Repowering the economy with artificial intelligence | Deloitte UK has been saved
Limited functionality available
Weak productivity has long been a challenge for the UK economy. Despite the promise of a productivity boom with the dawn of Industry 4.0 technologies offering interconnectivity, automation and real-time data, the 2010s were the worst decade for productivity growth since the early 19th century.
Enter 2020 and COVID-19 has dealt a further economic setback. With a global recession and unprecedented levels of uncertainty, cost reduction was the top priority for CFOs coming into Q4 2020. However, with COVID-19 continuing to impact every aspect of work and life globally, business leaders face a tricky balancing act of prioritising exactly where to innovate and where to cut back.
So why now can they look to AI to deliver the productivity gains the UK economy so urgently needs?
While the early 2010s were characterised by digital innovation, it is the adoption of technologies, not their invention, which ultimately drives economic growth. AI has been around since the 1950s and has often been criticised for failing to deliver on the industry hype. Some analysts are speculating that the COVID-19 winter will bring an ’AI spring‘. However, the reality we are seeing at Deloitte is that AI is already a driving force helping our clients respond to the challenges of the pandemic.
The rising business adoption of AI in recent years is finally delivering a step-change in results. According to our latest State of AI in the Enterprise research, four out of five UK organisations indicated that implementing AI has made their employees more productive, improved their decision-making and made their processes more efficient. Three-quarters have also lowered their costs.
Strikingly, this was not just the big industry players: over 70 per cent of AI adopters achieved a typical payback period of less than two years for their AI projects, irrespective of their size or digital maturity. The gradual democratisation of AI, as applications become easier to develop and implement, means organisations of any size and AI ability can now benefit.
Where can AI deliver the most value?
AI thrives on automating repetitive, mundane tasks, as well as detecting patterns and anomalies in large volumes of data. Unlike humans, it does not suffer from boredom, fatigue or sickness – key attributes in a global health pandemic.
As digital interactions replace physical ones, we are seeing a major uptick in automation strategies and remote management tools across all industries. Robots, chatbots and self-service tools have come into their own, allowing organisations to more with less, from afar.
AI powered virtual assistants, for example, have enabled banks to cope with soaring demand for their online services, including surges in queries on financial support and mortgage payment holidays. E-commerce companies, meanwhile, have managed unprecedented levels of online orders by leveraging advanced AI to improve supply and demand forecasting, as well as increasing the use of automation and robotics in their supply chain, warehouses and distribution centres.
However, when thinking about AI as a driver of productivity, we should not just consider machine learning or tools that automate human processes. Given productivity is usually measured as output per hour worked, we see the greatest value of AI lies in employee empowerment – helping people do their jobs better.
The best use cases of AI not only free up employees from low-value and routine tasks, but also improve their performance in the difficult, strategic tasks that typically require the most effort and attention.
For example, with thousands of research papers being published daily in the global race to find a vaccine for COVID-19, the Allen Institute for AI has partnered with leading research groups to create CORD-19. This open research dataset allows researchers to create and apply natural-language processing algorithms and, hopefully, accelerate the discovery of a vaccine. This should make it much easier for scientists to keep informed of other laboratories’ progress, saving them months or even years by either reinventing the wheel or avoiding blind alleys. It is also hoped that automated analysis will enable AI to identify patterns and suggest hypotheses or even treatments that might otherwise have been missed.
Is historical data exactly that… history?
As COVID-19 has accelerated AI adoption, it has also the highlighted the challenges of data-driven decision making in the ‘new normal’. The events of 2020 have seen a drastic change in human behaviours. Data from the past, and the algorithms that rely on it, are no longer representative, presenting a major hurdle for those that develop and use AI. While there will be many AI applications that steer clear of this concern, there are many cases where this inflection point in our history will call into question the feasibility and reliance on these systems for decision-making.
Now more than ever, organisations need good data science, leveraging the best AI and the best human expertise to make optimum judgements in such uncertain times.
The need to align objectives
While AI powered machines are excellent at meeting the objectives set for them, setting the appropriate objective is obviously key to driving value. Where many organisations struggle is in understanding what ‘second-order’ effects could arise from setting an objective that, at face-value, was a sensible one to follow.
The recent fallout surrounding the UK’s A-Level grading algorithm, is a prime example of the impact that purely algorithmic decision-making can have, when second-order effects are not properly considered. Even though the objective was to prevent grade inflation, the algorithm ultimately led to highly biased and unfair outcomes that negatively impacted students across the country.
While this is a particularly high-risk example, the underlying challenge isn’t one that exclusively impacts public sector organisations or human-focused decisions. Even the benign use cases of AI, such as the automation of how emails are handled in a call centre, require appropriate alignment of objectives and consideration for the downstream effects of introducing AI: How do we handle the emails the AI gets wrong? If the AI processes these emails really quickly, where will the next bottleneck in our process appear?
Organisations seeking to boost their productivity should take a holistic approach when building their solutions, putting risk management and ethical considerations at the front-and-centre of their AI solutions.
Collaboration and knowledge sharing will be critical to helping organisations overcome such challenges when adopting and scaling AI. Shared learnings within an organisation or even across sectors can help spread best practice, avoid common pitfalls and accelerate value generation from these technologies. Indeed the Bank of England chief economist Andrew Haldane has argued that a technology diffusion infrastructure could be the key to boosting the UK’s innovation strengths and closing productivity gaps.
This is why we have recently expanded the Deloitte AI Institute, bringing together an international network of AI specialists to share knowledge across markets and industries on AI development, helping organisations to identify and scale AI applications more effectively.
Solving the productivity puzzle
Despite a challenging start, the new decade brings new opportunities. The pandemic has forced organisations to reinvent themselves, change their business models and embrace new ways of working. AI is finally coming of age to play a key role in this ‘Great Reset’.
Organisations which leverage AI technologies to boost operational performance and employee productivity, will emerge more resilient and best prepared to return to growth in this new unstable normal.
Kishan is Head of the Artificial Intelligence Studio, part of Deloitte Ventures, and a Consulting Senior Manager specialising in advanced analytics and data science. Kishan works across industry sectors and has experience at all stages of Artificial Intelligence development; from the evaluation of strategic choices relating to the adoption of AI, through to rapid prototyping and scale implementations of solutions that employ cutting-edge techniques. Kishan is a member of the Advisory Board of the UK All Party Parliamentary Group for AI.