Your Employees Know How to Prompt AI. But Do They Know When Not to Trust It?
Is your team AI-ready? Discover the importance of AI literacy vs. fluency in the workplace.
Article
The management team of a growing service business has decided that its employees need to become more confident using artificial intelligence. The company has introduced several approved tools, employees are beginning to experiment with them and leadership wants to make sure the workforce has the skills to benefit from the technology.
A training provider is engaged, a programme is organised and employees attend sessions covering prompt writing, document summarisation and some of the more useful features of modern AI assistants.
The feedback is encouraging. People enjoy seeing how quickly the technology can prepare a report, organise information or turn rough notes into a professional-looking document. Several employees immediately recognise opportunities to reduce repetitive work, and managers begin discussing how the tools could improve productivity across different departments.
By the end of the programme, the company can report that a substantial proportion of its workforce has completed AI training.
A few weeks later, an employee in the operations team uses the newly approved assistant to prepare a weekly performance report. The system reads the information provided, produces a clear summary and identifies what appears to be a concerning decline in customer service performance.
The employee checks the language, adjusts the formatting and forwards the report to their manager, confident that the AI has done exactly what it was trained to do.
During the management meeting, somebody challenges the conclusion. The underlying figures don't actually show the deterioration described in the summary.
One measure has changed because the business altered how it categorises a particular type of enquiry, and the AI has interpreted that difference as evidence of declining performance. The report looks convincing, but the explanation isn't supported by the information available.
The employee followed the approach demonstrated during training and produced a professional document in much less time than before. What they hadn't been prepared to do was recognise that the system's conclusion needed a different kind of scrutiny.
I think this illustrates a question businesses will increasingly need to answer. When we say our employees have been trained to use AI, what exactly are they capable of doing as a result?
Can they produce a useful prompt, or can they also recognise when a model has misunderstood the information, made an unsupported assumption or generated an answer that isn't appropriate for the work?
I was thinking about this while reading BCG's September 2026 report, *BALANCE: Thriving in the Age of AI*. The report distinguishes between AI literacy and AI fluency, proposing different levels of capability as employees move from understanding and using the technology to designing, managing and supervising more sophisticated applications.
I think that's a useful distinction. Becoming comfortable with an AI tool is a worthwhile starting point, but it isn't the same as developing the judgement needed to use it responsibly in an operational setting.
## Being able to use AI isn't the same as knowing whether it's right
Let's return to the operations employee preparing the weekly report. The AI assistant has successfully completed several visible parts of the task. It has organised the information, produced readable commentary and reduced the time the employee needed to spend drafting the document. From the perspective of basic tool use, the interaction appears successful.
But the real purpose of the report isn't to produce well-written paragraphs. It's to help managers understand how the business is performing and decide whether anything requires attention. A report that looks professional but reaches an incorrect conclusion may actually make that decision more difficult, particularly when the presentation gives readers confidence that the information has already been analysed properly.