A Clever Hack to Guard Against AI Hallucinations

While large companies have been experimenting with generative AI, small ones are moving slowly. Are small businesses missing the benefits of AI chatbots such as Microsoft Copilot and ChatGPT? Or is a more tentative approach wise for small businesses? Most businesses are holding back. Roughly five percent of companies in the U.S. are using generative AI, according to a Census Bureau survey featured in a report by The New York Times. One small business is finding value and risk in generative AI. Win-Tech, an aerospace manufacturing company with 41 employees in Kennesaw, Georgia, is using ChatGPT for "writing emails to employees, analyzing data, and drafting basic procedures for the company's front office," Allison Giddens, a Win-Tech co-president, told the Times. Win-Tech faces challenges in its efforts to deploy generative AI to boost productivity. For example, ChatGPT sometimes gives "off base" responses, the company must take care not to share proprietary information with the chatbot, and there are no AI applications for boosting manufacturing floor productivity. "There's not a whole heck of a lot of use cases for the shop floor yet," she explained to the Times. Climbing the AI Value Pyramid There is no one-size-fits-all approach to generative AI. Your generative AI strategy depends on your objectives, your capabilities, the needs of your customers, and the strengths and weaknesses of your competitors, as I described in my new book, Brain Rush: How to Invest and Compete in the Real World of Generative AI. One approach to generative AI is to climb what I call the value pyramid--a set of three broad uses for AI chatbots whose adoption by companies is less common as you approach the top. The generative AI value pyramid has three levels. Overcome creator's block At the base of the pyramid are the many ways people use AI chatbots to get started on an activity--such as writing an email or a report, creating a photo or video, or coding software. By helping overcome creator's block, AI chatbots can increase your people's productivity. This is the easiest way for small businesses to get value now. However, since other companies can do the same, your advantage will be fleeting. Boost customer service and sales productivity The middle layer of the value pyramid can increase your productivity. For example, in the temporary job placement industry, generative AI can reduce significantly the number of candidates a recruiter sends to a company before finding a match. AI chatbots can also help a company to resolve customer questions much more quickly. However, such customer-facing applications have risks--most notably, the risk of sharing proprietary information with the world and hallucinations--incorrect responses to a prompt that risk a company's reputation. CEOs fear generative AI hallucinations could threaten their company's reputation. For instance, Google's AI advised people to add glue to pizza, Forbes careers contributor Jack Kelly noted. And Air Canada's AI chatbot made up a refund policy for a customer -- and a Canadian tribunal forced the airline to issue a real refund based on its AI-invented policy, Wired reported. If you can overcome these risks, the second level of the value pyramid can make you more productive in a way that is difficult for rivals to copy, giving you a more sustainable competitive advantage. Create new growth curves The top level of the pyramid is the most valuable one. I think companies aspire to create new growth curves for their business by using generative AI to help their customers grow faster. For example, Boston-based Bullhorn is striving to help recruiting firms to grow faster by making them much more productive at selling, so each salesperson can sell more, as I wrote in a June Inc. post. Focus on the first two levels for now. Protect your customer relationships from AI hallucinations Technology companies are providing solutions that could protect your customer relationships from AI hallucinations. An example is Aporia, a Tel Aviv- and San Jose, California-based AI control platform. In May, Aporia introduced "real-time guardrails for multimodal AI applications," according to SiliconAngle. The new guardrails let engineers add a layer of security and control between the app and the user. "The system detects and mitigates 94 percent of hallucinations before they reach users via chat, audio or video," reported Israel21c. Aporia says its technology works well. "Our competition is trying internal workarounds. They are over-engineering prompts. 'If there is profanity, don't answer.' Aporia offers better detection in real-time," Aporia CEO Liran Hason told me in a June 3 interview. As your company climbs the value pyramid, use AI control to protect your reputation. EXPERT OPINION BY PETER COHAN, FOUNDER, PETER S. COHAN & ASSOCIATES @PETERCOHAN