Microsoft VC warns about the dubious and increasing corporate AI claims.

• Recent earnings call transcripts of S&P 500 companies show a significant increase in the use of the term AI compared to the five-year average, according to an analytics firm.
• Michael Stewart, managing partner at Microsoft's venture arm M12, states that it is not difficult to incorporate AI into a slide presentation.

The AI industry is filled with numerous players, but not all of them are legitimate. Some of these players engage in a practice known as "AI washing," which involves making false claims to investors about the use of new technologies, as explained by Gary Gensler, the chair of the Securities and Exchange Commission, in a video.

In March, the SEC fined two investment advisers, Delphia (USA) Inc. and Global Predictions Inc., a combined total of $400,000 for marketing AI-enabled investment predictions to their clients, despite not actually using this technology. The companies paid the penalties and complied with SEC orders without admitting or denying the Commission's findings.

The phenomenon of companies using the term "AI" in their earnings call transcripts has been observed across industries, not just the financial services space. An analytics firm, FactSet, found that 179 S&P 500 companies used the term "AI" in the three months ending in mid-March, which is higher than the five-year average of 73 companies. While it is unclear which companies are being dishonest about the depth of AI technology in their operations, experts say that using the acronym has become commonplace across industries.

Michael Stewart, managing partner at Microsoft's venture arm M12, who specializes in AI, gaming, and deep tech, stated that it is not difficult to incorporate AI into a slide deck or use it because it can be done easily through any platform without having to build it from scratch. However, he noted that there is no long-term competitive advantage to doing so.

The breakdown of confidence between vendors and their consumers, enterprise partners and investors is a result of AI washing in reality.

The real risk lies with individuals on the back end not comprehending technology, as stated by Timothy Bates, a professor of practice at University of Michigan-Flint College of Innovation & Technology with expertise in AI and emerging technology, and a former chief technology officer at Lenovo and General Motors.

Bates argues that button-pushing applications are also considered AI washing, as they do not contribute to the development of AI. According to Bates, AI learns by consuming information and receiving varying inputs, or prompts. He explains that when the same question is repeatedly asked, which is essentially a button push, it does not contribute to the creation of a database. Furthermore, third-party applications that rely on generic natural language processing models are not effective in the long run.

If an AI assistant is not based on a unique, unbiased database specifically trained on law and actively learning, it may stop working in a matter of months when bought by a law firm to replace a human.

A 'good' corporate AI litmus test

Be cautious of any company that uses the term AI in a broad sense, advises Toby Coulthard, chief product officer at Phrasee, a generative AI solution for enterprise clients such as Sephora and Macy's.

Coulthard suggests businesses should define and specify their AI usage.

Coulthard advises paying attention to when a company first started discussing its AI usage. He suggests that businesses that were talking about AI before ChatGPT are likely to be more ethical in their use of the technology. Additionally, it's a positive sign if a company has a policy in place regarding what they won't do with AI.

Coulthard stated that the more a business talks about their use and non-use of AI, the more accurate their representation will be.

Bates advises examining the AI model used by the company to determine if it is self-generated or reliant on a third-party model. If the latter, he suggests reviewing the service-level agreement or key performance indicators over the next one to two years. He emphasizes that AI companies require maintenance, monitoring, and adjustments to function effectively.

Microsoft's VC approach to AI scrutiny

M12's AI-focused partners, including Stewart, evaluate startups using the four D's: data, dividends, distribution, and delight.

If a startup lacks access to its customers' crucial data that the AI will process, any competitor can obtain the same data, according to Stewart.

Identifying whether an AI's output is contributing to the bottom line can be helpful. It is possible to look at the gross profitability of any AI business, knowing that one advantage of the technology is extremely high profit margins. Stewart claims that even a fully AI company with limited human intervention can achieve 80-90% gross profitability.

The sustainability of a startup's longevity is determined by the distribution and delight elements of M12's investment analysis process, which are not directly related to AI washing, but play a crucial role in the evolution of young businesses as they mature.

To succeed in the competitive AI startup market, companies must have robust and reliable distribution channels to effectively reach customers. Additionally, providing exceptional user experiences is crucial in fostering customer loyalty and encouraging repeat business.

Unlike previous AI advancements, natural language processing technology has commercial potential and is therefore more likely to be marketed and sold. This has led to increased investment and customer interest, which in turn increases the likelihood of obfuscation.

The SEC has mainly focused on investment advisors and broker dealers, but AI washing is a concern across the marketplace. "We don't want our fund to be associated with a technology that goes beyond what's acceptable," Stewart stated. "If we've missed that the AI is not what's truly producing the magic, we need to reevaluate our entire diligence process."

In 2022, M12 erased the chalkboard and created new guidelines for analyzing AI startups after discovering that many computer vision companies were not as proficient as they claimed.

Bates stated that the excitement surrounding AI can attract investors, increase stock prices, and generate consumer interest, providing these companies with a temporary advantage. However, regulatory bodies may intensify their scrutiny and enforcement as many resources are being invested in superficial AI claims rather than tangible advancements in the field.