A few years ago, while serving as executive director at an international trade and investment services firm, I noticed that one of our graduate interns was consistently completing tasks more quickly than usual. A tedious task that typically took a week was done in a day. Turns out he was using various AI tools to handle most of the heavy lifting — without seeking anyone’s permission.
This scenario is occurring in many companies worldwide, and it poses significant risks to businesses.
The AI Landscape Today
AI consists of machines and systems that imitate human reasoning and actions. In business, AI has become a transformative tool that enhances efficiency, automates repetitive tasks, cuts costs, and boosts productivity. A 2023 IBM report shows that AI-powered chatbots reduce customer service costs by 30%. The market for Generative AI is projected to reach US$1.3 trillion by 2032.
Risk 1: Data Privacy and Information Security
Most publicly available generative AI models are “black box” models. When employees upload business plans, presentations, or internal documents for AI review, there is a risk that a company’s intellectual property and trade secrets could become part of the AI model’s memory. Attackers have tricked some LLMs into revealing sensitive training data or reconstructing user information.
Risk 2: Outdated Information
Most free public LLMs are trained on data with a cut-off date. In fast-changing sectors such as healthcare, finance, and law, relying on AI to draft policies or create forecasts can lead to misleading decisions and regulatory issues. An HR officer depending on a free LLM to draft an employment contract might overlook recent amendments to labour laws.
Risk 3: AI Hallucinations
LLMs can confidently generate information based on “facts” that do not exist. They may cite nonexistent papers, refer to laws that don’t exist, or quote fabricated figures. In the AI community, we say: “The danger is not that AI will lie, but that it will do so with absolute confidence.”
Risk 4: Bias and Prejudice
AI systems can harbour inherent prejudices based on their training data. A bank’s AI system predicting loan defaults could unfairly classify applicants based on demographics rather than creditworthiness. The potential risks from bias could be especially significant for African companies, given our region’s lower representation in global data repositories.
The Bottom Line
As Derek Bok, a former president of Harvard, once said: “If you think education is expensive, try ignorance.” The risk of not being an AI-enabled business far surpasses the entire spectrum of AI risks. AI will reshape the global economy, and businesses without a core AI strategy will face strategic and competitive disadvantages.
