Lesson 12 - The Future Analyst – Working With and Around AI
Learn how use AI as Your Co-Pilot in Data Analytics
Estimated Read Time: 1 - 1,5 Hours
Learning Goals
In this lesson, you will learn:
Technical & Analytical:
- Learn how to apply AI to perform tasks across the analytics pipeline.
AI Literacy & AI-Proofing:
- Compare strengths and weaknesses of different AI platforms for data analytics.
- Develop strategies to use AI responsibly, maintaining credibility and trust in professional contexts.
- Recognize that AI is a co-pilot — it amplifies your skills but cannot replace structured thinking or business sense.
Artificial intelligence is not the future; it is the present. It is transforming industries, reshaping business decisions, and re-imagining how analysts approach their tasks. For those working in data analytics, AI is no longer merely another tool — it is a genuine game-changer. It can compress hours of manual effort into minutes, surface patterns across thousands of rows instantly, and even generate slide decks or reports tailored for executives.
Yet amidst this transformation lies one enduring truth: AI is not here to replace you. It is here to empower you. The professional value of the analyst now lies not in mechanical execution but in structured judgment — in asking the right questions, verifying evidence, and framing insights responsibly.
1. Excel’s AI – “Analyze Data”
Excel’s Analyze Data (formerly known as “Ideas”) is a built-in AI feature available in modern versions of Microsoft Excel (Microsoft 365). It allows analysts to explore datasets using natural language queries. Rather than building formulas from scratch, you can ask plain English questions like:
“Are there any outliers in Sales?”
Excel will then scan your data using behind-the-scenes statistical techniques to return charts, summaries, or highlighted insights.
But here’s the crucial point: the feature suggests patterns—it does not confirm truths. As an analyst, your responsibility is to verify every suggestion, interpret its business relevance, and decide whether any action is necessary.
1.1. A Practical Scenario: Detecting Outliers in Rossmann Sales
Recall from Lesson 5, suppose you’re working with rossmann-sales.xlsx, and you’ve already cleaned and standardized the dataset.
Accessing Analyze Data Feature
- Open your Rossmann dataset (if not open already).
- Click the Analyze Data button on the Home tab. (Figure 1)
A panel appears on the right side of your screen.
Now you type: “Are there any outliers in the Sales column?”
What Excel Does Behind the Scenes
- It calculates descriptive statistics: mean, standard deviation, quartiles
- It looks for values significantly different from the rest of the dataset
- It creates charts (like histograms or line graphs) and summaries such as:
“For ‘DayOfWeek: 6’, ‘Sales’ changed significantly from ‘Date’: 14.12.22 to 21.12.22 with 6 outliers.”
Consider Figure 2 that shows the prompt and the results.
Note: This is not a final verdict. It is a lead — an insight you must investigate and validate using your domain knowledge and Excel tools (e.g., calculating standard deviation, applying filters, checking against store promotions or holidays).
1.2. Step-by-Step: From Insight to Analyst Judgment
Let’s walk through the complete thinking process.
1. Ask a Specific Question
Avoid vague prompts. Instead of “What’s wrong with this data?”, ask:
- “Are there stores with unusually high sales?”
- “Which stores had the largest weekly fluctuation in customer counts?”
- “Are there dates where Sales values spike significantly?”
2. Interpret the Output
Excel might highlight that Store 98 had €72,000 in sales on a single day—much higher than usual. Is this an error? Or a holiday spike?
At this stage, treat this like a hypothesis, not a conclusion.
3. Verify with Standard Techniques
Use traditional Excel tools to confirm or refute the insight:
- Use AVERAGE() and other statistical functions to calculate the normal range for that store
- Use conditional formatting to highlight extreme values
- Cross-reference with the Promo and StateHoliday fields—was there a special event on that date?
4. Check for Patterns
Is the anomaly isolated?
- Do other stores show similar spikes on the same day? (Suggests a holiday)
- Does this store often show such variation? (Suggests volatility or error)
5. Document Your Conclusion
Whether you keep the value, adjust it, or flag it for review, record your decision in the cleaning log.
2. External AI Tools
If you do not have Excel’s built-in AI tool, you can also use one of the external tools. As you know, there are many AI tools out there, some of those can also be used for data analytics. Sometimes it becomes overwhelming to decide which one to use. However, names such as ChatGPT, Claude, Perplexity, Gemini, and DeepSeek matter less than understanding their respective strengths and limits.
Different AI systems excel in different modes of analysis. For instance:
- ChatGPT is versatile and strong in explanatory narrative, though at times overconfident.
