Companies these days are discovering that their employees are more productive when they use AI assistants in their day-to-day work. According to Accenture, 40% of all working hours can be impacted by large language models (like ChatGPT) “because language tasks account for 62% of the total time employees work, and 65% of that time can be transformed into more productive activity through augmentation and automation.” Furthermore, researchers at OpenAI, alongside Daniel Rock of the Wharton School, found that “with access to an LLM, about 15% of all worker tasks in the US could be completed significantly faster at the same level of quality.”
Such productivity, of course, usually proves to be a competitive advantage in business. In addition, it comes with some major socio-economic impacts. So, while few analysts would question the importance of companies having a workforce that is trained and capable of getting the most out of LLMs, the problem is that nobody seems to know exactly what that means or how to do it.
For the moment, AI is mainly being used in business in the form of generative assistants that help carry out administrative tasks, from writing a generic email or form letter to making a presentation or putting together a spreadsheet. These capabilities come from the likes of Microsoft’s Copilot and Google’s Gemini. Though this is helpful, the reality is that this technology and what it does for workers is barely more sophisticated than the “office automation classes” of a couple of decades ago.
But the workforce is still using AI, however, basic it may be and thus, as a result, many companies mistakenly believe they are providing on-the-job training for their employees in the use of AI when all they are really doing is upskilling them with more or less glorified versions of prompt engineering. There is more to know than four basic recipes for how to ask a generative assistant to do something. And while to someone lacking experience in AI learning, prompt engineering may indeed seem like magic or rocket science, it’s not. Prompt engineering is easy and will actually be quite unnecessary soon. As Rick Battle and Teja Gollapudi of VMware point out, algorithms are best left to develop and optimize the prompts themselves.
What business leaders and professionals need to remember is that Big Tech wants us to see AI as a highly complex dark art, best bought from them. The truth is, however, that the entry barriers to AI are much lower than Big Tech would have us believe – meaning that your company’s competitiveness does not lie in being able to make the best possible use of services provided by Microsoft or Google but in creating and developing proprietary algorithms that are trained with your own data generated by your business activity. Don’t be fooled.
The job of leaders across industries is to help employees stop seeing magic where there is only technology.
So, considering this, it is essential to design an AI training program that goes beyond superficial knowledge. The training should focus on helping employees learn how to differentiate between the actual capabilities of AI and the “magical” perceptions often associated with the technology. By providing a clear understanding of AI’s possibilities as well as its limitations, leaders can enable the workforce to make informed decisions, increase their productivity in a meaningful way, and drive innovation within the company.
To achieve this, the training should begin with basic machine learning concepts and developments – presented in a manner that does not require extensive programming knowledge. This approach will allow employees to grasp the fundamental principles behind algorithms, to learn how to visualize the importance of data, and to learn the basics of statistics, or at least refresh what they learned at school or college.
From there, training can progress from machine learning to generative algorithms and this is simply a matter of understanding the possibilities of LLMs. It should also cover the role of dimension in AI, meaning how to optimize the various parameters and structures within AI systems so as to achieve the best performance from them (because while their capabilities may indeed seem impressive to us, their performance could actually be much, much better.) By understanding these concepts, employees will be better equipped to conceptualize AI applications and understand what is realistic to ask for and what is not, and, for example, be able to redesign products or services with more competitive features.
Employee training should also focus on the differences between various AI development approaches, for example RAG, LoRA, thin wrappers, and the complete development of a model. This knowledge will help employees understand the intricacies of AI, including its limitations and potential for error, and methods for enhancing performance.
This may seem intimidating to some workers at first but, by moving beyond simple prompt engineering and focusing on the fundamental principles of the technology, organizations can cultivate a workforce that has the skills to adapt and thrive today, in a workplace that is driven by AI.
Again, throughout the training process, it is crucial to emphasize that AI is not magic, but rather a technology that must be understood and then harnessed. This is why that first step of demystification is so important. Empower employees to know more about the technology they use, and foster a culture of innovation by giving them the tools necessary to make well-informed decisions. Also – and perhaps this will be difficult for some organizations – let employees and teams make mistakes in their journey of learning how to use AI effectively. This means you need to train workers to be more than prompt-writing machines.
Leadership plays a critical role in driving AI adoption and training within an organization. Thus, support and buy-in from the executive management is important when training the workforce to appropriately use generative AI. Resources must be allocated – for the technology and for the training – and priorities set. Furthermore, executives must walk the talk and actively engage in AI learning themselves. Showing that they too take the time to learn about how to use this technology will set the tone for the entire organization.
The ongoing AI revolution is a perfect illustration of Arthur C. Clarke’s third law: “Any sufficiently advanced technology is indistinguishable from magic.” The job of leaders across industries is to help employees stop seeing magic where there is only technology, so they can understand how it works and what they can expect from it. That’s all there is to it.
© IE Insights.
