AI and the labour market: What the headlines are getting wrong

Research evidence excerpt

AI and the labour market: What the headlines are getting wrong

13 July 2026

AI and the labour market EconomicsGlobal

What the headlines are getting wrong

◆ Many of the studies on AI exposure are skewed by the Bethan Ellis

same biases… Global Economist HSBC Bank plc

bethan.ellis@hsbc.com

◆ … tending to overstate the risk to white-collar jobs… +44 20 7991 6714

◆ …and understate the extent to which AI could help people do

their job, rather than replace them

The devil’s in the details

The impact of AI on the labour market is something we’re asked about often, and

something that we have written about previously (see Gamechangers, 20 April 2026 and

Is AI already impacting employment, 14 October 2025).

This being the case, we track the research around the topic pretty closely too. There have

been a lot of papers looking at the impacts of AI on the labour market – and a lot of them

follow the same broad structure:

1. Apply an AI exposure variable across a set of jobs

2. Follow trends in those jobs, analysing results and making findings according to their

AI exposure variable

The problem is that a lot of the studies on AI and the labour market use the same

methods of calculating AI exposure. This amplifies the biases of a very limited number of

papers across a huge swathe of the analysis on AI and the labour market – while missing

a lot of nuances. On top of this, studies sometimes conflate a job being ‘exposed’ to AI

(where AI may help with the role) and a job being vulnerable to being lost, a distinction

that may then disappear altogether in subsequent media reporting.

In our view, because of this, the discussion around AI and its impact on the labour market

has a tendency to:

1. Overstate the present impacts of AI

2. Overestimate the exposure of white-collar workers

3.