Financial sector shows caution in using AI for code generation due to high operational risks

Despite the undeniable influence of artificial intelligence (AI) on modern society, there remains a sharp debate over whether its overall impact is positive or negative. Just a year or two ago, the internet was flooded with AI-generated images that quickly entered corporate advertising. Soon after, social media platforms like YouTube were filled with AI-generated thumbnails—visuals that initially attracted viewer interest but now increasingly cause audience fatigue and skepticism.

However, outside of consumer social media feeds, artificial intelligence holds significant potential to transform corporate operations, provided it is implemented strategically. In an exclusive interview with IOL, Paul Banting, co-founder of growthguru.ai, analyzed the common narrative that AI has become an overrated tool that does not deliver expected productivity growth.

Banting stated that the problem is not that AI is overrated, but that it is being judged against the wrong promise. He emphasized that productivity gains are real and supported by research. According to a McKinsey survey on AI adoption in corporations, 80% of respondents noted improved individual productivity, yet only 37% claimed that AI directly contributed to increased profit for their organization—a metric that has remained virtually unchanged over the past year.

Therefore, in Banting's view, it is more accurate to say that AI is not overrated, but fundamentally misunderstood. Leaders mistakenly assume that saved time automatically translates into revenue growth. In reality, companies must actively direct the hours freed up by AI toward achieving tangible, high-return commercial opportunities.

While most public attention naturally focuses on optimizing outdated workflows, Banting warns that task efficiency should not become the primary goal of the enterprise. Around 79% of executives expect AI to drive substantial revenue growth by 2030, compared to 40% today. Nevertheless, only 24% have a clear idea of where this future revenue will originate. The true objective, Banting explained, is using AI to identify urgent commercial gaps and capture high-value market opportunities ahead of competitors.

The Growthguru.ai platform was developed precisely to bridge this strategic gap, focusing on business development rather than general workflow automation. Instead of offering a single universal assistant, their platform utilizes specialized AI agents targeting specific commercial levers: Perky helps partners access relevant incentives, discounts, and co-financing from suppliers; Maya develops go-to-market strategies and co-marketing initiatives; Alex guides partners through certifications and supplier tiers; and Lexi tracks important enterprise updates, such as critical Microsoft platform releases.

Nevertheless, alongside the commercial potential lies a darker side—AI-enabled crime, where issues of accountability and legal liability remain unresolved. Banting argues that criminals using AI for financial fraud must be held responsible for their actions, regardless of the tools they employ. However, if proprietary or third-party enterprise AI causes operational or financial damage, the responsibility lies directly with the business that implemented it. Legal liability cannot simply be shifted to the model provider. While software developers must create secure and compliant architectures, the choice of implementation rests entirely with the enterprise. BCG confirmed this finding in its applied AI benchmark.

S&P Global, the credit rating agency, issued a warning indicating that the largest global technology corporations may have levels of debt related to artificial intelligence (AI) higher than currently reported values.

The agency faces challenges in monitoring certain financial transactions used to fund the expansion of AI infrastructure. According to Reuters, Mark Habib, S&P's global head of digital infrastructure, pointed out that highly intricate financing arrangements may be concealing part of these giants' financial liabilities.

Despite this concern, S&P does not currently consider it necessary to modify the credit ratings of these companies. Habib stated during a panel in London (England): 'It is safe to presume there is more debt out there that we are not seeing.'

The executive referred to interconnected financial operations where large technology companies secure the construction of data centers or provide financial support to smaller AI firms. Such agreements can complicate the determination of who actually assumes the financial risks.

Habib clarified that when an operation is not made public or does not pass the agency's risk analysis, S&P may not have full visibility over the financial commitments involved.

In July, S&P had downgraded Oracle's credit rating to BBB-, the lowest level in the investment-grade category. This change was motivated by concerns regarding the company's spending on AI and the uncertainty about how these investments will translate into profits.

The agency was also questioned about the recent depreciation of debt securities linked to AI infrastructure growth. This movement raised questions about whether investors perceive these securities as riskier than suggested by the credit ratings.

Mark Habib reiterated that, for now, S&P does not detect a substantial disparity that would warrant changes in the assessments, given that the agency's ratings encompass a long-term perspective. However, this outlook could change if the investments prove unsustainable. Factors that could intensify risks include new increases in the costs of AI companies and competition so fierce that it could limit revenue growth.