ResearchXL by Speero

Ask Me Anything Q&A Speeroamaquestionrepo
1

Based on the Research XL blueprint from Speero, where and how do you apply AI? What AI tool do you use? What steps do you take? How do you QA/Confirm that the AI is not hallucinating?

Do you automate things to go through an AI and build a deck for you? What specifically do you automate and send or AI to minimize the extra time of the copy and pasting into AI?

2

Hey @bigbear420 , good question.

We have used ChatGPT 4o and Gemini a ton for qualitative and quantitative analysis across research pieces. We also use them for refining research ideas and test hypothesis like finding an iteration on an idea or a white paper for the hypothesis.

The steps are quite varied based on the type of data, Qualitative analysis requires a lot more context to be provided to the tool. For quantitative data is as simple as asking for common patterns and summarizing the data.

We confirm that AI is not hallucinating by

  • Always providing structured data or clear context (the less ambiguous the input, the better the output)

  • Spot-check AI results against source data, especially for qualitative clusters or trend summaries

  • Never take AI-generated stats or summaries at face value—they’re used as a guide, not a final decision

We’ve experimented with partial automation to minimize the repetitive parts:

What we do:

  • Export post-test analysis data (from VWO, Kameleoon, spreadsheets) and perform a quick analysis.

  • Use AI to generate a rough narrative, highlight insights, and create bullet points for slides

What we don’t do (yet):

  • Fully automate decks—we found it strips away nuance or hides the “aha” insights that only come from close reading.
1 Like