Buyer Beware: The Risks of AI in Business Intelligence

Organizations are increasingly harnessing the power of the latest AI algorithms by adopting privately-held AI models to better align their business strategies and fuel growth.

The distinction between private and public AI is crucial in this scenario. Many businesses are understandably cautious about granting public AIs access to sensitive information such as HR records, financial data, or operational details.

When an AI is trained on specific data, its output becomes more pertinent, aiding decision-makers more effectively in strategizing. Utilizing private reasoning engines is a logical approach for companies to achieve optimal AI performance while safeguarding their intellectual property.

By leveraging enterprise-specific data and fine-tuning local AI models, companies can provide bespoke forecasting and operational adjustments that reflect the daily realities of their operations. A Deloitte Strategy Insight paper refers to private AI as a "bespoke compass," emphasizing the competitive edge gained through the use of internal data. Accenture suggests that AI is set to bring about a major economic transformation akin to the agricultural and industrial revolutions.

However, there's a risk that, much like traditional business intelligence, relying on historical data collected over years could trap decision-making in past patterns. According to McKinsey, companies may find themselves "mirroring their institutional past in algorithmic amber." The Harvard Business Review highlights the technical complexities of customizing AI models to make them more relevant, warning that it might be a task suited only for those highly proficient in data science and programming.

MIT Sloan offers a balanced view, suggesting AI be treated as a co-pilot in business strategy, with continuous verification and questioning of AI output, particularly when high-stakes decisions are involved.

Decision-makers considering adopting AI in a private and cautious manner should scrutinize the motivations behind advice that strongly advocates for AI adoption.

Deloitte, for instance, develops and oversees AI solutions using custom infrastructures like its factory-as-a-service, while Accenture offers dedicated AI strategy practices, partnering with major platforms to create custom solutions. Both companies have vested interests in promoting AI as a transformative force.

Although inspired by grand claims like "the most significant change to work since the agricultural and industrial revolutions," businesses must consider the commercial interests behind these statements.

Advocates for AI generally highlight its superiority in recognizing trends and statistical patterns, a task at which AI excels compared to humans. With the massive amount of data available to businesses today, AI's ability to efficiently process information is a significant advantage. By eliminating the need for manual data analysis, AI provides actionable insights more quickly and accurately.

AI's ability to understand queries in everyday language and make data-driven predictions is particularly advantageous for organizations employing private AIs. Even team members without specialized skills in statistical analysis or database management can gather insights that would traditionally require input from various departments. This time-saving measure allows companies to concentrate on strategy rather than on assembling data points and manually querying information.

Nevertheless, McKinsey and Gartner caution against overconfidence and data becoming obsolete. Historical data may not be suitable for current strategies if it's too dated. Overconfidence in AI can stem from uncritically accepting AI outputs or posing poorly phrased queries without further investigation.

For any software algorithm, broad instructions like "base your findings on our historical data" can be interpreted differently, unlike more specific commands such as "analyze the last year's sales data while ignoring significant outliers, yet marking them for review."

Organizations might complement private AI solutions with established, robust business intelligence platforms. While SAP Business Objects has been around for almost 30 years, and SAS Business Intelligence has been in use since before the internet's heyday, even newer entrants like Microsoft Power BI have over a decade of experience in real-world business analysis. Therefore, private AI should be seen as an enhancement to a strategist's toolkit rather than a rapid replacement of traditional tools.