaks to the sustainability concerns of Gen Z and Millennials (Beliefs trust), and the feedback loops and gamification are designed for regular interactions (Consistency trust).
Another example is a data-driven materiality assessor developed in conjunction with Professor Paresh Date and Brunel University of London. This is a machine learning AI for auditors, investors and regulators that correlates public sustainability conversations about companies with their share prices daily. The results need to be highly trustworthy in order to inform regulated corporate reporting. The tool specifically explains how and why the most material drivers are surfaced (Ability trust), presents the data in ways that are useful to the job of the auditor and investment analyst (Beliefs trust), and shows the evolution of the drivers' materiality over time in both the conversations and the share price (Consistency).
What is noticeable about both these examples is that their underlying methodologies have been peer rev iewed and are bring published. The teams behind these examples (disclosure: the author works on both) have consciously chosen to be open about the Intellectual Property in order to build trustworthiness with end users.
What other examples are out there of sustainable investing being driven by truly trustworthy AI? It would be great to feature more examples throughout the series.