Generative AI is disproportionately affecting administrative, care, and coordination roles in British business — work that's essential but often uncounted.
The story of AI and employment, when it gets told, almost always gets told about the headline jobs — software engineers, illustrators, journalists, lawyers. But the work that gets done at most British SMEs isn’t those jobs. It’s the booking, the chasing, the rota, the policy update, the email to the client about the email to the supplier about the invoice that’s late.
This work is overwhelmingly done by women. It is overwhelmingly under-recorded. And it is — based on the conversations Underfold is having every week — the work most exposed to generative AI right now. Not by replacement; by quiet erosion. The chatbot does the booking. The drafted reply gets sent without editing. The policy update is summarised and forgotten.
The story of AI labour displacement is, for SMEs, the story of pink-collar work disappearing into the model.
Pink Collar is original citizen research grounded in three streams:
Field interviews — anonymised conversations with people in administrative, care, and coordination roles across 5–50 person businesses in the UK. Currently 17 conducted, target 60 by end of phase 2.
Process maps — granular task-level mapping of what these roles actually do, against what is now being delegated to AI. Sourced from real Tech-Xecutive client engagements (anonymised & consented).
Sector data — pulling published ONS, Resolution Foundation, and IPPR data on UK workforce composition, weighted toward SMEs (where most of these roles live).
What it isn’t: a survey. There are enough surveys. This is depth-first research.
Research design, ethics review, target cohort defined.
Interviews, process maps, sector data pull. 17 of 60 interviews complete.
Pattern mapping, sector breakdown, first findings.
External review with practitioners and academics.
Full open-access report, supporting microsite, dataset release.