Human advantage in the age of AI and autonomy

Generative AI is only the latest hype cycle in digital technology. Remember the cloud, big data, the Internet of Things, blockchain, cryptocurrencies, and the Metaverse?

AI boosters talk a big game, and it is true that the technology will be a game changer, shaping how people work, and even what they do in the modern workplace. But it will not be the end of human workers, let alone of the human world as we know it.

The age of AI is here, and it is allowing businesses to reshape their workforces. What was previously routine and time-consuming research, synthesis, coding, and report compilation can now be done with an AI agent in a fraction of the time, accelerating production for review by sector-experienced humans. There are many claims of significant improvements in individual productivity from AI pilots, though some gains have yet to be felt at corporate level.

In finance, investment managers now set the strategy, which is then coded, and leave the computers to make the investment decisions. Algorithmic trading has been here for some time. Specific AI tools, such as Hazel, produced by Altruist, will repeatedly challenge current tax and wealth management business models. For financial advisors and wealth managers, the underlying fear is that AI will drive down fees.

In education, the use of AI by university students is becoming a new norm and remains contentious. Students may present brilliantly curated work, but have they learned anything if algorithms have done the heavy lifting? The learn-by-doing process of research, sifting sources, and testing ideas is key to acquiring critical thinking and judgement. Education is what is left behind after a person has forgotten everything they were taught – learning how to filter facts and curate them into knowledge.

But the companion model is more likely than AI’s replacement of human workers. In engineering, AI will not replace the designer or the engineer. The IET takes the view that AI tools will be the engine, reducing hard work and accelerating draft designs. Yet it is the engineer, having used a bespoke AI-powered companion tool, who reviews and refines the draft, ensuring it is fit for purpose and attractive to people and the market. Specialist AI tools, with personae matched to the needs of the designer, engineer, and scientist, will add sophistication to design processes, shifting manufacture from Industry 4.0 to Industry 5.0.

The more likely way ahead is boutique specialist AI tools, assistants, and companions. In specific engineering or medical settings, AI lifts human performance: it enables rather than replaces.

AI is driving economic growth, primarily in the US, reflected in stock market prices and competition to build data-centre capacity. But the volume of compute needed for AI comes at a cost: financial, environmental, social, moral, and economic. Society will most likely push back as those costs increase. Studies, including by Barclays Bank in the UK, suggest AI is not yet boosting company productivity. Huge capital spend has not yet turned into profits.

Very high-capability super-compute and agentic AI will be expensive to develop and may be used mainly by governments and large commercial enterprises. The US is seeking to block non-Americans accessing Anthropic’s powerful Fable 5 and Mythos 5 models.

But a cheaper and simpler AI model developed in China could challenge US business models and investment assumptions. The cost of running AI models has risen 41 per cent since the start of the year, according to research company Silicon Data.

The contours of China’s strategy to become dominant in AI are becoming clear. It is difficult to protect expensive US large language models from competition. Deepseek and the potential challenger Kimi K3, developed by Moonshot and boasting 2.8 trillion parameters, show the strategy in practice. Kimi K3 is almost as powerful as the most advanced US models launched by OpenAI and Anthropic, while being cheaper to run. It is also free to download and modify, which could lead to Chinese-based AI becoming dominant in the global market. Alibaba may soon launch a similarly competitive version of its open-source model Qwen.

Taken together, these variables create a febrile market, with sharp shifts in stock values for chip makers and leading technology companies. We may be living in an AI bubble with parallels to the dotcom bubble. If it bursts, the size and shape of AI providers will change dramatically, but AI as a capability will still exist.

There is a common perception that AI is causing a jobs bloodbath in professional services. There has been a fall in jobs in professional services, finance, and creative industries, though how much is due to AI rather than indirect economic factors is not yet clear. It could also be transitory as AI becomes normal and human roles evolve to accommodate it.

The cumulative effect of smartphones, social media, and endless scrolling has not improved feelings of connection and community. There is widespread concern about the effect on human attention spans.

The technology-driven dystopian view of the near future is a sticky idea. But disruptive technologies tend not to become dominant overnight; they almost always take longer to arrive and cost more than most people can afford. AI adoption may be slow, steady, and uneven rather than instantly transformational.

There are valid questions around the AI investment bubble: the performance of LLMs themselves, “maths with bugs,” circular financing, and competitive threats from China. The market is fuelled by hype seasoned with anthropomorphism: things that are difficult to understand appear to be magic or scary.

