Moderate pressure on general practitioners. The shape is what matters.

Primary care is harder for AI than radiology was. The job isn't reading one thing; the job is sitting with a patient and figuring out what's actually wrong from a fragmented story. That's the hard problem, not the easy one.

Automation is medium-low and slowly rising. Routine prescription renewal, standard test ordering, admin documentation, referral letters. GPs in 2026 are getting their evenings back; the paperwork is mostly machine-handled.

Augmentation is medium-high and rising. Differential-diagnosis support, pattern flagging on long histories, longitudinal patient summarisation. Tools like Glass Health and Microsoft Dragon Copilot are genuinely useful when used as a consult-room second opinion.

Resilience stays HIGH because clinical examination, relationship-of-trust care, ethics, end-of-life judgement, and the community context of family medicine are not in any model. The shape of GP work changes — more time per patient, less paperwork, more reliance on AI for the differential — but the role doesn't go anywhere. The risk is the opposite: too few GPs because the AI can't replace them and the training pipeline is too slow.