AI technology has been around us for some time. It is fast becoming an essential part of workforces around the world.

Attitudes towards artificial intelligence (AI) vary radically from the AI zealots to the AI sceptics. The sceptics tend to focus on the ‘threat’ that AI represents to the human workforce. Gartner research, Surviving the rise of ‘smart machines’¹ (see ‘Online resources’ for the web addresses of the reports cited in this article), estimates that, ‘By 2030, 90% of jobs as we know them today will be replaced by smart machines’. Nowadays, this kind of statement or forecast can be heard almost everywhere, even in the mainstream media, while in reality AI technology is still far away from being mature enough to entirely replace human labour. At this stage, most forms of AI technology are very far from matching human intelligence.

Nevertheless, AI colleagues (though rarely in a physical form) are already working with us in an increasing number of workplaces, and whatever the level of ‘intelligence’ AI has achieved, the disruptive changes it has brought are already being felt across all industries. AI will be able to complete increasing amounts of work that used to be done by human employees in a more efficient way. Currently, this is mostly the repetitive tasks which previously required intense human effort but limited cognitive skills, such as reading a contract, finding the key numbers and putting them into an excel sheet. Higher-level work requiring complicated skill sets such as planning, judgement and reasoning have proved difficult for AI. So AI is performing very well within a well-defined scope and ‘artificial general intelligence (AGI)’ still only exists as science fiction at this stage.

Despite these limitations, companies ignore AI’s disruptive impact at their peril. The second edition of Deloitte’s State of AI in the Enterprise report, published in 2018, found that 42% of executives believe that AI will be of ‘critical importance’ within two years. This ‘critical importance’, however, will probably have less to do with saving costs by replacing human employees, and more to do with optimising work processes, increasing returns on investment and supporting better strategy.

Three types of AI colleagues

Machine learning, deep learning technologies, voice recognition, speech synthesis, image recognition and natural language processing (NLP) are evolving and upgrading rapidly. AI can now outperform humans in voice and image recognition. While we have not seen NLP develop to this level, AI is making progress here too. AI can now see, hear, speak, understand and even think, to a certain degree, like a human being. With these tools, AI colleagues are able to play three typical roles in the work environment – they can automate business processes, augment business decision-making and they can facilitate engagement with customers and other employees.

1. Automation

The most common application of AI is the automation of human tasks and business processes. Robotic Process Automation (RPA) was introduced to the China market in 2017. Now RPA has integrated with cognitive components and evolved to Robotics & Cognitive Automation (R&CA). Robots can now handle both of these tasks according to predefined rules. This includes doing ‘smart’ tasks, for example:

• recognising named entities, such as a company or a person’s name
• extracting structured information from unstructured documents, such as contracts and financial statements
• automatically classifying invoices according to their product name in natural language, and
• searching websites to find target information.

These types of tasks are increasingly carried out by AI in fraud detection, for example, assisting forensic teams in searching unstructured documents for target terms and information. This enables teams to deliver relevant work faster, more accurately and at a lower cost. RPA and R&CA are very likely therefore to increase human unemployment from a long-term perspective, especially in the offshore business-process outsourcing industry. If you have the ability to outsource a project, you can probably automate the whole process.

2. Assisting decision-making

By using machine learning algorithms to detect key patterns and relationships from billions of data sources, AI can derive deep and actionable insights to support the business decision-making process. There is nothing new, of course, in using a wide variety of data sources to predict consumer behaviour – there is a significant difference, however, in the scale, speed and accuracy with which AI can process data. Big data technologies integrated with machine learning or deep learning algorithms can give surprising and useful insights. Moreover, this process of insight generation is not driven by user queries but by a continuous monitoring mechanism that is proactive and real-time based.

AI colleagues who provide cognitive insights can be considered senior consultants, but they won’t pose threats to the existing employees since their tasks are far beyond human capabilities. The cognitive insights that AI colleagues produce are useful to executives since they improve the quality of strategic decisions. Executives need to understand the importance of, and make strategic investments in, this type of AI.

3. Facilitating engagement

Digital agents are becoming increasingly popular in daily work and life. The Turing test is a test of AI’s ability to exhibit intelligent behaviour equivalent to, or indistinguishable from, that of a human. Although digital agents are, as yet, unable to pass the Turing test, they have already made our work and life easier in many areas. Digital agents are already present in the following industries.

• Consumer and industrial goods. Digital agents enhance the customer experience by answering questions contextually based on past behaviour, preferences, weather, vacation plans, etc.
• Financial services. Robo-advisers give investment advice and provide personalised investment options to customers, and monitor and alert users about changes to portfolio risk.
• Healthcare. Digital agents deliver personalised, collaborative and timely medical advice to patients, such as providing prescription refill reminders, thereby improving patient satisfaction and health outcomes.

