A fork in the road: how AI can help or hinder the path to better work

Louise Marston, Director of Ventures

It’s become a staple of investor conferences to hear how AI is changing “everything”. But it’s rarer to have a discussion on how it’s coming for our own jobs in the form of AI-powered investment fund managers (if you haven’t already come across this – get ready to meet Boardy).

Last week I was pleased to speak to a discussion group hosted by 3Space, a partnership between Impact VC, Rothschild and Uplink, on how AI can help and hinder impact on social goals, specifically on economic empowerment.

With fellow panellists James Anthony and Mario Barosevcic, we discussed how AI can impact education, financial resilience, but especially work in this discussion – something of particular interest to me as the manager of the WorkerTech Fund.

I believe there are three main ways that digital technology, including AI, can create positive impact on the world of work:

• Better information – which can create more power and agency, or remove inefficiencies.

• Greater connection – helping people to organise, find support, or to find the right type of role for them.

• New ways to organise work – that could mean devolving power, improving flexibility or reducing administrative burdens.

All these can have a real impact on quality of work for people – something we think about in terms of increased pay, better prospects, greater power and improved wellbeing.

In the book ‘Is This Working?’ by Charlie Colenutt, which documents the experiences of workers across the UK, one of the key themes that emerges is the story of ‘this used to be a great job, one I really enjoyed, but now I seem to spend all my time on admin’. That theme connects teachers, and advertising, and any number of other jobs, and speaks to a combination of power over their work, and wellbeing impacts of doing tasks people dislike.

One way that AI could make a big positive difference, is to remove or make much more efficient the sort of administrative tasks that people enjoy less, to make room for the tasks that involve more creative thinking or personal connection.

And this is more than theoretical. We’ve seen some great examples already in the WorkerTech fund of companies who are putting AI to work in the service of workers. Earlybird has shown it can remove 70% of the admin time for employment support workers, improving their engagement with the people they are supporting, and helping them get more people into jobs. And it’s loved by its users. Valla makes it possible for more people to access legal support with employment claims by using AI to make the process much more efficient and lower cost.

The potential for negative impact is also there in this drive for efficiency, however. When tasks are automated, there are choices about how they are divided between humans and technology. In previous waves of automation, some workers found that the interesting and engaging parts of their jobs were automated, while the repetitive ones increased in intensity – think of micro-managed warehouse workers being instructed on which product to take out of which bin, or call centre handlers being rapidly passed from one call to the next, with a series of programmed prompts. These types of ‘low-discretion automation’ (a term from the Institute for the Future of Work typology) can make jobs worse, not better.

For me, the greatest barrier to taking the positive impact route to AI implementation now is lack of power and agency by workers.

Daron Acemoglu and Simon Johnson in their book ‘Power and Progress’ outline the example of automation in car manufacturing as an illustration of how different structures for worker power influence different outcomes.

 To paraphrase their account, introducing robots to US manufacturing reduced both employment levels and wages for those frontline roles. In Germany, where companies had to negotiate with unions, workers were represented on company boards, and there were fewer skilled workers available, there were more efforts to retrain workers and reallocate them to new tasks. The number of workers in the German car industry went up in the same period those jobs were declining in the US.

 To make the choices that ensure there is more upskilling and reskilling and less wholesale replacement, or degradation of roles requires the workers to be represented in the rooms where those decisions are taken, and to have a voice, and to be ‘AI confident’ to participate.

 The potential for positive impact of AI on work remains present. And there is still time to make choices about how we implement it, who is involved, and what future of work we want to see. But we need to put in place the structures and ‘institutional furniture’ to make that happen. It’s not the robots that are coming for our jobs – its other people, making individual decisions across thousands of workplaces. And that is something that we can influence for the better, if we choose to.