At a glance
- Source
- Budzyń, Romańczyk, Kitala and colleagues (2025). Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study. The Lancet Gastroenterology & Hepatology, 10(10), 896–903. doi:10.1016/S2468-1253(25)00133-5
- How it was done
- The authors took routine quality records from four clinics and compared colonoscopies done without AI in the three months before AI was introduced with those done without AI in the three months after. Data analysis. It builds a measure from existing data, such as job and task records.
- Who took part
- 1,443 colonoscopies done without AI: 795 before AI was introduced and 648 after.
- Where
- Four endoscopy centres in Poland taking part in the ACCEPT trial, where AI that marks possible growths on screen was introduced at the end of 2021.
“Continuous exposure to AI might reduce the ADR of standard non-AI assisted colonoscopy, suggesting a negative effect on endoscopist behaviour.”
What it found
The study measured the adenoma detection rate: the share of colonoscopies in which the doctor finds at least one adenoma, a growth that can turn into cancer. Clinics record this rate as a standard check on quality.
- Before AI. In 795 colonoscopies done without AI, doctors found at least one adenoma in 28.4%.
- After AI. In 648 colonoscopies done without AI in the three months after the clinics introduced it, the figure was 22.4%.
- The size of the drop. The rate fell by 6.0 percentage points. The likely range runs from 1.6 to 10.5 points.
- After other factors were counted. The authors adjusted for the patients’ sex and age, among other things. Having worked with AI still went with a lower detection rate.
The drop appeared within about three months.
When the result holds
The skill in question is spotting something on a screen, which the AI was doing for the doctors on other days. Any job where a tool flags problems and a person used to find them alone has the same structure: visual inspection, document review, checking AI output.
Most workplace studies measure how people do with the tool. This study is one of very few that measured how people do without it after using it.
Limits
- The study compares two periods. It does not assign doctors to AI by chance. Other changes between the periods, such as in patients or workload, could explain part of the drop.
- It covers one medical specialty at four clinics in one country.
- It does not show whether the skill comes back once people stop using the tool.
Practices that use this study
This study supports:
- Measure time saved per task against a baseline Record how long a task takes and how good the result is before you add AI. Without that starting figure you cannot tell a real gain from a hopeful one. Do first · Effort: low (days, one person)
- Check now and then what staff can do without the AI On a sample of cases, have people do the work without the AI and compare the result with their usual work. Start with the people who check AI output and the people still learning the job, and set the tools that learners use to give hints, not answers. Do later · Effort: medium (weeks, a few people)
- Keep the entry-level tasks that newcomers learn on Before you automate routine work, check whether it is how juniors learn the job. If it is, keep some of it for them, or replace the learning another way. Do later · Effort: medium (weeks, a few people)
This study limits or contradicts: