AI has been adopted by businesses at pace, but it’s been partly driven by FOMO. With every competitor seemingly getting on board, nobody wants to miss the train. The promise of productivity benefits and the opportunity to cut costs is too tempting, even if businesses don’t have an AI strategy to speak of, or wholly understand how best to implement it.
At the same time as AI is being adopted apace, however, discontent is brewing among the workforce. People not only fear being replaced by AI, but also increasingly have their own experiences with and distrust of AI tools. In this climate, could the adoption of AI by businesses be fomenting a quiet revolution among employees—and ultimately undermining the potential benefits AI offers?
Old dog, new tricks
AI isn’t as new as it might seem. The tools we use today (both large language models like Claude and those built into other software) are just the latest developments in a field that’s existed for decades.
What we used to call machine learning—where computers were trained in pattern recognition for things like sorting items on a production line or analysing CT scans—has now been applied to language and image generation, using those learned patterns to assemble content from a vast library of similar examples.
What has changed is the range of applications this is applied to. The accessibility of LLMs like Gemini, Claude and ChatGPT means that the barrier to entry for things like content creation or basic research is lower than ever. The role Google played in finding information has been simplified even further.
Instead of needing to click through links to read original sources and interpret information, LLMs now do this for you. Instead of researching a topic and writing an article, an LLM can assemble an original-sounding article for you from similar pieces it has ‘read’, without you ever needing to explore or even understand the topic.
Smoke and mirrors
There are obvious practical benefits to this. Not every task requires particularly deep thought or focus. Many of the jobs people do (or did) on a regular basis involve less interpretation of data, and more mundane and rote copying, pasting or formatting.
Plenty of tasks we do in office-based work take time not because they require the application of certain skills, but simply because they require manual effort. Nobody would argue that colouring fields in a spreadsheet or applying heading styles to a document is a particularly creative endeavour.
At the same time, there is a general misunderstanding of the capabilities of AI. There’s an old saying that any sufficiently advanced technology is indistinguishable from magic, and that appears to apply to AI.
The ability to produce an image or video from a simple prompt, or get a comprehensive-sounding answer to a question in seconds, has hoodwinked many people into believing that AI is broadly infallible. Yet the reality is that AI is inherently and irrevocably fallible. The very nature of the technology means it can never be 100% reliable.
How AI really works
The best way to think about LLMs like ChatGPT is that they are an advanced version of predictive text. An LLM doesn’t ‘know’ or ‘understand’ anything; all it looks at is context. If you ask it a question, it looks at that string of words, and gives you an answer that corresponds with answers in its data bank that have been given for similar questions.
That doesn’t mean the answer is right, or that it’s actually researching the question that’s been posed. It’s just assembling a string of words that seem to fit together based on all of the combinations of words it knows.
This is why AI not only frequently gets things wrong in its answers, but also hallucinates information, and even lies and excludes information. However many guard rails the companies put up to prevent it from saying certain things, it isn’t applying any actual intelligence or rational thought to the answers it provides.
All it’s doing is assembling words in an order that makes sense based on what it’s read, wherever those words might have come from. Whether that’s from Wikipedia, Reddit or Twitter makes very little difference in the AI’s ‘mind’; it’s just regurgitating words that it thinks you want to hear.
The dangers of unchecked AI for businesses
What this has resulted in, in many cases, is employees actually being less efficient with AI than without it. We’ve talked before about the idea of ‘workslop’: low quality AI work that then has to be corrected, requiring more time and effort than just doing it properly in the first place. This assumes that the errors are actually caught.
Plenty of AI work goes unnoticed until it is published, at which point it can impact the business in some subtle and less subtle ways. If it isn’t outright noticed by customers, it might end up on your website or blog, and subsequently be detected and penalised by search engines, or end up misrepresenting the tone of voice and messaging you are aiming for.
This is all a worst case scenario. There are areas where AI can legitimately save time, when used with a full appreciation of its limitations. Local models with strict guardrails can be used to complete certain repetitive tasks without sending data to foreign servers, and produce results without inventing sources or silently changing figures in a spreadsheet.
But doing this requires a level of understanding and investment in AI strategy that ignores many of the biggest players in the AI space, and implements it on a much more limited scale than most organisations feel they require to be competitive.
Most importantly, it also benefits businesses to implement it in a way that accentuates the talents of its employees, and doesn’t look to replace them. The pitch the AI companies themselves are making in large part is the replacement of labour. However much they may claim that AI can’t replace people, that is very much the subtext.
The tasks it claims to be able to complete and the speed and quality it claims to do them with have the capacity to make many low-level positions in organisations redundant. We’ve seen this come to pass already with the current high levels of youth unemployment, and the frustration many jobseekers feel with a process that now frequently uses AI to judge job applications.
Is AI worth the hassle?
All of this is anathema to a healthy working environment. It isn’t that AI can’t be beneficial to businesses, but the growing perception is that it isn’t beneficial to employees. This is something business leaders need to be more conscious of: the benefits of AI can’t just be weighed against the financial cost of implementation, but also the impact it could have on morale.
This is, in many ways, a classic case of change management. But rarely has the change being managed felt so apocalyptic, or so wide-reaching. The impact is maybe only comparable to the Industrial Revolution, or the advent of machine automation. Whether or not AI lasts, the sense is that the AI boom could lead to a period of significant job losses, and that this may have started already.
This is where leadership needs to fill the gap, and address these latent concerns. Change management training can certainly help with this, helping to address resistance to any AI tools being adopted in a way that diffuses tension. But it’s also where more typical leadership skills can come into play.
Communication is a key one. The concerns of employees don’t come from nowhere, but they may not be relevant to the kind of AI software you intend to adopt, or the way you intend to implement it. Courses like our Management Development Programme can help to build communication skills that allow you to better communicate and empathise with your team, seeing their perspectives and relaying them to management.
It’s also where your motivational and delegation skills can come into play. Identifying where AI is best deployed, who can make the most of it, and ways it can be applied to augment and enhance different people’s skillsets can all multiply the benefits of AI adoption. It’s this kind of management that can take a tool with a broad set of capabilities, and turn it into something that provides more granular and comprehensive benefits.
–
There are certainly legitimate benefits to using AI in some areas, but also legitimate concerns. The problem is when businesses embrace AI wholesale without thinking about the way it is received internally. Done incorrectly, it can not only cause undue concern among employees, but create an adversarial relationship that might undermine its adoption.
That might prompt some consideration about how heavily you invest in AI, but it’s as much a question of proper planning and leadership. Through managing the AI onboarding process and listening to people’s concerns throughout, you can not only utilise AI in a way that’s effective for your organisation, but get the most out of every employee.