Challenges

While AI is becoming an important element of the workforce in many different organisations, there are risks associated with the use of AI technology. According to our State of AI in the Enterprise report³ mentioned above, cybersecurity is viewed as the number one risk (see ‘AI challenges’). Cyber attacks targetting AI systems may cause data breaches and information security problems. This is not a problem unique to AI, but as dependence on AI increases, addressing cybersecurity concerns will become all the more important.

Source：Deloitte State of AI in the Enterprise, 2^nd edition, 2018)
Note: This survey interviewed more than 1,100 IT workers and executives from US-based companies in Q3, 2018.

Failures of AI systems are another critical concern, particularly where mission-critical data is involved. A photocopying error in an internal document might not pose a significant risk to an organisation, but an error in the key numbers in a financial statement would be another matter. In the medical field there can be ‘life-and-death’ situations where the risk of an AI failure cannot be tolerated. Human monitoring and control are still needed in many scenarios.

How should we get ready?

Despite the concerns discussed above, many companies that have invested in AI technology have already reaped economic benefits. The key is to take a practical approach to the benefits available from these technologies. This approach usually starts with implementing process automation via AI. This can lead to the adoption of RPA or R&CA technology, which is relatively cheap and can be applied to business processes within a managed scope. These benefits can be gained immediately and they can help establish confidence in the management team about future use of AI.

The question of finding the right mix of talent to successfully adopt AI should be the first consideration. Although China is in a leading position in the AI industry globally, there is a significant shortage of AI talent in China as compared with the US.

According to a 2017 Linkedin report, Talent solutions⁴, the US is ranked number one and China is in seventh position in terms of the available talent. It is calculated that there are only around 50,000 AI engineers in China. Another report from McKinsey, The future of artificial intelligence in China⁵, shows that most US data scientists have more than 10 years of work experience, while 40% of Chinese data scientists have less than five years experience.

Due to the high demand in the market, the salary levels of the data scientists and AI engineers keeps increasing. However, many companies still find it very hard to find the right talent to support the implementation of their AI strategy. There are typically three different ways to address this problem.

1. Companies can target global talent, especially the established Chinese researchers working in the US, for recruitment.
2. Companies can train existing employees to become AI experts. People with a background in mathematics, data mining and software development are well suited to a career in AI.
3. Companies can collaborate with university laboratories. Although fresh graduates from universities lack industry experience, their academic training gives them a good knowledge of AI. They can be good candidates for a new AI research team in the company, with enough in-work training.

Conclusion

It seems certain that increasing numbers of AI colleagues will join the workforce in the future and take over many labour-intensive tasks, but they are unlikely to eliminate humans from the workforce as many tasks require high-level cognitive skills. Further in the future it is harder to predict what may happen, but it seems likely that AI will continue to evolve and redefine the relationship between technology and humans.

The impact of these developments will be very significant for businesses and for the workforce. Organisations that adapt quickly to the new digital business model by embracing AI technology can increase their competitive advantage. As discussed above, AI can not only save on labour costs but can also assist executives with their strategy formulation and decision- making. Management teams need to assess what impact the increasing popularity of AI will have on the organisation’s workforce and plan ahead for the training that will be required to ensure the human workforce is adapted to the new AI era. Opportunities and threats are coming together; it will be the survival of the fittest.

Meet AIME

Digital agents are becoming more common in workplaces around the world. The first digital agent developed by Deloitte China was recently developed for a healthcare client to support employees with their internal policy inquiries. The Deloitte office in the Netherlands recently unveiled ‘AIME’, a digital employee with natural language processing and natural language generation capabilities. She supports the human workforce across different departments in Deloitte, internally and externally. Perhaps most important of all, however, AIME is able to learn on the job. She currently delegates tasks she is unable to handle to her human colleagues, but she is able to learn from these situations.

Endnotes

1. "Surviving the rise of 'smart machine", Maverick Research, 23 September 2013, https://www.gartner.com/doc/2594820/maverick-research-surviving-rise-smart.
2. "State of AI in the enterprise, 2^nd edition", Deloitte Insights, October 2018.
3. Ibid
4. "Talent Solutions", LinkedIn, https://business.linkedin.com/zh-cn/talent-solutions/s/sem-report-resources/ai-report.
5. "The future of artificial intelligence in China", McKinsey, 31 March 2017, https://www.mckinsey.com.cn/中国人工智能的未来之路/.