- Claude offers extended reasoning and is particularly good with structured documents.
- Perplexity integrates real-time search, making it valuable when freshness is essential.
- Gemini is deeply integrated with the Google ecosystem and often excels at technical tasks.
- DeepSeek emphasises transparency in its stepwise reasoning, though it may be less polished in communication.
- Excel’s own Analyze Data feature (introduced in Lessons 4 and 5) – provides quick suggestions for trends, patterns, and charts directly within the workbook.
The key is not to master every available platform, but to maintain a working diversity: using at least two tools in parallel allows you to cross-check results, hedge against quota limits, and counteract model-specific biases.
Rather than asking, “Which AI is best?”, the more precise question is, “Which AI is best for this task, in this context?”
Let’s explore one of these tools to further explore the Rossmann Sales dataset.
3. Rossmann Dataset in Action
Let us apply this to the Rossmann Sales dataset. Recall that it contains daily sales, customer counts, promotions, state holidays, and store attributes.
Step 1 – Upload and Verify
Using your chosen AI tool, upload rossmann-sales.xlsx. Begin by validating:
“Read the file and tell me how many rows and columns are present. List the column names.”
If the AI misstates the size or structure (e.g., claiming 200 datapoints instead of 91,262), correct it and insist upon re-evaluation. This simple step illustrates a core truth:
AI can mis-parse files. Verification lies with you.
Step 2 – Ask Business Questions
Next, pose the same business questions we solved manually in earlier lessons:
- “Which stores had the highest average sales during promotions?”
- “Do state holidays significantly affect sales patterns?”
- “What correlation exists between customer count and sales?”
Compare the AI’s responses to the outputs you generated in Lesson 4 (Cleaning), Lesson 5 (Outliers), and Lesson 6 (Descriptive Analysis). Where do the answers align? Where do they diverge?
You can also provide the stores’ file and ask AI to merge information, like you did in Lesson 8 on Lookups.
Step 3 – Explore Insights at Scale
Push further with exploratory prompts:
- “Identify which stores consistently underperform compared to peers.”
- “Summarise how promotions and school holidays interact in affecting sales.”
AI may surface insights you hadn’t considered, but it may also hallucinate patterns.
Treat AI not as a truth machine but as an accelerator for inquiry.
Step 4 – Build an AI-Generated Report
Extend the experiment by requesting a stakeholder-ready artefact:
“Create an executive summary with key Rossmann sales trends, risks, and recommendations. Format it as a 5-slide presentation.”
Examine whether the slides are consistent with your manual analysis. Revise outputs where the AI over-generalises or omits business context. This illustrates AI’s role in storytelling: it can deliver a polished draft, but the strategic framing must always originate with you.
4. Best Practices for Analysts Using AI
1. Validate First
Confirm data import, row counts, and column names before proceeding.
2. Define Clear Business Questions
Broad prompts yield shallow answers; precise queries bring sharper insights. Frame your analysis goals before engaging AI tools to keep results focused.
3. Cross-Check Outputs
Always compare AI responses with your manual calculations or Excel outputs. Check formulas, filters, and visualizations for accuracy and business relevance.
- Check the Data Range: Is AI analyzing the correct time period? For example, does the pivot table include all 3 years of Rossmann data or just a subset?
- Verify Filters and Groupings: Are promotions correctly flagged? Are store types properly classified?
- Assess Completeness: Are key metrics missing? For example, is customer count included when analyzing sales to understand sales per customer?
4. Stay Aware of Limitations
AI cannot truly “know” your business context; it only infers from patterns. Use AI to handle repetitive or computational tasks but apply your judgment to interpret and contextualize results.
5. Protect Sensitive Data
Never upload confidential or customer-identifying information.
6. Document AI Usage
Note which parts of your analysis leveraged AI assistance to demonstrate transparency and rigor. Transparently documenting when and how you use AI tools demonstrates analytical integrity and ethical responsibility, distinguishing you as an elite analyst who blends technology with sound judgment — a valuable aspect of your personal brand.
5. Guardrails for Responsible Use
AI is fast — but it is not flawless. YOU are accountable for the reports you send out to your stakeholders.
You can use all the AI that you want, the ultimate responsibility of correctness of results lies with you. Not AI.
Three principles should guide your use of AI tools in analysis:
- Verification before Adoption – Always test AI-generated outputs against your own calculations and logic.
- Task–Tool Alignment – Choose the AI tool whose core strengths match the analytical need.
- Transparency in Communication – When AI has assisted your work, note this clearly, and retain ownership of the conclusions.
