“40 per cent of jobs will be lost to AI”. That was the sort of headline inadvertently prompted by a report by Oxford University academics on ‘The Future of Employment’ which examined the coming impact of Artificial Intelligence on the workplace.
Whilst the researchers later clarified that the reality of AI’s growth would be slightly less apocalyptic, the consensus remains that more than a third of jobs will be at risk of significant automation by the mid-2030s. Regardless of the exact numbers affected, we may currently be living through a period which will enter the history books as the start of the fourth industrial revolution. What will this mean for the role of humans in decision-making in the workplace?
The transition to new manufacturing processes in the first industrial revolution saw machine replace muscle, with machines doing the heavy lifting previously undertaken by human - or animal - power. The second and third revolutions saw electrically-powered flow production and computerisation transform the workplace respectively. Some people believe we are in the midst of a fourth industrial revolution – one defined by replacing brain power with machines.
That may prompt a bit of existential angst, but let us focus in this discussion on the impacts for knowledge workers.
A family doctor or general practitioner is an exemplar knowledge worker, building on long theoretical training with daily cases adding to their bank of experience and insights. However, they are limited in what they can know by their training, patient interaction, and outside study.
“Thinking” machines on the other hand have an outsize advantage. They can absorb data from thousands, even millions, of case notes without flagging. Moreover, those case notes can track a patient well outside of the normal scope of a family doctor, taking in the outcomes of downstream diagnosis and treatment by specialists. Like humans, AI can develop biases but, unlike many of us, AI algorithms will actually change their predictions as the information changes.
The transition to automation is already well underway. In the medical field, Google Health’s Streams app is exceeding – in certain contexts – experienced consultants’ ability to assess multiple streams of data and make effective predictions. It can of course do this tirelessly, round-the-clock.
AI is undoubtedly encroaching on many jobs. Professional services firms, for example, are worried as they see AI pick up the fee-generating tasks historically carried out by the base of their pyramid. However, AI is simultaneously opening up many new opportunities for human experts.
The advantage of appropriate AI
Businesses implementing AI well are already benefitting in areas that span the entire value chain. From improved customer experience and reduced agent costs, to more-insightful analysis that can support management decisions, the impact that AI delivers to businesses today can be game-changing. Those late to automation will lose out against their competition.
Consider the example of a mobile network customer. An unrealistic usage value can now be spotted and diverted by AI for human expert review, which means that the customer isn’t landed with an outrageous bill. Not only will this save the time of the service agent, it will also avert the bad press from a negative Twitter complaint. In this case, AI can help maintain a satisfied customer, and prevent resources having to be diverted on rectifying downstream errors.
Management teams will increasingly rely on AI to do the analytical heavy-lifting. Exception reporting won’t need to be driven by thresholds set by management. Instead, AI can learn to recognise what is considered ‘normal’ and ‘good’, and flag early warning signs when something diverges from the norm. By using AI in this way, management teams will be able see what the road ahead looks like and make predictions and corrections accordingly.
Complementing not compromising
AI excels at performing tightly-constrained, bounded, tasks. However, it often falters on more complex activities that humans find straightforward. When faced with complications, the human brain can reason outside the parameters of a given situation.
One reason driverless vehicles haven’t been rolled out extensively is the need for drivers to synthesise multiple inputs and respond appropriately to uncommon situations. To succeed in this area, machines would need to be able to react to the variables around them in the same way a human driver would e.g. “the person currently standing safely on the pavement looks distracted and is about to step into the road”.
A typical corporate example would be a business that is moving into a new market or product area. Past performance of established business units may be a guide, but a predictive algorithm built on it will be a far from infallible guide to future success in the new ventures.
Overseeing such challenges means that the need for human intelligence won’t disappear anytime soon. In fact, the demand for human expertise should actually grow as companies utilising AI free up more time to focus on the “value-add” activities. As times goes on and technology takes up more and more of the grunt work of analysing large data sets and providing guidance, more opportunities will come up for human input to synthesise these predictions and define the next action to take that supports the firm’s strategic and operational goals.
A ‘gotcha’ for all of this is that algorithms will reflect biases within the given data sets. For example, there is evidence that algorithms used during recruitment processes have treated applicants less favourably based on their gender or ethnicity because of historical decisions taken by recruiters and line managers. Flaws like this are a reminder that predictions from an algorithm should be taken as guidance, not fact, and sometimes could be simply wrong.
The role of AI in the workplace, therefore, will largely complement and augment the human decision-making process. Managers don’t need to waste time sifting through available analyses; instead, they can take advantage of AI-generated insights, with algorithms flagging potential issues and suggesting follow up action. As such, AI is not about relying solely on the outcome of algorithms to make decisions, but rather using it to guide us towards greater innovation while protecting us from costly mistakes. Whilst AI may not always get it right, businesses that use it properly will certainly reap the rewards.
Xavier Fernandes, Analytics Director, Metapraxis