AI Risk and Overspend: Evaluating high-risk finance tech for CFOs

Featured image for Ai-risk article. Shows a man walking a tight rope over a ravine.

Artificial intelligence (AI) isn’t just showing up on the finance team’s expense sheet — it has infiltrated every department’s budget. A CFO asked to approve “cutting-edge” AI-driven technology often has to fully trust in a colleague’s presentation on the new software ROI without a clear roadmap to evaluate AI risk. Already $2.3 trillion is wasted on failed digital transformations worldwide, and the stakes are higher during pivotal changes in the economy.

The burden of determining cost-effectiveness largely lands on the CFO and finance team. They are asked to juggle their own workload and research into new AI vendors and software. Sometimes it’s easier to just say yes, especially when immediate “savings” can be found in downsizing labor.

Approving AI spend without significant scrutiny can potentially increase costs. And there’s the problem of lost time, productivity, and growing risk.

Why AI is not a competitive advantage in today’s market

From finance to marketing, AI developers proclaim that these bots are the salve for all business bottlenecks. Consumers and independent organizations differ on the real impact of AI.

Around 61% of Americans have used AI in six months, but 80% prefer people over bots. The same study, from Menlo Ventures, show that the 39% of Americans who have used AI over the last six months don’t see a use for it, don’t trust it, and are anxious about privacy concerns.

They are right to be worried. The recent court case of The New York Times Company v. Microsoft Corporation et al revealed that all chat logs from ChatGPT can be preserved, archived, and released in court. These records are not protected by HIPAA, attorney-client privileges, or similar privacy laws. Risk-wise, the lack of transparency and privacy opens up organizations to significant security risk.

AI as a marketing tool also offers lackluster results. While AI writing and designing platforms claim high ROIs with polished case studies, consumers tell a different story. A study from Interactive Advertising Bureau (IAB) found that consumers associate AI with negative descriptors, such as: Fake, disconnected, inauthentic, and unethical.

But nothing is more concerning that the MIT AI Risk Repository, that lists over 1,600 risks posed by using AI technology. Some organizational threats include:

  • Increased employment inequality
  • Decline in employment quality
  • Copyright infringement
  • Unclear liabilities
  • Increased emissions and pollution
  • Potential to leak sensitive information
  • Vulnerable to cyber attacks

Short-term savings riddled with long-term costs

Technology is neutral, but a misalignment can have significant costs. We have several of examples from 2025 alone, but here are our top three:

These are all exceptionally expensive mistakes. Human error and technology misuse can account for a small segment of the problem. In many cases, the technology is not designed or documented enough to ensure security or accuracy. And revenue loss only snowballs when a company’s reputation takes a hit from the resulting PR.

Employee turnover from bad brand perception alone can tank a business. A LinkedIn study found that a company of 10,000 can expect to lose $7.6 million from high turnover caused by bad reputation. Another study found that bad press can cost 48% of small business between $10,000 and $100,000 in lost revenue.

These do not account for additional costs, such as:

  • Fees for noncompliance in case of an information breach
  • Lost time, labor, and other resources from correcting mistakes or editing substandard outputs
  • Reduced productivity and increased employee turnover due to replacing labor with AI

Preventing overspend: Evaluating alternatives

No one wants to pass up on a growth opportunity, especially when a colleague offers a detailed presentation on a new tool’s potential. Rather than sit on a staunch rejection, there are some questions CFOs can ask departments consider when proposing changing the tech stack:

  • Is this new tool Traditional AI or Generative AI (GenAI)? Traditional AI options (think Siri or Alexa) operate on a stricter set of rules. Nothing is “created”, and they operated on a smaller database, making them more environmentally efficiency and less prone to error. Opting for traditional AI variations work just as well for most repetitive tasks and enable you to say you’re AI-friendly, if that’s an important pivot your organization is making. However, it doesn’t come with the same growing list of risks as GenAI.
  • Can this be done with simple automation? Artificial intelligence and automation are not the same, and automation has been around for decades. Similar to using Traditional AI, using a Robotic Process Automation (RPA) model significantly reduces risks while streamlining most repetitive tasks.
  • What is the realistic number of cuts we can make with this technology? One of the biggest misconceptions is that you can replace labor with a bot. Even if you scale a team down from 10 to 5 in marketing, for example, the remaining 5 employees now are expected to still complete the work for 10 people. This model rapidly leads to turnover, increases long-term hiring costs, and devalues the organization’s reputation.
  • Does this software align with our customers? If investing in specific technology, AI or not, would upset or confuse your customer base because it doesn’t align with your shared values, it’s not the right software for you. Betting on acquiring new customers to offset the turnover isn’t a strategy, it’s a risk.

There’s a real fear of “falling behind” and missing out on a “competitive advantage.” However, this strategy only works to build revenue and boost costs during the tail-end of the early adoption period. We are now approaching mainstream “adoption”, although this doesn’t mean “acceptance.” For the first time, we are experiencing how this technology is used in practice on a massive scale, and the dust hasn’t yet settled.

More Insights on AI