LO
Statistician — tasks, one by one
The unit of analysis is the task, not the job title. Each one below carries its direction, whether the judgement rests on evidence or on platform inference, the reasoning, and what it does not establish.
Every task on this page#
Designing surveys, samples and trials
Still human-led✓ Evidence-backedDeciding who to sample, how many, how to ask and how to analyse, and writing the analysis plan before data arrive.
No record found shows AI designing samples or trials in production. In trials, Europe's medicines agency requires any AI used in the analysis to be pre-specified and frozen in the statistical analysis plan, which keeps design decisions with the people who write it.
The absence of a record and one regulator's guidance; it does not measure whether AI tools help draft designs or plans.
Processing and coding data
Automating✓ Evidence-backedCleaning, editing and classifying responses so that they can be counted, such as coding written answers into occupations and industries.
Machine coding is in production at statistics offices: the US Census Bureau's language-model autocoder cut the cases sent to clerical coders by 500,000 a year, and Statistics Canada has coded part of its Labour Force Survey by machine learning since October 2021. Much of this work was done by clerical coders rather than statisticians.
Two agencies' accounts of coding; they do not show how statisticians' own time on processing changed.
Statistical analysis and modelling
Being augmented✓ Evidence-backedEstimating, modelling and testing, and judging what the data can and cannot support.
Britain's statistics regulator says AI in official statistics is still focused mainly on research, testing and administrative support rather than routine statistical production. In trials, AI models used to analyse data count as part of the statistical analysis and must follow the same statistical principles, and a 2024 benchmark found the strongest model answered 58 percent of data-based statistical and causal questions correctly.
A regulator's observation for one country and one agency's guidance; they do not measure how much analysis AI tools now draft.
Quality assurance and methodology
Being augmented✓ Evidence-backedChecking estimates for errors and breaks, setting error tolerances and documenting methods.
Where coding was automated, the checking stayed with statisticians: Statistics Canada compared 13,665 published estimates with and without machine coding and found no breaks, and the Census Bureau set its threshold so the error rate matched the 6 percent limit clerical coders must meet.
Two agencies' accounts; they do not show whether quality-assurance work grew overall.
Advising and communicating results
Still human-led✓ Evidence-backedExplaining what results mean and how far they can be trusted to policymakers, clinicians and the public.
No record found shows AI advising on or explaining statistical results in place of statisticians. The regulator's assessment places AI in support roles, not in producing or explaining official statistics.
The absence of a record; it does not measure how statisticians' time on communication is changing.