Corporate AI6 min read

How Do Data Analysts Use AI?

On this page

Data analysts use AI to write and explain Excel formulas and SQL queries, plan data cleaning, suggest and build charts, and explain results in plain words for managers. The analyst still checks every number against the source, because an AI tool can give a confident formula or total that is wrong. Customer, staff and client data stays out of public tools unless it has been masked.

This guide is for anyone who works with data in Kenya: analysts, M&E officers, finance and sales teams, and managers who build their own reports. Each sample prompt follows the four Cs we teach: Clarity (the task), Context (your data), Constraints (the rules) and Customisation (the format).

Six tasks where AI helps data analysts today

1. Writing and explaining Excel formulas

You rarely need to paste real data. Column names are usually enough.

Write an Excel formula for column F. Context: Excel for Microsoft 365; sheet "Sales" has A Date, B Branch, C Product, D Units, E Amount (KES). Constraints: one formula that returns total Amount for the branch in H2 and the month in H3; explain each part in plain English; list two cases where it would give a wrong answer.

If your licence includes Copilot in Excel, Microsoft says (as of November 2026) that it uses Excel's own tables, charts, PivotTables and formulas to complete requests, with edit, plan and chat modes. Chat mode analyses your workbook without changing it. Our step-by-step guide on how to use Copilot in Excel covers licences and setup.

Check: test the formula on rows where you already know the answer, including some that should fail.

2. Writing SQL queries

Give the tool your table structure, never your connection details.

Write a SQL query for PostgreSQL. Context: tables customers(id, county, joined_on) and payments(id, customer_id, amount_kes, paid_at). Constraints: total payments per county for October 2026; include counties with no payments; explain the join and any risk of counting a payment twice.

Check: run it first on a small sample or a test copy, using a read-only account. Watch for joins that multiply rows, missing NULLs and date ranges that cut off the last day of the month. Never paste passwords, API keys or connection strings into a prompt.

3. Planning data cleaning

Here are my column headers and 10 sample rows with names removed. List every cleaning problem you can see and the steps to fix each one in Excel or Power Query. Context: phone numbers appear as 07..., 7... and +2547...; dates come from two systems. Constraints: do not delete any rows; tell me what to check before and after each step.

Check: write down the row count and the column totals before and after cleaning. If they change and you cannot explain why, stop.

4. Explaining results to people who do not read spreadsheets

Write a five-line summary of the table below for our operations manager. Context: monthly sales by branch, July to October 2026. Constraints: use only these numbers; state the biggest change first; say what the data cannot tell us; no guesses about causes. Customisation: simple English, no jargon.

Check: match every figure in the summary to the table. AI tools can round, swap two branches or describe a fall as a rise.

5. Choosing and building charts

Which chart best shows this to a board: monthly revenue for four branches over 12 months? Give me two options, say what each is good and bad at, and the exact steps to build the better one in Excel.

Microsoft says Copilot in Excel can create and edit charts and PivotTables with editable links to the source data.

Check: look at the axis start, the labels, the units (KES or KES thousands) and whether the colours still make sense when printed in black and white.

6. Reviewing a formula or query someone else wrote

Explain this formula line by line, as if to a new analyst. Then tell me what inputs would break it. Formula: [paste].

This is often the quickest win: inherited workbooks full of formulas nobody understands.

Check: test the explanation by changing one input and seeing whether the result moves the way the tool said it would.

How to check every number

AI writes a wrong answer with the same confidence as a right one. Build these habits:

  1. Reconcile totals. Your summary total must equal the source total.
  2. Count rows before and after every cleaning or join step.
  3. Spot-check ten rows by hand, picked at random.
  4. Recalculate one key figure a second way, such as a PivotTable against a SUMIFS.
  5. Take official figures from the official source. If your report needs October 2026 inflation, use the Kenya National Bureau of Statistics release, not a chatbot's memory.
  6. Keep a note of which steps AI helped with, so a reviewer knows where to look.

What AI should not do for a data analyst

  • See personal data in a public tool. Payroll, M-PESA statements, ID numbers, health records and customer lists stay out. Use masked or made-up sample rows.
  • Hold your credentials. No passwords, API keys or database connection strings.
  • Produce a final number you have not checked.
  • Pick the method for a formal study. For statistics that go into donor, regulatory or academic reports, the method is a human decision, made with someone qualified.

Data protection in Kenya

Kenya's Data Protection Act, 2019 is regulated by the Office of the Data Protection Commissioner (ODPC). Its stated mandate includes regulating how personal data is processed and making sure processing follows the principles in section 25 of the Act. For an analyst, that means three habits: mask before you prompt, use tools your organisation has approved, and keep raw files in approved storage. Our guide to AI and Kenya's Data Protection Act explains the duties, and NGO teams can read How Can NGOs Use AI for M&E Data Analysis?

How FuKazee trains data teams

Our Excel and AI training teaches solid Excel first and AI on top. The live page lists cleaning data with Power Query, lookups with XLOOKUP and INDEX with MATCH, PivotTables and simple dashboards, reading a formula before trusting it, Copilot in Excel on licences that include it, and a free route that uses only your column headings. Teams work on their own workbooks with sensitive columns masked, in groups of about 20 or fewer. Finance teams can also look at AI training for finance teams. We are rated 4.9 from 38 Google reviews, as shown on our reviews page on 14 November 2026.

Questions people ask

Will AI replace data analysts?

AI makes formulas, queries and first drafts faster. Someone still has to know the business, choose the right question and stand behind the numbers. In our view, those skills matter more as drafts get easier.

Can I upload client data to ChatGPT?

Not raw personal or confidential data. Use a masked copy, or describe the columns and ask for the formula or query instead. Follow your organisation's approved tools list.

Is Copilot in Excel free?

It depends on your Microsoft 365 plan and your organisation's settings, and Microsoft changes the plans often. Our Copilot in Excel guide explains how to check what you have.

Do I still need to learn Excel and SQL?

Yes. If you cannot read a formula or a query, you cannot tell whether the AI's version is right.

Can AI do statistics for my report?

It can run calculations and explain tests. You still need to check its assumptions, and for formal reports, agree the method with someone qualified.

Sources

All sources last checked 14 November 2026.

Where FuKazee fits in

FuKazee teaches teams to use AI on the spreadsheets and reports in front of them, with checking habits that keep every number honest.

Email us at lead@fukazee.com

Want guided support instead?

Plan AI training around your team's real work. Answer a few quick questions about your sector, team size and goals, and we will recommend the right programme and format. Sessions are built on your workflows and include a usage policy.

Rodney MainaFounder & Director, FuKazee

Rodney has run FuKazee's training since 2022 and still teaches inside every programme line: corporate sessions, the postgraduate cohort and the KCSE examiner webinars. First Class Honours in Journalism and Mass Communication, University of Nairobi.