Robin Bordoli is Chief Executive Officer of leading machine learning company, Figure Eight. He has spent the past two decades helping high-growth technology companies launch and scale platforms and products into rapidly transforming markets. Here, he gives a glimpse of what a workforce enriched by artificial intelligence (AI) may look like.
We know AI is driving step changes in business. In manufacturing, for example, it’s improving productivity with increasing automation and the arrival of the "digital factory." It’s also enabling innovation in such areas as the automotive industry (with autonomous vehicles) and in healthcare, where some commentators believe AI has the potential to improve accessibility, reduce costs and administration, and revolutionize diagnosis and treatment.
AI is blurring the dividing line between human and machine. Humans create training data that teaches machines to be smarter and in turn, machines are increasingly able to replicate elements of human intelligence and learn more about the human experience. In the workforce of the future, then, humans and machines will be learning about each other simultaneously.
Some fear that AI will result in machines replacing people altogether, which is understandable but inaccurate. Crucially, AI is about enabling machines to augment humans — in other words, to complement human skills — rather than merely replace them. In most cases, AI’s skills are different than a human’s, so we are moving towards a symbiosis of the two.
Despite all these opportunities, companies face significant risks if they do not implement AI successfully. It’s critical for organizations to work out where human and machine intelligence can each add the most value. Firms need to get this balance of strengths right from the start, so they can design business processes that make the best use of both resources. This will help empower their employees to work smarter and more productively.
Companies may choose to implement a lot of AI or only a little. Whatever the level of implementation, they must ensure workers and AI are comfortable with each other. Ideally, they will help employees understand and feel comfortable about how they fit and work alongside the machines. This includes getting them to trust — rather than fear — the technology that’s involved.
Employees need to feel they are retaining enough control. It’s important for them to realize that, depending on the business process involved, humans still play a key role in making final decisions and are accountable for the consequences. In some instances, machines can’t confidently make the right decision. That’s when the process calls on humans to take the next step, applying their personal judgment and experience.
The US Postal Service is a good example. It’s been modernizing mail sorting for decades and has now automated 99% of the process. However, that 1% — when a machine cannot be confidently read an address — requires a human to step in and accurately sort the mail. Given that the US Postal Service handles 500 million pieces of mail a day, some 5 million items per day might not get delivered correctly without human intervention.
This “human in the loop” concept demonstrates that intelligent systems can only go so far in automating some processes. At the same time, it enhances training data every time it is used. So, if a human corrects a problem, the machine learns from that correction in a way it wouldn’t be able to without human input. This means the overall model gets more intelligent with every iteration, and the process gets more effective. It’s this iterative process of active learning that is so important in training, testing, and tuning machine learning models.
Let’s say a process is 100% human — categorizing technology support tickets through human-to-human conversations, for example. A firm might start to automate parts of this task, and as the “human in the loop” element improves processes, the machine can handle more and more work. This also means the firm can redirect more work through that business process, making it more productive.
To achieve an optimal hybrid model of machine-human interaction, companies will need to look at redrawing organizational structures and rewriting or redefining job descriptions. To this end, demand for skills such as creativity and critical thinking is already on the rise.
AI will even invent completely new jobs, which will need to become part of this mix. Gartner's Predicts 2018: AI and the Future of Work report says AI will create more jobs than it will eliminate, notably in healthcare and the public sector.
Accenture has even identified three new types of employment in the future of work: “Explainers,” who will be needed to interpret the output of AI systems; “Sustainers” to optimize their effectiveness; and “Trainers” to feed AI systems’ capacity for judgment. In a separate study published in the MIT Sloan Management Review, Accenture’s team even gave some potential new jobs titles. For example, a Context Designer will “design smart decisions based on business context, process task, and individual, professional, and cultural factors,” while an Automation Ethicist will evaluate “the non-economic impact of smart machines, both the upside and downside.”
Employees need to see the benefits of AI for the company and for themselves individually, in terms of learning new skills relevant to the digital age. Learning is likely to become continuous, partly because of the speed at which technology is developing. As a result, organizations will need to develop more flexible structures and workplace philosophies, so employees can continue to learn and develop on the job.
This means business leaders must plan short and long-term training for all teams — not just for those in IT — so everyone is part of the AI-driven world of work. This might involve deploying user-friendly, AI-powered training applications that have already been created for this purpose.
Regular training should include keeping employees up to speed on regulations that govern how data is managed, stored, shared, and used. Breaches of rules and regulations can cause customers to quickly lose trust in an organization, which can be commercially fatal if it goes too far.
This article is part of our series on the impact of Artificial Intelligence on business. For more insights, check out this piece on 3 reasons why AI will boost US productivity by 35%; this discussion with Peter Norvig on how AI can deliver the 'magical difference' in customer experience; and this piece on managing the integration of humans and AI in your workforce.