Remember:
YOU are the one responsible for the accuracy and trustworthiness of what you share. Not AI!
6. AI Integration Throughout the Pipeline
AI tools can enhance your effectiveness at every pipeline stage, but human judgment remains essential:
6.1. Where AI Excels
- Requirements: Brainstorming potential business questions
- Collection: Automating data extraction and formatting
- Preparation: Identifying patterns in messy data
- Analysis: Performing complex calculations quickly
- Visualization: Generating initial chart drafts
6.2. Where Humans Add Value
- Requirements: Understanding organizational priorities and constraints
- Collection: Navigating organizational relationships and permissions
- Preparation: Making judgment calls about data quality and handling
- Analysis: Interpreting patterns within business context
- Visualization: Crafting compelling narratives that drive action
6.3. Building AI-Proof Skills – Your Edge as a Human
Focus on developing skills that complement rather than compete with AI:
Strategic thinking and business acumen:
AI lacks knowledge of Rossmann’s strategic priorities and operational nuances. For example, it may not recognize the importance of promotional effects or store segmentation unless explicitly guided.
Stakeholder communication and relationship building:
There is no alternative to human connection and relationships.
Ethical reasoning and judgment:
AI-generated outputs can omit key variables, apply incorrect aggregations, or misinterpret data relationships. You must critically evaluate AI suggestions before accepting them.
Creative problem-solving and hypothesis generation:
Thinking outside the box requires context and knowledge not directly avaiable in the data.
Project management and cross-functional collaboration:
Prioritization, resource management, alignment across teams and people management are too diverse and complicated, at the moment for AI to handle. When we need to know the rules and know when to break them – this is where your human instinct will come in.
6.4. Reflection: What AI Can—and Can’t—Do
|
AI Can |
AI Can’t |
|
Analyze structured data |
Decide what matters to your client |
|
Summarize unstructured reviews |
Identify the business use case |
|
Generate visual-ready slides |
Replace your domain understanding |
|
Suggest ideas or hypotheses |
Judge feasibility or ethics |
AI is a multiplier of your skills, not a replacement. It is powerful only when paired with your business judgment, critical thinking, and ethical lens.
Summary
In this lesson, you learned:
- AI accelerates analysis but does not replace human judgement.
- How to integrate AI into data analytics pipeline while verifying results against Excel.
- How to compare and select AI tools based on task strengths and limitations.
- Best practices for responsible AI use: validate, be specific, cross-check, protect data, and document outputs.
Suggested Readings & References
Exercise
Estimated Time to Complete: 3-4 hours
Task 1: Working With and Around AI
Business Context
You’ve now worked through the full analytics pipeline with Walmart data: cleaning → outliers → integration → storytelling.
In this exercise, you’ll see how AI handles part of that workflow, compare it to your own work, and reflect on your unique value as an analyst.
Sub-Task 1 – Automate with AI
Choose one stage from the pipeline below and redo it using AI (Excel Copilot, Ideas, or another assistant).
Option A – Data Cleaning (Lesson 4)
- Ask AI to detect missing values, fix inconsistent dates, or flag formatting issues in walmart-sales.
Option B – Outlier Detection (Lesson 5)
- Ask AI: “Find outliers in Walmart weekly sales and explain what they mean.”
Option C – Data Integration (Lesson 8)
- Ask AI to join walmart-sales with walmart-features by Store and Date.
Option D – Storytelling (Lesson 11)
- Ask AI to “Summarize Walmart’s holiday sales performance for executives in 3–4 sentences.”
Deliverable: Save the AI’s raw output (Excel Worksheet).
Sub-Task 2 – Compare & Reflect
Compare the AI’s result to your own earlier work from that lesson.
- What was useful about the AI’s approach?
- What was missing, wrong, or oversimplified?
- Which part took less time with AI?
- Which part required your judgment as an analyst?
Deliverable: Write 5–6 bullet points of reflection.
Sub-Task 3 – Analyst Value-Add
Write a short executive memo (3–5 sentences):
- Explain what you learned about AI’s role in analytics.
- Highlight where you would trust AI where human expertise is essential.
- End with one recommendation for Walmart executives on how to responsibly use AI in decision-making.
Deliverable: Executive memo + recommendation (slide deck or a document report)
Task 2 – Choosing Your AI Partner as an Analyst
Business Context
The market offers many AI tools — ChatGPT, Claude, Gemini, Perplexity, DeepSeek, and more. Each claims to help analysts, but their strengths and weaknesses vary depending on the task (e.g., coding formulas, summarizing, business insight, data storytelling).
