How human-AI hybrids will change work forever

In early 2019, US-based artificial intelligence research firm OpenAI published details of its natural language processing text generator GPT-2.

The AI model demonstrated an impressive ability to generate contextually relevant passages of text virtually indistinguishable from those written by humans. In fact, it is so effective that OpenAI declined to release the full version of GPT-2 because it was concerned about its technology being misused. 

Whether it is chatbots or facial recognition, medical diagnosis apps or digital assistants, AI is infiltrating and shaking up our lives. The result is a brave new world of human-machine interaction.

It is a world which offers many benefits and opportunities for organisations and individuals. But it also poses some substantial challenges, not least in the way it is transforming the performance of work tasks and supercharging the gig economy.

AI is the ability of machines to perform cognitive functions usually associated with human minds, such as reasoning, learning, making decisions, problem-solving and even being creative.

Despite exponential gains in AI technology, humans have not been rendered entirely redundant. AI is good at identifying patterns and making predictions, but humans are still required to intervene and take responsibility for action. For example, AI algorithms fall short with the 'why' question (as opposed to the 'how'). They are not good at explaining why a certain decision is made, or performing counterfactual reasoning. These remain the realm of human intelligence.

For the time being at least, rather than replacing people completely, the competencies of humans and AI are being combined in the form of human-AI hybrids; intelligent systems comprised of both artificial agents - software enabled by machine learning algorithms - and people.

In radiology, for example, machine learning technology helps to alleviate the shortage of specialised radiologists by processing radiological images and identifying high-risk cases. These are then referred to the expert radiologists for diagnosis.

Or in investment banking, artificial neural networks, a type of machine learning technology, are used to predict the performance of different financial products. In both cases the volume of data being processed is too vast for the human mind to handle at the speed needed to produce the desired results.

Inevitably, AI technology will radically change the way people work over time. But this process is being dramatically accelerated by the development of a new generation of digital platforms. These platforms create greater possibilities for human-AI hybrids to customise and improve the user experience, facilitated by AI technologies, and fuelled by data collected via the platforms from past and current user interactions.

The disruptive potential of human-AI hybrids, enabled by digital platforms, is clear from the development of digital labour platforms.  Firms like upWork, freelancer, Fiverr and Toptal are transforming work through the virtual outsourcing of a wide range of organisational tasks. The effectiveness of these platforms is assisted by a number of factors which facilitate the ability of AI-human hybrid technology to disrupt task performance.

One is the automated generation of data. For example, the increasingly sophisticated Internet of Things generates a trove of task-related data that can be used to feed AI learning algorithms and generate new ideas and services. Equally, ease of online data storage and lack of constraint encourages people to generate and store more data than they used to.

How will AI and Humans work together?

Innovations in AI technology, coupled with increased computing power, and decreasing costs of technology have led to substantial advances in AI learning methods, whether that is supervised learning or unsupervised learning, reinforcement learning or neural networks.

A third factor is task modularisation. Most digital platforms adopt a modular architecture with digital modules - think Google Maps and other apps - designed and developed on a plug-and-play basis, so that they can easily be discontinued or continued, removed, added, adapted or improved at relatively low-cost and with low technical complexity. 

This modularity does not just operate at the level of the technology. It also applies to tasks. Just as apps are becoming modularised, so too are the tasks associated with those apps. Work becomes increasingly more granular. This makes it much easier for businesses to outsource elements of their work to human-AI hybrids able to perform that task more efficiently and effectively. 

Together, human and artificial agents are able to perform tasks at greater speed, scope and scale than people can do on their own. They do so in three ways: task substitution, where AI substitutes for humans; through task augmentation, where humans complement one another; or via task assemblage,where AI and humans are dynamically brought together to function as an integrated unit.

Task substitution is the most commonly encountered via ubiquitous virtual assistants. While with task augmentation human designers and engineers in manufacturing, architecture, and engineering, input design information into AI-powered generative design software to quickly produce a range of design alternatives. Or humans might augment software by applying a moral perspective to ensure AI delivers task objectives within certain acceptable ethical boundaries.  

And, although task assemblage is yet to make inroads on digital labour platforms, people and AI agents are already working as integrated units elsewhere. For example, to carry out minimally invasive surgery, or in manufacturing where humans wear robotic devices to enhance their performance.

These three factors help to create the conditions in which human and AI agents are able to reshape the way work tasks are performed. Understandably, the disruption of task allocation and performance, using human-AI hybrids, raises a number of important issues for both organisations and society more broadly.

Perhaps the most obvious implication is that in many areas of work - including those previously thought beyond the reach of robots - workers will be in constant competition with software agents. This will drive people to reskill and gain more specialised knowledge as they jostle for position in the world of work or else risk being outcompeted by more specialised human or even AI agents.

More broadly there is the challenge of managing the tension between the dominance of digital platforms and the empowerment of workers. At the moment there is effectively a digital platform arms race where the ammunition is data and the weapons are AI technologies.  

Powerful digital platforms like those owned by Amazon, Google and Apple make people's working lives easier. They provide a cloud infrastructure for convenient data accessibility and storage, as well as many freely available tools and ready-made algorithms to be used in work activities. This might be seen as democratising digital innovation, providing software tools that businesses might not have access to otherwise, for example.  

At the same time, though, these platforms are amassing more and more data and improving their AI capabilities. As a result, these platforms are gradually monopolising task substitution, augmentation and assemblage. The more dependent businesses are on these platforms the more value these platforms are able to capture. And the more they will squeeze out freelance service providers in the gig economy.

At some point, a backlash is likely. Policymakers are already focusing on dominant digital platforms with a view to reining in their power.

In the meantime, though, it seems that those individuals able to harness the potential of AI to optimise task performance will be the individuals most likely to thrive in a world of digitally disrupted work.

Further reading:

Rai, A., Constantinides, P. & Sarker, S., 2019. Next Generation Digital Platforms: Toward Human-AI Hybrids. MIS Quarterly. 43, 1, p. iii-ix.

Constantinides, P., Henfridsson, O. & Parker, G., 2018. Introduction—Platforms and Infrastructures in the Digital Age. Information Systems Research. 29, 2, p. 381-400.

Panos Constantinides was Associate Professor of Digital Innovation a at Warwick Business School and lectureds on Strategic Global Outsourcing and Offshoring on the Distance learning MBA and Global Outsourcing and Innovation on the MSc Management of Information Systems & Digital Innovation. He is now Professor of Digital Innovation at the University of Manchester.

For more articles on the Future of Work sign up to the Core Insights newsletter here.