Logan Hutchinson
In my BA work, I’m increasingly doing data analysis that previously required dedicated data analysts. The combination of accessible AI tools, mature self-service BI platforms, and growing business demand for evidence-based decisions means I now answer questions like “what is the conversion rate by segment last quarter?” without filing tickets to a data team. In my view this is good for BAs - more agency, more strategic value, faster decision cycles. It is also a skill area where I see most BAs have meaningful gaps.
AI Data Analysis for Business Analysts: SQL, Excel & Beyond
I wrote this guide for BAs who want to expand their data analysis capabilities with AI without becoming full-time data analysts.
The Data Analysis BAs Actually Do
BA data analysis typically falls in these categories:
- Volume questions: how many users do X per day?
- Comparison questions: how do segments differ?
- Trend questions: how is metric X changing over time?
- Diagnostic questions: why did metric X drop?
- Cohort questions: how do new users behave vs old?
- Funnel questions: where do users drop off?
Most BA data work is descriptive and diagnostic. Advanced statistical or predictive work usually goes to dedicated data analysts.
AI for SQL Without Becoming a SQL Expert
AI tools generate SQL from natural language:
- Hex’s AI assistant.
- Mode’s AI features.
- Snowflake Cortex.
- ChatGPT or Claude with schema description.
A useful prompt:
“I have a database with these tables: [paste schema]. Write SQL to answer: ‘What is the conversion rate from sign-up to first import, segmented by acquisition source, for users who signed up in Q1 2026?’”
The BA reads the SQL, understands what it does, runs it, and validates the output. AI handles the syntax; the BA owns the question and the interpretation.
Strong BAs build SQL fluency over time. AI accelerates but does not replace this. Reading SQL is a fundamental BA skill in 2026.
AI for Excel: Beyond Pivot Tables
Excel with AI (Copilot, ChatGPT) handles:
- Generating formulas from natural language.
- Building pivot tables.
- Suggesting chart types.
- Explaining what existing formulas do.
- Cleaning messy data.
A useful prompt:
“I have a column of customer signup dates and a column of first purchase dates. Write the Excel formula to compute the median time-to-purchase, excluding outliers more than 2 standard deviations from the mean.”
The BA validates the formula and uses it. Excel remains the most-used analysis tool by BAs in 2026.
AI-Augmented BI Tools
Self-service BI platforms with AI:
- Hex: notebook-style with AI assistant.
- Mode: similar, stronger SQL.
- Tableau Pulse: AI-driven insights on Tableau dashboards.
- Power BI Copilot: AI for the Microsoft stack.
- Looker Studio with Gemini: Google’s stack.
For most BAs, Hex provides the broadest capability for ad-hoc analysis. For BAs in companies already on Tableau or Power BI, those tools’ AI features suffice.
The Question-First Approach
Strong data analysis starts with a clear question. AI cannot rescue unclear questions. Patterns:
- Bad question: “Tell me about user behaviour.”
- Good question: “What percentage of users who sign up complete onboarding within 7 days, and how does that vary by acquisition source?”
A useful prompt:
“I want to understand whether the new pricing model is working. What 5-7 specific data questions would I need to answer to address this? For each: question, data sources, expected analysis.”
The BA refines the questions before doing the analysis.
Common Patterns BAs Encounter
Conversion rate analysis: numerator / denominator with appropriate segmentation.
Cohort retention: cohort by signup month, retention curve.
Funnel analysis: stages with drop-off rates.
Segmentation: defining and comparing segments.
Trend analysis: metric over time with seasonality.
A/B test reading: control vs treatment with statistical significance.
For each pattern, AI generates the SQL or BI query. The BA validates and interprets.
Tools That Help
| Tool | Use case |
|---|---|
| Hex | Notebook-style ad-hoc analysis |
| Mode | Stronger SQL, BI features |
| Tableau Pulse | AI on existing Tableau |
| Power BI Copilot | Microsoft stack |
| Excel with Copilot | Day-to-day Excel work |
| ChatGPT / Claude with data tools | One-off file analysis |
| Mixpanel AI Insights | Product analytics specifically |
For most BAs, Hex plus Excel with Copilot covers most needs.
The 60-Day Skill Build
Days 1-15: SQL fluency. Use AI to generate SQL, then read and understand each query. Build pattern recognition.
Days 16-30: BI tool fluency. Pick one (Hex or your company’s tool). Build 5 dashboards.
Days 31-45: statistical literacy basics. Mean, median, percentiles, standard deviation, statistical significance.
Days 46-60: question framing. Practise turning vague questions into specific data questions.
By day 60, the BA can answer most BA-level data questions independently.
Common Failure Modes
These are the failure modes I see most often when BAs lean on AI for data analysis. What I tell BAs starting out: the first two on this list account for most of the bad analyses I’ve reviewed.
- Trusting AI-generated queries blindly. In my experience this is the most common error. Always read and validate.
- Wrong question, fancy answer. AI cannot fix unclear questions.
- Statistical illiteracy. AI does the math; the BA needs to understand what means significant.
- Privacy violations. Pasting customer data into consumer-tier tools.
- Skipping data quality validation. AI assumes clean data. I’ve seen reality be much messier.
- Over-engineering analyses. Most BA questions need basic analysis, not advanced techniques.
- Not communicating uncertainty. Real analyses have confidence ranges. Communicate them.
Statistical Concepts BAs Should Know
A working knowledge of:
- Mean vs median: when each is appropriate.
- Percentiles: especially p50, p90, p95.
- Standard deviation: and what it tells you about distribution.
- Statistical significance: confidence intervals, p-values basics.
- Correlation vs causation: the most important distinction in business analysis.
- Sample size adequacy: when small samples mislead.
- Confounding variables: why apparent relationships disappear.
These concepts let BAs read AI-generated analyses critically rather than accepting them blindly.
Building a Personal Analysis Library
Maintain a personal library of:
- Common SQL patterns for your data sources.
- Excel formulas that solve recurring problems.
- BI dashboard templates by question type.
- Statistical methods reference.
- Question-framing checklists.
This library compounds across years. After 24 months, most BA-level analysis questions can be answered in under an hour.
- 25+ years at the intersection of complex systems and human needs
- SAFe Certified Product Manager (Scaled Agile)
- Certified Scrum Master (Scrum Alliance)
- Pragmatic Marketing Certified PM
- Expert in AI-driven product strategy and delivery
- Deep experience scaling products at enterprise level
- Known for balancing detail-orientation with big-picture thinking
- Passionate about engineering tailored solutions for customer needs
Questions people ask
Published