As a future analyst, you need to critically decide which AI tool adds real value to your workflow — and when to rely on it vs. when to rely on yourself.
Sub-Task 1 – Explore AI Tools
Pick at least two AI tools (e.g., ChatGPT, Claude, Gemini, Perplexity, DeepSeek or another of your choice).
- Upload necessary files and use the same prompt/question with both tools. Suggested prompts (choose one of the following or create another):
- “Summarize Walmart’s holiday sales trends in 3 sentences, highlighting risks for executives.”
- “Suggest how to clean and prepare the Walmart features dataset for analysis in Excel.”
- “Explain whether high fuel prices could impact Walmart sales, using reasoning and evidence.”
Deliverable: Note the responses from each AI tool in a doc or slide deck.
Sub-Task 2 – Evaluate Strengths & Limitations
For each tool you tried, answer:
- What did the tool do well? (clarity, speed, depth, technical correctness, business framing)
- What were the limitations? (oversimplification, hallucination, lack of Excel specificity, shallow analysis, vague recommendations)
Deliverable: A comparison table with at least 3 strengths + 3 limitations per tool.
Sub-Task 3 – Choose Your Go-To Tool
- Based on your experiment, decide which AI tool you would rely on most as an analyst.
- Explain your reasoning in 3–5 sentences:
- Which types of tasks does it handle best?
- Where do you need to double-check its outputs?
- How will you combine it with your own skills moving forward?
Deliverable: A short written recommendation.
Submission Guidelines
Submit your solution as a worksheet and a presentation slide deck / a document report:
Workbook:
- appropriate worksheets with analysis and visuals, if needed
Filename Format:
- YourName_Lesson12_Walmart_AI.pptx
- YourName_Lesson12_Walmart_AI.xlsx
- YourName_Lesson12_Walmart_AI.docx
When you’re ready, submit your completed exercise to the designated folder in OneDrive. Drop your mentor a note about submission.
Important: Please scan your files for viruses before uploading.
Submission & Resubmission Guidelines
- Initial Submission Format: YourName_Lesson12_…
- Resubmission Format:
- YourName_Lesson12_…_v2
- YourName_Lesson12_…_v3
- Rubric Updates:
- Do not overwrite original evaluation entries
- Add updated responses in new “v2” or “v3” columns
- This allows mentors to track your improvement process
Evaluation Rubric
|
Criteria |
Exceeds Expectations |
Meets Expectations |
Needs Improvement |
Incomplete / Off-Track |
|
Task 1.1 AI Use |
Uses AI effectively, documents output clearly, may test multiple prompts or outputs. |
Provides one AI output relevant to task. |
Provides incomplete/unclear AI output. |
No AI output provided. Submission is plagiarized. |
|
Task 1.2 Reflection & Critique |
Sharp, multi-dimensional critique (5–6 points) covering technical accuracy, statistical nuance, business implications, and AI limitations. |
Reasonable critique (at least 3–4 solid points) showing awareness of oversimplification or gaps. |
Superficial critique with fewer than 3 relevant points. |
No critique provided. Submission is plagiarized. |
|
Task 1.3 Executive Memo |
Concise, executive-ready, highlights both AI’s strengths and human analyst’s role, plus a creative recommendation. |
Clear summary with at least one human vs AI insight and one actionable recommendation. |
Generic note, misses the human–AI distinction, or weak recommendation. |
No memo provided. Submission is plagiarized. |
|
Task 2.1 Exploration |
Tests 2+ tools with thoughtful prompts; includes full outputs clearly. |
Tests 2 tools with provided prompt(s) and shows outputs. |
Only 1 tool tested or outputs incomplete. |
No tool outputs provided. Submission is plagiarized. |
|
Task 2.2 Evaluation |
Comparison table is detailed, with nuanced critique (3+ strengths & 3+ limitations per tool). |
Comparison covers basic strengths/limitations (at least 2 per tool). |
Evaluation is vague, missing, or unbalanced (focuses only on positives or negatives). |
No evaluation provided. Submission is plagiarized. |
|
Task 2.3 Recommendation |
Clear, strategic, and personalized; explains how chosen tool fits workflow, with awareness of risks/limitations. |
Provides a logical choice with reasoning tied to tasks. |
Choice is generic, weakly justified, or ignores limitations. |
No recommendation provided. Submission is plagiarized or thoughtlessly AI-generated. |
Got Feedback?
Contact
Talk to us
Have questions or feedback about Lumen? We’d love to hear from you.