LLMs may be considered giant approximation tools. They achieve remarkable mimicry by scraping and sampling data, often overlooking IP ownership. Models also sample opinion or conjecture with no foundation in truth or fact. As they sample the outputs of other models, they may repeat errors, pollute synthetic text, generate “AI slop,” and risk model collapse.

Humans commonly probe the limits of AI models with unreal or silly questions. The answers betray a lack of humour, empathy, and simple common sense.

More fundamentally, data centres need power and water at a scale that outstrips available infrastructure: substations, switching gear, and transmission lines. A single data hub in Hays County, Texas, can use 10 million gallons of water a day for evaporative cooling and power generation. On one industrial estate in Slough, UK, more than 30 data centres have been built, prompting complaints about noise, heat, and light pollution.

The illusion of intelligence can be dangerous, especially where humans lack the knowledge to recognise and filter algorithmic advice. So what enduring advantages do people possess to help them navigate, survive, and thrive in the age of AI?

In the age of AI, human judgement remains the premium product. The human advantage is the ability to apply experience when assessing context, especially under uncertainty. Smart humans frame the right questions, challenge outputs, explain and defend decisions, and adapt in real time to changing circumstances. People can routinely say “I don’t know” and “Help me understand.”  Humans choose to break rules and take risks.

Humans have evolved an uncanny capacity to fuse information from all their senses to make decisions. In the modern workplace, this allows leaders to infer intent, absorb context, and balance explicit and implicit inputs from all sources. This all-source intelligence is key to effective decision-making, especially where volatility, uncertainty, complexity, and ambiguity are the normal conditions of work.

Humans augmented by AI tools can deal with large-scale, complex problems at the tempo of relevance. The human brings discernment, taste, judgement, and critical thinking.

It is ethically and legally important to protect brand integrity in business and reputation in public life. Only human decision-makers can be made responsible and held to account. The ability to assess a situation, decide to do the right thing, and stand by that choice shows the integrity essential to trust.

People like people. They trust personal recommendations. There is a deficit of trust in AI outputs because hallucinations, truth, and accuracy are context-specific problems.

Humans may also distrust AI because they are not consuming it; it is consuming their job. Agentic AI learns from each interaction. It is designed to replicate business methods and logic and could become a substitute for the human worker. Companies will therefore want firewalled models on approved systems to protect workflows and domain knowledge.

An AI agent can emulate a personality, which for many is worrying. But a machine or autonomous AI-powered robot cannot be accountable in law, cannot create, and cannot cope with novel situations and the unknown. Customer-service chatbots have become commonplace since the pandemic, but a chatbot-first user experience can devalue customers if they get stuck in a doom loop of automated responses or if the chatbot cannot handle complex queries. Intelligent AI should improve the experience, not trap people inside the machine.

People will pay more to speak to humans for advice on important matters because they value trusted interaction.

AI agents lack deep human connection and emotion, although they can increasingly emulate it and parrot phrases. AI may generate a wedding speech, even including phrases the celebrants would not have thought of. But humans respond badly to beige, homogenised content. They respond best to sincerity, humour, originality, and even errors: to err is human.

Humans have a flair for innovation. An AI LLM often struggles to generate a genuinely new idea; it regurgitates and reuses ideas it has seen before. People have an appetite for the unusual and quirky, and that appetite can generate genuinely new fashions and trends.

People have almost unbelievable abilities, reinforced with practice and experience, to specialise in certain topics or fields. These may be knowledge-based, behavioural, or physical. Put a carefully curated group of people together with the right mix of knowledge and skills, in a team or on a board, and they can achieve remarkable things.

And finally, there are the all-rounders: the polymaths, and more commonly the people who balance IQ, EQ, and SQ, combining intelligence with emotion and social skill. Some people have huge innate relatability. Others develop the ability to coordinate multiple activities across projects and programs, using tools skilfully to coordinate humans and machines in an adaptive manner.

Jobs will change and the job market will evolve. AI-savvy people who can excel in the new normal workplace will always be in demand. Employees will still need to learn and gain experience. Data will become commonplace, but judgement, the ability to infer knowledge, distil wisdom, and apply it in context will remain the differentiator between humans and machines. Humans aided by their digital companions will win every time.

Get in touch to discuss how your organisation can make the most of AI while strengthening the human judgement, creativity, and trust that technology cannot replace.

Connect with Mark Baines on LinkedIn.