EKALAVYA ACADEMY | SHORT BOOKS
When a data shortcut becomes the rule
How a quick fix turns into habit, how to judge data quality, who owns data and how to make a hidden step visible.
Ekalavya Academy by Almost Magic
Short Book 5 of 10 | Free to read and share
How to read this book
Before You Start
A few slides explain one idea in full sentences.
A short, made-up example gives the idea a face.
A ten-minute exercise in the tools you already use.
Three questions. Guidance is at the back, not next to the question.
A four-sentence summary to come back to.
Two levels in every chapter. The main slides are written so a first-time reader can follow them. A "Going deeper" slide gives the same idea at the level an experienced reader would argue about.
This book works with any AI tool you already have. It never asks you to buy or sign up for a particular product.
Where this book sits in the series
The Plan
| Book 1 | What AI is, and what it is not | Foundations |
| Book 2 | Productivity with AI: writing, summaries, research, meetings | Practical use |
| Book 3 | From one task to a repeatable workflow, and AI deputies | Workflow automation, agents |
| Book 4 | Handling data safely | Data handling |
| Book 5 | When a data shortcut becomes the rule | Data quality and governance |
| Book 6 | Staying secure with AI | Security |
| Book 7 | Rules, risk and what you must know | Risk and compliance awareness |
| Book 8 | Your AI rules on one page | Policy and accountability |
| Book 9 | When AI gets it wrong | Recovering from AI mistakes |
| Book 10 | When not to use AI, and who stays in charge | Judgement and oversight |
Each book works on its own. Read them in order or pick the one you need.
In this book
Contents
How a shortcut becomes the rule
Is the data any good?
Who owns the data?
Making the invisible visible
Guidance and answers for every question and exercise
How a shortcut becomes the rule
Chapter 1
By the end of this chapter you will be able to:
Explain how a one-off fix becomes a standing habit.
Spot a shortcut that nobody decided to keep.
Ask of any number: where does it come from?
Decide whether to keep, document or retire a shortcut.
Time: about 40 minutes, including the exercise.
A useful shortcut and a quiet habit
Learn
A shortcut starts as a good idea. A deadline is close, the data is messy and a quick fix works. The report goes out and nobody minds.
Next month someone reuses the file. The month after, that is simply how the numbers are made. At no point did anyone decide it should be permanent.
AI tools make this easier than ever. A messy column can be tidied in seconds, and nothing records what the tool changed.
IN PLAIN WORDS
Nobody decided. It just kept happening.Signs a shortcut has become the rule
Learn
Ask who set it up. If the answer is "it was always like this", look closer.
You cannot say exactly what was changed, or with what settings.
Reopening a number is possible but takes effort, so nobody does it.
People reuse the cleaned file rather than go back to the source.
Four questions to ask of any number
Learn
Name the source and every step between the source and the figure.
Name the person or tool that changed it, and when.
Could someone else get the same number from the same source?
If nobody would, nobody is checking.
Why speed hides doubt
Going Deeper
Time saved is easy to measure. A quiet loss of trust, such as more queries and more arguments about figures, shows up slowly.
A tidy dashboard looks sure of itself. Its confidence is a claim, not evidence.
AI can make weak data look neat and authoritative, which makes it harder to challenge.
Each month a shortcut runs undocumented, more decisions depend on it and it gets harder to unpick.
A short example
See It
Mei runs the finance side of a small manufacturer. Two years ago she used an AI tool to tidy a supplier-name column for a board pack.
The file has been reused for every pack since. When a new accountant asks why two similar supplier names are merged, nobody can say whether that was deliberate.
Mei realises that one quick fix has set the way a whole report is built.
THE LESSON
A fix that is not written down is a rule nobody chose.This is a made-up example for teaching.
Trace one number
Try It
YOUR TOOLS
Use your own report. You do not need an AI tool for this exercise.GUIDANCE
What to look for is in the Guidance section at the end.Check yourself
Check It
How does a one-off fix become a standing habit?
Name two signs that a shortcut has become the rule.
Why is a confident-looking dashboard not evidence?
Answer in your own words before you read the Guidance section at the end.
Chapter 1 in four sentences
Keep It
KEEP THIS
A shortcut starts as a good idea and becomes the rule when nobody decides otherwise. Signs: no known origin, no written steps, awkward checking and reuse of the cleaned file. Ask of any number: where does it come from, who changed it and can it be reproduced? Confidence is a claim, not evidence.
Is the data any good?
Chapter 2
By the end of this chapter you will be able to:
Name five qualities of good data.
Run quick checks on a data set.
Explain why AI cleaning needs a record.
Decide how good is good enough.
Time: about 40 minutes, including the exercise.
Five questions about quality
Learn
Does it match what really happened?
Is anything missing?
Is the same thing written the same way everywhere?
Is it up to date?
Can you tell where it came from?
Quick checks you can do by eye
Learn
Does the number of rows match what you expect?
Look for impossible values, such as negative quantities or dates in the future.
Look for the same record twice, or one thing spelled two ways.
Do the totals match another source you trust?
