AI, jobs and competencies: results from a study in the telco sector

Introduction
Most of the literature devoted to the rise of AI is expressed in quantitative terms. It generally opposes the emergence of new jobs against the spectrum of the disappearance of jobs that have become obsolete or fully replaced by the machine. The most notable example of this trend is the study of Frey and Osborne (2013). They apply this approach to the American labour market, identifying risk sectors and scan more than 700 jobs more or less likely to be threatened. Similarly, in a longitudinal and macro-economic perspective, Autor (2015) points out a polarization of the labour market due to automation, artificial intelligence and robotic process automation. This polarisation, on its bright side, lead to the emergence of highly sought jobs, among which the “data scientist” is considered to be number one, according to management magazines and websites such as Harvard Business Review (2012) or LinkedIn (Le Monde, 2017).
Many of these works also highlight the importance of soft skills, said to be central because of their presumed irreplaceability by the machine: social intelligence, creativity (Frey & Osborne, 2013), empathy, adaptability, integrity (Manyika et al., 2017), human experience, contextual knowledge or common sense (Günther, Mehrizi, Huysman, & Feldberg, 2017). Autor (2015) support these trends and conclude that the use of machine will strengthen what is unique in human skills: flexibility, judgment, common sense, problem-solving, or creativity.

Our position and research questions
There is no denying that with the rise of AI some jobs will appear while some other will disappear. However, we believe setting the debate solely in those terms is quite mechanical and reduces the richness and complexity of changes that every job can undergo.
In this regard, Schatsky, Muraskin and Gurumurthy (2015) or Brynjolfsson, Mitchel and Rock (2018) call for a shift in the debate on artificial intelligence and its impact on employment. They offer insights into job redesign and business process reengineering, with a task-centric approach and potential automation.
Therefore, we propose to shift towards a reflection on the transformation of occupations via the analysis of the tasks performed and their evolution. Beyond employment rates in general, our aim is to consider the diversity of tasks occupations can cover and to investigate the extent to which they can vary.
This emphasis on jobs transformation involves focusing on their very content and how the human-machine collaboration modifies the tasks that compose them. Brynjolfsson and Mitchel (2017) identify several distinctive criteria for tasks that are easily handled by the AI, and those that are less so, if at all. Such reflexions also involve considering the organizational phenomena that occur within organisations that choose to invest in an AI strategy. In other respects, to consider the impacts of AI through an analysis of the tasks and their evolution leads to focus on the competences needed for their accomplishment, and their respective weight within the impacted functions.
Based on the literature, we distinguish three types of skills that can be mobilised in relation with AI: technological skills, specific occupational skills, and soft skills (Bughin et al., 2018; France Stratégie, 2018; Hajkowicz et al ., 2016; Kolbjørnsrud, Amico & Thomas, 2016). We then develop two hypotheses in connection with these skills. On the one hand, these skills might evolve in varying proportions under the influence of AI, leading to a change in their weighting. In other words, the respective weight of occupational, digital and soft skills will vary depending on the collaboration of human and AI within the tasks performed. On the other hand, the development of AI applications for occupations will lead to a hybridisation of skills between the data-jobs (compiling data and generating the algorithms that are at the basis of AI applications) and user-jobs (end user of the applications and data analysis). This hybridisation will ensure that applications are tailored to the need of end-users considering their specific contexts.

Empirical data: context and methodology
In order to consider these research questions empirically, we conducted a survey, as part of a study commissioned by a major European group active in the telecom sector.
Two complementary and successive questionnaires were developed and distributed to 7800 employees and managers of the five job families targeted in this study (technical intervention, finance, marketing, customer relations and human resources) between September and October 2019. The response rate was respectively 11,8 (924 questionnaire – 1st wave) and 4,2% (326 questionnaires - 2nd wave).
Through closed-ended single or multiple-response questions, attitudinal scales and open-ended questions, we tried to identify the working practices of respondents in relation to the AI, the possible transformations of their jobs and skills. We also provided respondents with hypotheses about how the future will evolve in relation to the IA, likely to influence their jobs and activities in the company. On the basis of the first questionnaire, these hypotheses were further developed and detailed in the second questionnaire, through a breakdown in concrete terms of the mobilisation of AI in the company's activities and the perception of the impact on the jobs, skills and activities of the Group.

Results
First, respondents point to the perception of a transformation of their jobs much more than to a perception of appearance or disappearance of jobs. Appearance is often linked to the rise of not-so-new jobs but of existing ones that are more talked about and are actually the subject of a re-labelling process (such as data analyst or data scientist etc). Disappearance is seen as the result of organisational phenomena related to digitalisation (not necessarily AI) or to strategic business choices (e.g. outsourcing of some low added value activities). Transformation is considered essentially as a human-machine collaboration in which AI is considered as a support not exceeding 50% of the tasks. It is also seen as a repositioning process of the jobs on certain operations in particular that are seen as more complex such as piloting operations or management of exceptions.

Second, with regard to the use of skills currently held and the need for new ones, the need for soft skills seems, according to our results, of a lesser importance compared to what is generally assumed in the literature. The need for soft skills (and, therefore, its relative weight) is supplanted by the need to acquire digital and occupational skills, both in terms of strengthening the ones respondents already possess and acquiring new ones. Moreover, the need for digital skills (understood in a broad sense and not necessarily linked to AI) comes first. It demonstrates a form of hybridisation from users-jobs towards data-jobs via the acquisition of a basic data literacy that would ensure a better understanding of the analyses provided by the machines. This perception is however weaker within effective users of AI, who are less willing to consider the possibility of a general training to AI for all.

Conclusion
Our results highlight the relevance of taking a step back from the trends in organisational choices as regards the rise of AI. They are generally expressed in terms of percentages of loss in the workforce, through an emphasis on soft skills development and through massive recruitment of data scientist. As an alternative to these trends, we propose to consider the evolution of tasks inside occupations. An analysis of the occupations in terms of tasks reveals the hybridisation and weighting processes happening in every function. Theses processes ultimately have an impact on the definition of profiles that are sought and on the development of a digital acculturation for all.
Results also show a continuity in strategic business choices already made before the development of the AI, that are likely to be pursued or to be amplified. This calls for analysing the managerial discourses that might disguise such choices behind an alleged determinist and technological wave they could not counter. Eventually, AI remains a tool that can be mobilised by the company for its organizational choices in one way or another.