Skip to main content

Working with AI Answers

Once you've connected EDITED to your assistant, you don't need a query language or a manual to get real value — you just ask. This page is about getting good answers: how to ask, how to read what comes back, and how to lean on the assistant itself to guide you.

What's happening when you ask​

An answer usually blends two things: facts and numbers from EDITED (retailers, brands, prices, assortment, trends) and the assistant's own wording and interpretation — how it summarises and what it chooses to highlight. Most of the time that blend is exactly what you want; knowing the seam between the two is what makes you good at reading the answers.

For how your assistant actually pulls that data — and what it means for your own data — see How it works, and your data.

Let the assistant be your guide​

You don't need to study documentation to use EDITED well — the assistant can explain itself. Ask it:

  • "What can I ask EDITED about?" — to see what's available before you dig in.
  • "How should I ask for this?" — if you're not sure how to phrase a question.
  • "What does this result mean?" — to have a number or table explained in plain terms.
  • "How did you get that — what did you look up?" — to see the steps and the EDITED data behind an answer.

Treat it as a knowledgeable colleague who's happy to show their working. The more you ask it to explain, the faster you learn what EDITED can do — and the less you need pages like this one.

Ask good questions​

A little specificity goes a long way:

  • Name the specifics — which retailer or brand, which region, which timeframe. The assistant can resolve fuzzy or misspelled names, but the more precise you are, the sharper the answer.
  • Iterate — it's a conversation. Start broad, then refine with follow-ups: "just the UK", "only the last quarter", "compare that to a year ago".
  • Ask for the shape you want — a short summary, a table, the top five, a trend line.

Check what you'll act on​

For anything you'll make a decision on, give the answer a quick sanity check — the same way you would a fast-moving colleague's:

  • Ask where it came from — "Is that from EDITED, or your general knowledge?"
  • Ask for the underlying data — the specific figures or records behind a summary.
  • Cross-check when it matters — against what the data covers, the data dictionary, or the Reference.

The assistant is a fast analyst whose work you can spot-check — not an oracle.

The honest edges​

Knowing the boundaries makes you quicker, not more cautious:

  • Coverage and freshness have limits — EDITED tracks a great deal, but not everything, and data has an as-of date. See what the data is.
  • Confidence isn't proof — an assistant can sound certain even when it's stretching. A surprising answer is a good moment to ask it to show its working (above).
  • Outside EDITED's scope, it's guessing — ask about something EDITED doesn't cover and the assistant may answer from general knowledge. When it matters, ask whether the answer is backed by EDITED data.

If an answer looks off​

  • Rephrase — try the question a different way, or more specifically.
  • Ask it to re-check — "Can you look that up again in EDITED?"
  • Ask what it was based on — "What did you use to answer that?"
  • Check coverage — what you're after may be outside what EDITED tracks.

If an answer still seems genuinely wrong after that — not just unexpected — tell us what you asked and what you got back, and we'll take a look.