- Microsoft launched AI assistant Copilot across 365 apps.
- Though 70% of testers report efficiency improvements, others cite inaccurate data, odd behaviors, and AI hallucinations.
- Microsoft acknowledges Copilot’s complexity and aims to address flaws through further model refinement.
Microsoft recently launched its AI-powered assistant Copilot, integrating it across popular Microsoft 365 apps to help users generate documents, presentations, and more.
However, some early adopters of the $30 per user tool are questioning its value after experiencing unreliable performance.
Is Copilot meeting expectations?
While companies like Dow and Lenovo have praised Copilot’s “tremendous efficiency gains,” others told The Wall Street Journal (WSJ) that the assistant made mistakes in Excel, PowerPoint, and other programs.
Critics reported AI hallucinations, inaccurate answers and calculations, and odd behavior like scheduling weekend meetings. Hardware company Juniper remains unconvinced about purchasing subscriptions for its whole workforce after testing Copilot since November.
Microsoft’s commitment to refinement
Microsoft’s Jared Spataro acknowledged the complexity of Copilot given the state of AI technology today. However, he stated that 70% of early users report increased productivity and 68% see improvements in their work quality.
As an AI-first company investing billions in OpenAI, Microsoft aims to work through the feedback to develop Copilot over time further.
After launching Copilot exclusively to companies committing to 300+ licenses last November, Microsoft opened smaller subscriptions in January ahead of the internal rollout this month. Though Copilot highlights the possibilities of AI assistants, its flaws demonstrate adoption hurdles.
As a work-in-progress, Copilot’s ultimate success relies on Microsoft refining its underlying AI models to address reliability and accuracy issues surfaced by pioneering users.
