Opinion:
AI in the workplace – progress or displacement of the workforce?

Well first, what is AI?

Put simply, it is a technology that allows computers to learn highly complex patterns and relationships in data and from this learning they can provide solutions or further insights, simulating human like intelligence.

‘Generative AI’ goes one step further and not only learns the patterns and relationships in data, but can create entirely new data or content in the way a human would. ChatGPT is an example of this.

What impact has AI had on the workplace so far?

The Department for Culture, Media and Sport published a report in January 2022, which found that 68% of large companies, 34% of medium sized companies and 15% of small companies had adopted at least one AI technology at that time.  So, it has been entering the workplace for a while, but at different degrees.

The Institute for Public Policy Research (‘IPPR’) published a report on 27 March 2024, looking at how generative AI could affect work in the UK. The IPPR found in a large-scale assessment of 22,000 tasks in the UK economy that about 11% of tasks are exposed to generative AI right now.

However, at this stage, the introduction of AI largely feels to be a welcome addition to an existing human service.  Chatbots in a customer service context, for example, answer the more standard questions freeing up human advisors to deal with more complex issues.  They can also work 24 hours a day, providing a bridging service day-and-night.

So, the impact at this stage appears to be mostly positive.  Availability and efficiency is boosted, and there doesn’t appear to have been the displacement or relegation of humans to any significant degree at this stage.

What impact will AI have on the workplace in the future? 

The IPPR, in the report mentioned above, identifies three phases in the implementation of generative AI, with potential effects on different groups at different times.

In the first phase there is predicted to be a transformation of the routine back-office roles and some creative tasks, with occupations likely to be impacted including secretarial and administrative roles, as well as writers and translators.  The impact of AI in the first phase is likely to be felt more by women and young people, who are more likely to fill these types of roles on a part-time basis or as entry level positions.  For roles with a manual element, the expansion of AI will continue to be limited by how quickly robotics can evolve to keep up.

In the second phase, when generative AI has become more deeply integrated with existing processes, the number of impacted tasks is anticipated by the IPPR to potentially increase to 59%.

In the third phase, processes will be built around AI, but this is projected to only happen where societal norms and regulations change to allow for roles such as teachers and doctors to be provided by avatars, for example.

Importantly, however, there is no set path for how AI will be introduced into the workplace or how it will affect it.

AI clearly offers benefits and exciting potential to provide efficiencies, boost productivity, and guide business strategy in new ways with the ability to analyse huge amounts of data in quick time.   AI can also deliver consistent results, reducing the inevitable element of human error which occurs when analysing large data sets.

However, there are also some very clear problems with the implementation of AI. The technology is still very new and, whilst tools like ChatGPT are widely accessible, more bespoke products are likely to be slower to implement and to be expensive.  They require skill and expertise to build and maintain, which may create a race to the top for bigger organisations, leaving smaller organisations struggling to compete.

From a HR and legal perspective, relying on AI to manage the workforce gives rise to potentially significant issues.  There are many areas of the employment lifecycle where AI can be used, from recruitment (e.g. preparing job descriptions, screening candidates) and coaching and development (e.g. suggesting new positions, providing training) to performance management (e.g. measuring individual employee performance) and redundancy selection (e.g. comparing redundancy selection data).

However, there are risks with many of these uses, largely stemming from the lack of visibility and understanding as to how AI tools work and how a particular decision has been reached.  For example, in recruitment, if an AI tool is used to screen candidates it will do so based on the data available to it.  If the data is unbalanced or biased this can lead to a discriminatory result, for example, disproportionately rejecting job applications from female candidates for senior roles because more men have been hired to these roles successfully in the past.  If AI tools are used as the basis for decisions about such things as employees’ capability, conduct, pay or promotions, without managers fully understanding and being able to explain and justify the data and information behind the decision, that could cause issues from an employee relations perspective as well as when defending any potential legal claims.

The big question for most, however, is whether AI will replace humans in the workplace, creating mass unemployment.  This, of course, is a stark possibility.  But, whether the possibility becomes a reality will, we think, depend on the policies and regulations put in place in the next few years to set us on the right path.

So, what steps are being taken to steer us towards positive progress?

There is currently no specific legislation in the UK governing AI.  Of course, there are existing laws, such as the Equality Act 2010 and the Data Protection Act 2018, which will apply generally to the implementation of AI in the workplace.  But there is no ‘UK AI Act’, as such.

The Government to date has indicated that they intend to govern AI through a voluntary regulatory regime, rather than legislation.  One reason being that creating new law takes time, and with the rate AI is developing, new law could well be out of date before it even comes into effect.

Instead, the Government is taking a principles-based approach, setting out key principles in respect of managing AI that are intended to be implemented by existing regulators (e.g. the Information Commissioners Office, the Equality and Human Rights Commission, the Health and Safety Executive and the Financial Conduct Authority).

This is in contrast to the EU, which is taking a more interventionist approach and has introduced a new draft EU AI Act, designed to sit alongside other new and existing legislation, setting out a legal framework for the use and implementation of AI.

The Governments’ ‘agile’ approach, which in reality is likely to provide limited protection for workers in respect of the displacement of jobs, has been challenged and a draft Bill (the Artificial Intelligence (Employment and Regulation) Bill) was recently published by the TUC.  The stated purpose of the Bill is to regulate the use of AI systems by employers in relation to workers, employees and jobseekers to protect their rights and interests in the workplace.

The Bill puts forward the prospect of a more regulated approach, which to date the Government has rejected, continuing on its path to a voluntary regulatory regime.  However, there has more recently been an indication that legislation may be needed and will be introduced when the risks are fully understood.  ‘Will this be too late?’, is the question. And, following the general election, will there be a change in approach if there is a change in government?

Conclusion

Employees working longer hours coupled with an ever increasing risk of ‘burnout’, an ageing workforce and skills shortages in certain sectors might all suggest that there is enough work to go round, with scope for AI to be introduced into the workplace alongside a human workforce to boost and enhance productivity, skills and wellbeing, as well as improving business efficiencies and output, without the displacement of the human workforce as a result.

Whilst AI appears to have endless potential, ultimately, in our view, we are a long way away from total, or even substantial, displacement of the human workforce.  There are key limitations to AI at this stage.  In most cases, AI still requires a significant amount of human input, monitoring and auditing.  Ultimately, in many cases, there is no substitute for a human’s ability to see the context of a situation and provide an appreciation of the nuances that are so often required in the workplace.

Does this mean that we can lock that ominous beast in a cupboard and throw away the key, never to be concerned about the potential impact of AI on our role in the workplace again?  We wouldn’t go that far.  These are going to be a crucial (and interesting) few years, but ones that we’re choosing to embrace, with caution.