AI is set to transform the employment landscape and employers are already using AI to enhance efficiencies and collate data to monitor trends and the productivity of their workforce, as well as in recruitment and monitoring employee behaviour.
AI can help to remove both conscious and unconscious bias in decision-making and to ensure consistency of approach. However, will it ever be acceptable culturally for a machine to decide to fire an employee? Where should the line be drawn when important decisions need to be made about employees’ performance or disciplinary matters? Is the human element still important in this process
Employers have been using software to facilitate disciplinary proceedings and performance management processes for many years, but AI has provided new opportunities for employers to analyse the data collected, monitor employees’ performance and behaviours and produce reports on the information collected. AI is already being used as a management tool, for example in:
AI is also being used for monitoring purposes, to track employees’ performance and conduct and to protect companies’ business interests. This technology allows employers to record behavioural patterns of employees and build a picture of the routine tasks they undertake. Once an employee steps outside their usual behaviour, for example by collating and copying an unusually high number of company confidential files, the software can
Disciplinary decisions can be better informed by analysing these sorts of data. Prevention of a problem at an early stage is often preferable to curing the problem once the confidential information has already been misused or the company’s reputation already damaged.
As a manager, should you be concerned about developments in AI? A Deloitte report from 2014 found that only 58% of companies said that their performance management process was not an effective use of time. But why? Technological advances, as well as shifting cultural patterns, mean that the way we work has evolved. Companies have expanded their geographical footprint. Homeworking has become a widely accepted practice. This can make it harder to assess productivity in a traditional sense, so performance reviews may need to rely more on data.
AI could provide a solution. Tools can be designed to enable employers to instantly collate and analyse information and send feedback to management on an ongoing basis. This could enable managers to track a bad hire and identify early on if an individual is not a long-term fit for the business. However, such a system requires that the AI technology is embedded with the company’s ethical code and values. This may not be possible, in which case a combination of technological and human influence may be preferable.
Equally, many employees are filled with dread at the thought of spending hours filling out their annual appraisal form and waiting for the appraisal meeting. The time and cost associated with appraisals could be alleviated with the use of AI tools to gather and report data. Will annual appraisals as we know them become a thing of the past with employees instead receiving an automated report? But how far can we reduce ‘man management’ time before it affects morale and employee engagement, and threatens employees’ confidence and trust in decision-making?
There may also be regulatory compliance issues. For example, in the UK senior employees in the financial services sector must be certified as “fit and proper” on an annual basis. Can a computer properly perform this assessment on behalf of an organisation?
Employers will invariably determine the appropriate degree of AI involvement based on their organisational model. Some will decide to integrate it into many aspects of their HR processes. Others will use it in only limited ways.
The use of AI in objective decision making processes or assessments could potentially assist in defending an allegation of bias and to help demonstrate that the process was fair. However, AI is only as fair as the algorithms behind it and the algorithms used in software can be biased in the same way as individual managers. For example, the software may have a preference for certain skills or attributes and discriminate against employees without those particular characteristics. Nevertheless, provided the underlying technology is not crafted in a discriminatory way, it may be possible to significantly reduce the risk of discrimination in some decision making.
Further, multinational organisations could use AI technology to standardise and consolidate their processes. If businesses deploy software tasked with investigating or managing disciplinary processes, this could result in a more consistent and harmonised approach for their worldwide operations.
However, if the human element is entirely removed, will technology ever be sophisticated enough to understand and rationalise mitigating factors in a way that a human might? Whilst a machine may be able to identify that an employee simply had a bad day at work for personal reasons or underlying medical issues, and that this should be factored into the decision making process, the appropriate value to be placed on it may be more tricky? Can and should we circumvent human input entirely? How would a court view a decision made solely by a machine? And what about the court of public opinion? For example, lie detector tests have existed for decades, but they are not routinely used in anything but the tightly controlled and regulated environment of criminal investigations. It is unlikely they would be considered acceptable in the workplace decision making processes.
It may well be that a compromise position is preferred which involves, for example, a manager considering an initial recommendation or analysing reports produced by an AI tool.
If an employee successfully challenges a decision that was made with the use of technology, there may be uncertainty around who is liable – the employer or the technology provider. Employees are highly likely to be resistant to AI tools, which monitor and analyse their activities through company devices, particularly outside their normal working hours or beyond the workplace.
From a regulatory perspective, the EU’s General Data Protection Regulation and the data privacy regimes in many other jurisdictions place legal controls on automated decision making, including requirements to inform individuals about such activity.
In conducting a disciplinary process, businesses must behave fairly and lawfully in respect of monitoring activities and employees have a reasonable expectation of privacy in the workplace, even if they have been informed that monitoring may take place.
Given the potential scope of the use of AI by businesses and its wide-reaching implementation, there have been discussions at EU and national level about whether it should be uniformly regulated. A UK Government Select Committee was set up earlier this year to seek views on the possibility of AI becoming regulated. Although the Committee concluded that a blanket regulation would be inappropriate at this time and existing sector specific regulators are best placed to regulate matters arising from the use of AI, this does not rule out future AI specific laws or regulation.
AI is already widely used in the workplace to enhance efficiencies and reduce costs. Whilst concerns may be raised about the increased use of AI in performance management and disciplinary processes, including being “fired by a machine”, there are likely to be procedural and practical benefits to its use. Nevertheless, it seems likely that a degree of human involvement will continue to be necessary in such processes.