The missing piece of the AI puzzle could drive better outcomes across the capital markets
During the panic of 1873, 36 percent of corporate bonds in the US defaulted and over 100 railroad companies issuing bonds collapsed. There were no scorings or ratings to objectively assess the underlying financial strength of these bonds.
In 1909, John Moody founded the world’s first rating agency which issued ratings on railroad bonds to define and limit investor risk in investing in these bonds. Studies show that defaults and capital loss declined as a result of Moody’s ratings.
Like AI, railroads were the hottest investment topic of the times. Today, in many ways, we are in the same situation today as in 1873. Some estimates have it that up to 30 percent of the S&P market capitalization ($16,740 trn, as of 23 November 2025) is due to AI boosterism.
Yet many key players do not think it’s a bubble: ‘I sincerely think that there is no AI bubble,’ declared Larry Fink, Blackrock CEO, during the World Economic Forum’s meeting at Davos in January 2026. Debbie Weinstein, Google’s EMEA president, predicts that AI-based productivity gains for Europe alone will be $1.3 trn within the next years.
AI is becoming the defining technology of our times and future generations, reshaping productivity, decision making and entire business models. Deloitte’s study of CFO priorities for 2026 cites the ‘digital transformation of finance’ (driven by AI) as first on their minds.
But as AI systems become more powerful and more deeply embedded in critical processes, a fundamental problem has emerged: we lack reliable, standardized ways to evaluate their quality, safety, economic impact and ethics.
‘We see companies gearing up with AI but in many different ways. There’s a need for a standard to structure AI strategies and performance in companies,’ says Patrick Rozario, director at the worldwide accounting firm Moore Global in Hong Kong.
Just as financial markets depend on credit ratings – and sustainability strategies rely on ESG metrics – the rapidly growing AI economy now requires its own layer of scoring and assessment frameworks. These will help companies, regulators and investors navigate a landscape in which AI models differ dramatically in accuracy, risk, performance, cost and trustworthiness.
‘Investors with a long-term perspective and strong risk awareness, such as pension funds, should be able to assess and monitor a company’s likely future AI performance. Currently, there are no systematic guidelines or set of standards to do this,’ says Eduardo Lamers, Superintendent of ABBRAP, the Brazilian association of pension funds.
AI-driven scorings of companies’ AI strategies and actual performances would constitute an opportunity for IR professionals to report relevant and compelling contents to investors. Here are the key reasons for introducing this approach.
AI performance varies widely, but no standard exists
AI models differ massively in accuracy and error rate, hallucinatory behavior, domain knowledge, computational cost, security vulnerabilities, robustness and consistency.
Despite these differences, AI is described in vague terms such as ‘state-of-the-art accuracy,’ ‘enterprise-grade,’ or ‘trustworthy,’ and so on. Investors have no way to compare one AI system with another on a consistent basis.
AI scoring introduces standardized benchmarks, enabling businesses to evaluate models using real, comparable data rather than marketing language.
AI introduces new, hard-to-detect risks
AI risks exceed traditional IT risks in many ways:
- Bias and discrimination in model outputs
- Data leakage and unsecure training datasets
- Model drift over time
- Deepfakes and synthetic identity fraud
- Compliance violations, including the EU AI Act, GDPR or financial reporting rules
- Explainability and auditability gaps.
Boards and risk committees need clear risk classifications – similar to credit or operational risk” ratings – so they can govern AI responsibly, and IR professionals need specific and uniform guidelines for reporting their company’s AI plans and performance.
AI scoring systems categorize and quantify models by risk level and flag vulnerabilities early, reducing regulatory, legal and reputational exposure.
Investors need protection against hype
AI is driving stock market valuations, particularly among AI infrastructure, semiconductor and software companies. But investors currently lack reliable metrics and reporting frameworks on AI maturity, transparency on AI R&D pipelines, evidence of actual revenue impact or clarity on whether a company’s AI claims are real or inflated.
AI ratings could help investors differentiate between:
- Companies with real, scalable AI capabilities
- Companies overstating their AI readiness
- Companies at risk of disruption, and
- Companies which may lose substantial value in a stock market correction.
Just as accounting standards prevent financial manipulation, AI scoring frameworks prevent AI-washing – exaggerated or misleading AI claims.
Companies need AI support before deployment
When companies adopt AI systems, they must evaluate critical questions:
- Where shall we deploy AI and how to do it so it makes a financial difference?
- How accurate is this model?
- How likely is it to hallucinate or output harmful content?
- How does performance differ by demographic group?
- How much data does it retain?
- What governance and logging systems exist?
- What happens when the model fails?
Currently, most companies acquire AI tools with minimal technical due diligence, simply trusting vendors. AI scoring would provide a structured assessment – an independent quality and risk ‘nutrition label’ that guides procurement decisions.
Regulation is increasing
The EU AI Act, the US NIST AI Risk Management Framework and global AI safety institutes all require transparency, documentation, monitoring, testing and risk assessment around AI use.
But compliance becomes complex and costly without standardized evaluation methods. AI scoring could effectively become the compliance engine, providing regulators and companies with model classifications, structured testing, traceability documentation and audit trails.
This reduces the cost and complexity of AI governance and incites investor confidence.
Consumers need to trust AI
As AI expands into law, healthcare, finance, HR and education, customers are asking: Is this AI reliable? Does it protect my data? Is it safe for critical decisions? Will the model still be supported in 3 to 5 years?
Absent trusted indicators, adoption slows. Ratings could create trust in a similar manner to UL certification, ISO standards or cybersecurity ratings, which could accelerate safe commercial adoption.
AI scoring drives competitiveness and innovation
Companies with high AI scores could demonstrate their superior capabilities, stronger governance, higher operational efficiency, lower risk exposure or better long-term value creation.
These, in turn, become competitive differentiators, shaping procurement or investment decisions, M&A activity, talent attraction and executive strategy. In other words, AI scoring becomes part of the competitive identity of the firm.
As the rate of AI adoption increases, that is only set to become more important. Generative AI achieved a 39 percent adoption level in just two years, a milestone which took the internet five years and personal computing nearly 12 years. Even so, studies show that companies have difficulties deciding how and where to use AI, with only 13 percent or less being AI-ready, according to a Cisco survey.
An AI scoring system would help quantify and monitor companies’ adoption, benchmarking it against the latest developments in AI.
Giving companies, investors and regulators standards and confidence
AI is advancing faster than any previous technology. But without reliable ways to measure quality, risk, performance and impact, organizations operate blindly – and markets become vulnerable to volatility and hype.
Just as credit ratings enabled modern capital markets, AI ratings will enable the safe and effective scaling of the AI-driven economy.
Dr William Cox is a founding partner at Management & Excellence, a global partner at Hiqo Solutions and a global partner at All Scorings. He has consulted more than 50 blue chip companies and institutional investors over the course of decades on the financial impacts of corporate process, ESG and AI. He received his PhD from the London School of Economics and other degrees from Oxford, Boston and the Harvard Kennedy School.
