How to Use AI for Analysis

Data

Spreadsheets, sales, survey results, and metrics

Documents

Reports, contracts, research papers, and transcripts

Feedback

Reviews, support tickets, and comments

Markets

Competitors, trends, and pricing

Decisions

Options, risks, and trade-offs

AI finds patterns fast. You decide what they mean.

ChatGPT

Uploads and analyses files, runs calculations, and makes charts

Claude

Strong with long documents and careful reasoning, plus Claude for Excel

Gemini

Works inside Google Sheets and Docs, with very long context

Microsoft Copilot

Works inside Excel, Word, and Teams, with the Analyst agent for data

Vague

Analyse this data

Clear

Which three products drove the drop in revenue last quarter, and why?

A good question tells AI what to look for and what decision it supports.

• One table, with clear column names
• Remove personal data you do not need
• Fix obvious errors and duplicates
• Explain what each column means

Here is our sales data. Describe what is in it, check for missing or odd values, and suggest five questions worth answering. Do not draw conclusions yet.

• Trends over time
• Differences between groups
• Outliers and unusual events
• Possible causes, labelled as possible, not proven

Ask AI to show the calculations it used, so you can check them.

Compare these three reports. List where they agree, where they disagree, and what evidence each gives. Quote the page for every key point.

Compare these options on cost, risk, time, and impact in a table. Then give a recommendation, the main reason, and what would change your mind.

• Ask for the chart that best answers the question
• Bar charts for comparisons, lines for trends
• Start bar chart axes at zero
• Label everything clearly

• Recalculate key numbers yourself
• Ask AI to find weaknesses in its own analysis
• Look for correlation presented as cause
• Test conclusions against what you know
• Check sources for any outside facts

Always ask: what would make this conclusion wrong?

• Lead with the answer
• Three key findings with evidence
• One chart per finding
• Clear next steps
• Limits and assumptions

• Follow your company’s AI and data rules
• Use business plans for company data
• Remove names, emails, and IDs where possible
• Check data and training settings