History and AI: Why Microsoft and the World Economic Forum get it wrong

27 February 2026

I am a historian. I am not worried about this Microsoft report that rates ‘historian’ second in terms of occupations most at risk from AI. And neither should anyone who has (or wishes to) study history.

Let’s take the most obvious error first. What occupation will AI make redundant? A quick perusal of job sites under the heading of ‘Historian’ has never produced much. Even the fine detail of the report, based upon US Bureau of Labor Statistics, lists only 3,040 people in this occupation. Making it in the top ten smallest occupations of the list of 80 in the study.  We are not a dying breed. We never really existed as an occupation to begin with.

Studying or training in History is not vocational; that is, it rarely leads directly to a specific career as a ‘historian’. The world of politics, law, and business (all less at risk) are often peopled by history graduates. Historians are trained to analyse the cause of things, not just to explain that something happened but to weigh the factors that caused something to happen in the way that it happened. They know how to use valid and relevant evidence, and to identify reliable sources, something a world of proliferating digital data and fraud needs more than ever. Historians think about how humans operate within social and economic systems. According to the World Economic Forum the core skills in 2030 include ‘analytical thinking’ and ‘systems thinking’. Yet it rates as ‘out of focus’ things like ‘teaching and mentoring’. I’m at a loss as to how we will live in a world where these skills are vital yet somehow emerge out of our very essence, without training and honing.

Even a job like ‘researcher’, something history training has always helped with, might be presumed to be at risk because everyone can simply do their own research by asking AI. Yet anyone harnessing AI to research a given topic still needs to have well-defined parameters. We will still need people with the skill to ask the right question in the first place, to understand the connection between question and material utilised to answer it. This is a skill we work on every week in my undergraduate classrooms.

Then there are the most history-adjacent jobs. Most heritage sector jobs are about curating physical sites, leading community events, tours, and facilitating people’s interaction with physical objects. They are about connecting with history, not transactionally imparting information. Archivists and librarians are still needed to organise physical collections of artefacts. History teachers in primary and secondary school will still be needed for the foreseeable future, precisely because the process of studying history is one that teaches vital skills as well as knowledge. It is, in fact, still a thriving A-level and, presuming that governments and communities still value civic-mindedness, it will remain on the curriculum. For the smaller flock of historians who end up in these history-adjacent jobs, AI might be utilised to enhance – but will not displace – them because they are relationship-driven, a point AI consultants admit would mean the occupation was at far less risk.

At bottom is a fundamental misunderstanding of what history is and how it is made. It has been placed at the top of this report because it has been defined primarily in terms of ‘information work’. AI and, more specifically, large-language models (LLMs) sort data to summarise and reproduce. They impart information. They do not produce it. This is transactional, giver to receiver. The error, here, is that history is simply and straightforwardly the imparting of historical facts. For the small bunch of ‘historians’ who produce historical knowledge, the occupational definition and tasks this report is based upon misses crucial aspects of the job. We put objects to use in ways they have not been previously by asking new questions. We take what is absent from shipping records or court testimony to find the stories of people who are missing or in the margins of documentation. We dig through dusty folders to find evidence that can revise and revitalise the stories and digitised data that AI sifts through. Humanities interactions with each other and with the natural environment are recorded in song and dance, in objects and on parchment. The world’s knowledge is not digitised. If we do not recognise this, then we will continue to reproduce existing inequalities in our knowledge of the world. And even if it were, we need people with the skills to find the holes, ask creative questions, and consider contingency in how people and systems operate.

Whether as analytical thinkers or as knowledge producers, there will be a place for historians for a long time yet.

Leslie James is Reader in Global History at Queen Mary University of London. Her latest book, The Moving Word: How the West African and Caribbean Press Shaped Black Political Thought is now out with Harvard University Press.