Hardly a day passes without another piece of AI slop – a long post about a historical event or a cut-up movie clip – landing on your social media feed to steal a few more precious seconds of your time.
At least those may be entertaining. Worse are the “made it big” CEOs and self-proclaimed gurus on LinkedIn who keep telling you that they’ve made an app or generated some cute image “without any expertise”.
Use AI or die, they’d say. The reality for many office workers, however, is that they have already been using AI because of these top-down mandates from the CEO to do so, but things just haven’t improved. Sometimes, they have got worse with AI work slop.
In the past six months, the hand-wringing about the returns of investment for AI has come to the fore, with many businesses questioning when they’d see real gains from the technology.
In some cases, AI has made a real difference – think of the analysis of medical scans or the search for a system’s cyber vulnerabilities. Yet, in many cases, the ROI from AI just hasn’t met expectations raised by the hype.
All this is also happening at a time when AI seems to be carrying the economy – amid talks of a bubble, the reality is that without the AI-generated investments and growth, the United States economy would have hardly grown in 2025.
And many of those losing their jobs to AI aren’t being muscled out by a faster, better (but often less accurate) machine – kicking them out are CEOs eager to cut the fat to show shareholders that AI is working.
Some get hired back to simply correct the hallucinations by AI. Others, and these are usually from companies that have learnt, get rehired to work with the AI to teach it and collaborate in new ways.
This is how things could get better in 2026, as the hype dissipates and real gains have a chance to be realised. The bubble may yet burst for some companies – see OpenAI being outplayed by Google – but from that we’ll be clearer how the technology can move forward.
Indeed, for AI to be useful, it has to be simple to use. This has been the same for any technology, from a simple PC in the 1980s to the first smartphones of the 2000s.
Yes, people need to be upskilled to learn how to make use of data more meaningfully. To begin with, they have to be open to automation (companies that already have such a foundation do better with AI).
However, common tasks like prompting an AI to generate something for you shouldn’t be harder than surfing the Web and finding stuff. You shouldn’t have to learn how to type a prompt to get basic information or content you need.
Remember those days when online search was a difficult thing? Experts used to tell people to use inverted commas and words like “and” and “or” between search terms.
Today, you don’t need such rigid Boolean search methods – even though they are still used by librarians, for example – to find stuff on the Web.
Semantic search, enabled by natural language processing and more recently AI, now gauges your intent when you type certain words to return results that are relevant.
The same will happen with AI. Experts will still be able to extract more with a complex prompt but generative AI should become as easy as Web search so people can really use it for everyday work.
At the dawn of the World Wide Web in the 1990s, I worked as a part-time cyberguide in a library to teach people how to surf the Web in Singapore. Imagine having to learn to use the Internet now.
AI will go through the same gestation, though its development is advancing faster than even the Internet. In the short span of the past year, you can already see how much AI chatbots have improved.
Throwing an AI a bunch of content to crunch recently, I found that Google’s Gemini was smart enough to anticipate my next request and automatically generate it without any additional prompts. In our short interaction, it had learnt what I was going next in a series of steps and so carried out its task without prompting.
What is needed today isn’t a guide to making generative AI content but an appreciation of what the tools can and cannot do. That requires expertise learnt from a job, not in prompting AI.
What people require is a higher level of AI literacy, by drawing on their real-world experience, to analyse what’s accurate and what’s not from their AI counterpart.
Ultimately, humans have to make the important decisions – to go with one of three images generated by AI to use on their social media platforms, for example. Or choose which market to enter first, based on the best available data sent to an AI to crunch.
In the new year, the hope is that AI can finally make real progress as a tool, particularly as a trustworthy agent that can truly expand the capabilities of a human user.
Plus, you hope businesses will realise that AI often needs to work with a human co-pilot and generate new value. Collaboration is the way forward – it’s not a zero-sum game
To be sure, some jobs will be lost. Those in the creative industries, unlike, say, doctors or pilots, will find clients saying “good enough” from an AI is good enough.
Perhaps even more professions might be affected. Will a court submission done by an law intern with AI’s help be “good enough”? Will information generated by bots be seen as good enough compared to trusted journalists digging for facts in the real world?
We’ll know soon enough. In 2026, one big question is whether all the AI will make everyone lower standards and expectations to accept more work slop and fake news, or if the search for authenticity and meaning will intensify with all the information AI now gives us so easily.
