Your AI assistant sounds confident. But which month does the figure refer to?

But which month does the figure refer to? Is it CPI or CPIF? And does the link lead to the table where the figure actually appears?

That is often where the checking begins.

An AI assistant can help you find and analyse statistics. But a well-written answer does not show that the data are current or comparable. Even with web search, the right source, period and measure still have to be chosen.

If you are writing a board paper, a memo or a market analysis, the details are what matter. A figure can be correct in its own context and still be wrong in yours.

Three questions show what it takes for the answer to be usable.

What is inflation in Sweden?

In September 2026, the inflation rate was 1.5 per cent according to CPIF and 1.1 per cent according to CPI, according to Statistics Sweden (SCB). Same country, same month – but different measures.

An answer that only says “inflation is 1.5 per cent” leaves an important part unsaid. If the figure is going into a decision paper, it needs to be clear what it measures and which period it refers to.

Source: SCB, CPIF and CPI, retrieved via Sancera.

Which of Sweden's 20 largest municipalities have grown the most in population since 2015, in per cent? Give me the data in Excel.

Before opening a new location or assessing a market, a comparison like this can be a first step. But the answer takes more than a few population counts.

Which municipalities are included – the largest in 2015 or the largest in the final year? And is the same geographical boundary used throughout the comparison?

A useful answer states the selection and the final year, ranks the municipalities by percentage growth and shows the population at both points in time. The Excel file needs to contain starting values, final values, calculations and source references, so that the analysis can be checked and built on.

This is the kind of work where the path from question to finished dataset matters more than any single search result.

How does Sweden's unemployment compare with Finland and Denmark?

In August 2026, unemployment was 8.9 per cent in Sweden, 9.9 per cent in Finland and 5.9 per cent in Denmark, according to Eurostat's harmonised, seasonally adjusted series for people aged 15–74.

Here the comparison is based on the same measure and month. Picking each country's most recently published figure can give a different picture, since the data do not always refer to the same period.

Source: Eurostat, monthly unemployment, retrieved via Sancera.

So the hard part is not just finding figures. It is choosing the right data, making them comparable and producing a dataset that actually answers the question.

Sancera connects your own AI assistant to official statistics from 100+ sources, including SCB, Riksbanken, Eurostat, the ECB and the IMF. You ask the question in, for example, Claude, ChatGPT or Cursor and get access to source data in the same workflow, with source references and the option to export to Excel or CSV.

That shortens the path from question to analysis. You can use the assistant to process the data and trace the figures back to the source before they end up in the presentation.

A data connection does not make AI infallible. Even with an official source, the assistant can pick the wrong series or draw the wrong conclusion. That is why traceability is central: both the figures and the calculations need to be checkable.

Next time your AI assistant gives you an exact percentage, ask a follow-up question:

Which period does it refer to, and can you show me the original source?

And when you need more than a single figure: ask for the full dataset.

The figures were retrieved via Sancera on 25 October 2026. Statistics may be revised after publication.