The body and the skeleton

Why work is not a series of tasks and the implications for AI adoption

Have you ever wondered why it’s so hard to write your own job description?

It’s in part because no job is adequately summarised in a list of responsibilities or duties. Work is not a series of tasks.

The analogy I use is of the body and the skeleton.

To reduce the body of work down to the skeleton of tasks is to ignore the connective tissue, the circulation, the tendons and ligaments of interpersonal relationships.

Just thinking (for the moment) of white collar work – so much involves negotiation, management upwards or downwards, persuasion, nuance, contextual understanding, historical appreciation, emotional perception (if not intelligence).

Not to mention gossip, group dynamics and social norms, likeability, authority and the swathe of heuristics that determine impact, whether we like it or not.

A basic example from PR

The clue to successful media relations is in the name. It requires the development of relationships with journalists in order to add value to a message by mediation.

To reduce media relations to the tasks of drafting and sending a press release is to fundamentally misunderstand the nature of the work.

And yet, in so much of the analysis of employment and AI, and by extension in the thinking of organisations slimming down workforces to make way for AI, this category mistake is made again and again.

We can see the consequences already at organisations like Klarna, which swiftly rowed back on AI-related job cuts. More broadly, a Careerminds survey found that roughly a third of companies that conducted AI layoffs had rehired 25% to 50% of the roles they cut, and another 35% had rehired more than half - including one in three employers who spent more on restaffing than they saved from the layoffs[1].

Gartner projects that half of all companies that cut customer-service jobs for AI-related reasons will rehire people for similar roles within the next year.[2]

Sam Illingworth writes at length about the lack of evidence driving AI-related job cuts in his Slow AI newsletter.

This oversimplification of jobs isn’t a new problem. The traditional task and competency approach of defining job elements has significant shortcomings. It ignores the relational, contextual and political fabric of work. What Michael Polanyi called tacit knowledge[3].

The problem with tacit knowledge is that it’s really difficult to codify. Which means we return to reductive tropes. To equating the body with the skeleton.

This is a really longstanding issue for, instance, in curriculum design and workforce planning.

And because the standard approach to calculating the ‘vulnerability’ (for want of a better phrase) of jobs to AI is occupation and task based, it’s now an issue for AI deployment.

The consequences of which are going to be, I’d wager, profound.

A version of this post was published on my 'AI and Reputation' Substack on 3 June 2026.

[1] Careerminds (2026) 'AI-led layoffs: What HR leaders wish they knew before making job cuts', March

[2] Gartner (2026) 'Gartner Predicts Half of Companies That Cut Customer Service Staff Due to AI Will Rehire by 2027', 3 February

[3] Polanyi, M. (1966) The Tacit Dimension, Doubleday