Agents · Lesson 02 Pro ~10 min read Data → insight

Analyst: data without analyst on retainer.

Analyst is the quantitative sibling of Researcher. Where Researcher synthesizes documents and the web, Analyst crunches numbers — your Excel files, your CRM data, your dashboards. Statistics, charts, trend analysis, comparisons. The agent that handles 'help me understand this data' for everyone who isn't a data analyst.

Workflow 01 Excel file → clear insight

1

Hand it a spreadsheet, get back analysis

The canonical Analyst use case: complex Excel data, you need to know what it means. Analyst handles the analysis you'd otherwise do manually with PivotTables.

The prompt that works

Sales data analysisI have an Excel file (q1-sales-by-rep.xlsx in SharePoint) with sales rep performance data — 50 reps, 6 columns: name, region, deals_closed, deals_lost, total_revenue, quota. Analyze: 1. Top and bottom 5 reps by revenue and by quota attainment 2. Win rate (closed / (closed + lost)) by region — is there a regional pattern? 3. Is there correlation between number of deals closed and average deal size? 4. Any reps who appear to be outliers (very high or very low compared to their region)? 5. Build 3 charts that would explain this to executives in a meeting Give me a 1-page summary in business English. Save the charts as separate files.

Best use cases

  • Sales performance reviews
  • Marketing campaign result analysis
  • Operational efficiency reviews
  • Customer cohort analysis
Analyst's interpretations can be confidently wrong if your column names are ambiguous. Verify the analysis makes sense before sharing — check that the agent understood which column meant what.
Time savings: Excel → executive summary: 2-3 hours of pivot tables → 10 min.

Workflow 02 Cross-source quantitative comparison

2

Pull data from multiple files and compare

Analyst can pull from multiple files in SharePoint or OneDrive simultaneously. Comparing current quarter to historical data, or cross-team benchmarking, becomes a one-shot prompt.

The prompt that works

Multi-quarter comparisonCompare our Q1 2026 marketing performance against the last 4 quarters. Use: - Q1-2026-Marketing.xlsx - Q4-2025-Marketing.xlsx - Q3-2025-Marketing.xlsx - Q2-2025-Marketing.xlsx - Q1-2025-Marketing.xlsx (all in our Marketing Analytics SharePoint folder) For each channel (paid search, paid social, content, email, events): - Quarterly spend trend - Quarterly leads-generated trend - Quarterly cost-per-lead trend - Quarterly attributed-revenue trend Build a single combined chart showing each metric over the 5 quarters. Highlight which channels are clearly improving vs. declining. End with 'three things to ask the CMO based on this data.'

Best use cases

  • Year-over-year or quarter-over-quarter reviews
  • Cross-team performance benchmarking
  • Before/after intervention analysis
  • Multi-product portfolio analysis
Schema drift between files breaks this — if your Q1-2025 file had different column names than Q1-2026, Analyst might silently misinterpret. Spot-check the input mapping.
Time savings: Cross-quarter analysis: half-day → 15 min.

Workflow 03 Statistical analysis without a stats degree

3

Significance, correlation, regression — explained

Analyst can run real statistical analysis and explain the results in plain language. This unlocks proper analysis for non-statisticians.

The prompt that works

Stats with explanationI want to know if our customer-success outreach actually reduces churn. Data: - We track which customers got 'high-touch' onboarding vs. standard - 18 months of churn data per customer - Customer attributes: company size, industry, contract value, signup date Analyze: 1. Is there a statistically significant difference in churn rate between high-touch and standard onboarding? 2. Once you control for company size, industry, and contract value, does the high-touch effect still hold? 3. What's the magnitude — if we moved 100 standard-onboarded customers to high-touch, how many fewer would we expect to churn? 4. What are the confidence intervals? How sure are we? Explain the test you used and why. Plain-English summary at the end for a non-statistical leader.

Best use cases

  • Program effectiveness analysis
  • Marketing attribution
  • Customer success ROI measurement
  • A/B test analysis with proper controls
For genuinely high-stakes statistical decisions, have a real analyst verify Analyst's methodology. AI can run correct math on the wrong test.
Time savings: A real statistical analysis: days of analyst work → 30 min Analyst + review.

Replace one analyst request you'd normally make

Think of the last time you asked your data team for a 'quick analysis.' Take that exact same data and request, frame it as an Analyst prompt. Compare the result. The skill isn't replacing analysts — it's knowing when self-service works.

Back to Copilot Mastery → Try a prompt in the playground →

What you can do now

  • Use Analyst when you have data and need insight, not just a chart
  • Verify column understanding before trusting the analysis
  • Combine multiple files for cross-period or cross-team comparisons
  • Ask Analyst to explain the test it chose for statistical work
  • Escalate to a real analyst for high-stakes statistical decisions
Pro
Up next in Copilot Mastery

Lesson 03 · Facilitator — meeting management in Teams

Facilitator runs alongside you in Teams meetings — taking notes, capturing decisions, moderating turns. The agent that means you never have to choose between participating and documenting. See the track →