Sage Group, a company specializing in accounting and financial tech for businesses, temporarily disabled its Sage Copilot AI assistant after it mistakenly shared customer financial data. A customer discovered the issue when they asked the AI to display recent invoices and received information from other accounts, including their own. The customer promptly reported the problem, and a Sage representative confirmed the early-access AI was taken offline for several hours last Monday to resolve it.
Despite this, Sage appeared to downplay the incident. In a statement, a spokesperson referred to the error as a “minor issue” affecting only a “small number” of customers.
This comes hot on the heels of several stories about other AI-embedded tools sharing private emails from the CEO or company salary information to anyone who asked. These may all be teething problems for organizations, but while everyone is calling this, the year of AI, in many ways, it may feel like a step back to move forward. We still have a lot of tools that are not helping as much as we expect or creating unverifiable information, or in some instances where the AI is “hallucinating” and just coming up with false information.
Like all technology, this will go slow, slow, slow and then very fast. This is not a reason to avoid using AI tools. The challenge lies in the vast array already available and determining where to start. It is also difficult for many organisations who immediately shutdown any tool that may open up the company to risk. In reality, employees will continue using these tools—they will simply find alternative ways to access them. Here are four things your organisation can do today:
- Buy the 20 euro a month ChatGPT tool (or an equivalent). If you want to test one of the most amazing inventions in modern times, it is worth paying for it, especially for those staff members who are likely to learn from it and impart this information to others. (fyi you gain access to GPT4 not GPT 3.5 with the paid version).
- Talk to your competitors, suppliers, colleagues to find out what they are using – it is impossible to keep up. For now, it is not always about finding the perfect tool; it is about getting used to using these tools to augment your productivity/ efficiency.
- Implement small AI pilot projects. If something is not going to work, find out quickly and move on, we cannot treat these tools like large software implementations.
- Identify Strategic Use Cases: like any tool, don’t start with a solution such as “let’s get a chatbot.” Begin by pinpointing areas where AI can add the most value.
We will continue to hear stories from companies like SAGE that are surprised by how independently their AI operates. However, don’t let these stories deter you from exploring one of the most amazing inventions now ready to be unleashed upon us all.
