In a recent episode of the Real AI Use Cases podcast, business owner Richard Gearhart shared a frustrating experience with Microsoft Copilot, describing it as a "complete and total dumpster fire."
Richard, an intellectual property attorney, attempted to use Microsoft Copilot to analyze his firm's profitability and performance metrics. He fed it financial statements, lawyer time sheets, and invoice records, expecting accurate answers. However, the AI repeatedly returned incorrect figures, even contradicting the data he had provided.
"It kept coming back with wrong answers. It was even wrong compared to the information that I put in there in the first place," Richard said. He questioned the AI about revenue numbers and received responses that did not match his financial statements.
Co-host Jessica, drawing from her own experience with Claude, noted that when she catches errors, she corrects the AI and it often improves. Richard acknowledged that his prompts might not have been optimized but felt his questions were straightforward enough for the AI to handle.
This incident highlights ongoing challenges with AI reliability, especially when handling sensitive client data. Richard emphasized the need for platforms that don't compromise confidentiality.
"I really wanted to throw my laptop against the wall," Richard admitted, underscoring the frustration many users face when AI falls short of expectations.
The story serves as a reminder that while AI tools are powerful, they still require careful oversight and may not be ready for critical business analysis without human verification.