Skip to main content
Use chat when you want Sofie to act like an analyst for a CoSheet. Ask for the analysis you need, then decide whether Sofie should apply formulas, create charts, create pivot tables, or summarize results in a CoDraft.

​
Start with a data profile

Ask Sofie to inspect the CoSheet before changing it. 
Analyze this CoSheet. Profile the data shape, column meanings, missing values, duplicate IDs, outliers, inconsistent units, date issues, and suggested next actions.
Good first checks:
  • Number of rows and columns.
  • Header quality.
  • Missing values.
  • Duplicate identifiers.
  • Date formats.
  • Unit consistency.
  • Numeric ranges.
  • Outliers.
  • Categorical values.
  • Formula opportunities.

​
Ask for formula suggestions

Suggest formulas for this CoSheet. Focus on status, days open, percent difference from target, pass/fail logic, and missing required fields. Show formulas before applying them.
Apply only after review: 
Apply the Status formula to the Status column after showing me the first three expected results.

​
Ask for pivot tables

Suggest pivot tables for this CoSheet. Include source range, rows, columns, values, filters, aggregation, and the question each pivot answers.
Good pivot questions:
  • Open deviations by product and category.
  • CAPA actions by owner and status.
  • Findings by batch record section.
  • Average assay result by campaign.
  • Risk count by severity and detectability.

​
Ask for charts

Suggest charts for this CoSheet. Include chart type, data range, title, axis labels, and what the chart should help reviewers see.
Ask Sofie to create a chart only after you agree with the data range. 
Create the recommended line chart using the campaign trend range. Include clear title, axis labels, and data labels only if they remain readable.

​
Ask for data quality review

Use this before drawing conclusions. 
Review this CoSheet for data quality issues. Return a table with issue type, affected rows or columns, example value, likely impact, and recommended fix.
Common issues:
  • Missing values.
  • Inconsistent units.
  • Mixed date formats.
  • Duplicate identifiers.
  • Outlier values.
  • Text in numeric columns.
  • Hidden assumptions in calculated fields.

​
Ask for life sciences analysis

Examples: 
Analyze this CAPA effectiveness CoSheet. Compare baseline and post-implementation metrics, flag rows with insufficient observation window, and list conclusions that need SME review.

Analyze this batch record findings tracker. Group findings by section, identify recurring issue types, and suggest a chart for QA review.

Analyze this risk assessment table. Check severity, occurrence, detectability, rationale, and recommended actions for inconsistencies.

​
Create a summary tab

Ask Sofie to create a reviewer-friendly tab. 
Create a Summary tab with data quality findings, key trends, outliers, recommended pivots, recommended charts, and open reviewer questions.

​
Move results into a CoDraft

After analysis, ask Sofie to create narrative output. 
Turn this CoSheet analysis into a CoDraft section. Include data source, method, key findings, limitations, assumptions, and open SME questions.

​
Export after review

Before export, ask: 
Review this CoSheet before export. List formula risks, stale pivot tables, unclear chart labels, missing units, hidden assumptions, and rows needing human review.
Then export through CoSheet or ask Sofie: 
Export this CoSheet to Excel after I approve the review checklist.