What AI cleaning changes
Learn
Asking an AI tool to tidy a column is quick. The tool may fix real errors, but it may also merge things that should stay separate, drop values it does not understand or change formats.
It does not tell you every change it made, and it may not make the same changes next time.
Keep the original, record what you asked for and compare before and after.
RULE OF THUMB
Never overwrite the original.Good enough for what?
Learn
Data does not need to be perfect. It needs to be good enough for the decision it supports.
A rough count may be fine for planning a roster. The same data may be too weak for a price change or a payroll run.
Say what the data will be used for, and let that set how much checking it needs.
REMEMBER
Quality depends on the use.Sampling and silent errors
Going Deeper
You cannot read every row. Check a random sample each time and record the error rate you find.
A wrongly merged name or a shifted date still looks like valid data. Checks must compare against something outside the cleaned file.
Source data changes over time, such as a new supplier format or a renamed field. A cleaning step that worked last year may quietly fail.
If some customers or periods are missing more often than others, summaries lean toward the ones that remain.
A short example
See It
Dev runs a small wholesaler. He asks an AI tool to standardise the product codes in a price list. The new list looks tidy.
Before using it he compares 30 random rows with the original. Four codes have changed in ways that point to a different product.
He keeps the original, fixes those four by hand and writes down the request he used, so he can repeat and check it.
THE LESSON
Compare a sample against the original.This is a made-up example for teaching.
Run four quick checks
Try It
YOUR TOOLS
Use your own data, kept in your own tools. Do not paste private data into an AI tool for this exercise.GUIDANCE
What to look for is in the Guidance section at the end.Check yourself
Check It
Name the five quality questions.
Why must you keep the original when an AI tool cleans data?
Why can the same data be good enough for one decision and not another?
Answer in your own words before you read the Guidance section at the end.
Chapter 2 in four sentences
Keep It
KEEP THIS
Good data is accurate, complete, consistent, current and traceable. Run quick checks: count, range, duplicates and totals. AI cleaning is quick but can change things silently, so keep the original and compare a sample. Good enough depends on the decision the data supports.
Who owns the data?
Chapter 3
By the end of this chapter you will be able to:
Explain what data governance means in a small business.
Name an owner, a fixer and the users.
Write a simple record of where a data set comes from.
Decide who may change it.
Time: about 40 minutes, including the exercise.
Governance in plain words
Learn
Data governance sounds grand. In a small business it means a few simple answers: who owns this data, who may change it, who uses it and who fixes it when it is wrong.
Without those answers, data tends to be everyone's problem and nobody's job.
Good governance is not a thick policy. It is a short page that people actually use.
IN PLAIN WORDS
Governance is knowing who is responsible.Three roles
Learn
Owner
Steward, or fixer
User
Relies on the data and reports problems, but does not change it alone.
A one-line record for each data set
Learn
What the data set is called.
Where it comes from, such as a system, a supplier or a form.
What is done to it before it is used, including any AI tool step.
The person who answers for it.
When someone last tested it against a source.
Who may change it?
Learn
Changes to important data should be visible. A change note says what changed, when, who did it and why.
When AI changes data, the same rule holds. A person asks for the change, a person checks it and the note says a tool was used.
Keep change rights narrow. Fewer people who can edit means fewer silent changes.
REMEMBER
Every change leaves a note.Governance without bureaucracy
Going Deeper
Write the record first for the data behind big decisions, such as pricing, payroll and customer lists.
Pick a regular time to review the records, such as every quarter, and put it in the calendar.
An owner answers for the data. That does not mean they caused every error.
Some data comes with legal duties, such as privacy rules. Note any that apply, and ask an adviser if unsure.
A short example
See It
Chloe runs a small events company. Her customer list sits in three places and is edited by four people.
After a mailing goes to the wrong person twice, she names the office manager as owner, limits editing to two people and writes a one-page record: source, steps, owner and last checked.
The next time an error appears, everyone knows whom to tell.
THE LESSON
A name next to the data is half the fix.This is a made-up example for teaching.
Write a data record
Try It
YOUR TOOLS
A table in a document is enough. Use any tool you like, or paper.GUIDANCE
What to look for is in the Guidance section at the end.Check yourself
Check It
What are the three roles in a simple data governance setup?
What should a change note say?
Why keep the number of people who can edit small?
Answer in your own words before you read the Guidance section at the end.
Chapter 3 in four sentences
Keep It
KEEP THIS
Governance in a small business means knowing who owns the data, who fixes it, who uses it and who may change it. Keep a one-line record for each important data set. Every change leaves a note, including changes made with an AI tool. A name next to the data is half the fix.
Making the invisible visible
Chapter 4
By the end of this chapter you will be able to:
Document a shortcut so someone else can run it.
Give it a name, an owner and a change log.
Reproduce a number from its source.
Decide to keep, change or retire it.
Time: about 40 minutes, including the exercise.
Name it, write it down
Learn
The first step with a hidden shortcut is not to remove it. It is to make it visible.
Give it a name. Write down what it does, step by step, and with what settings. If an AI tool is involved, record the request wording and the tool.
Once it is written, anyone can see it, test it and improve it.
REMEMBER
Make it visible before you change it.What a documented step needs
Learn
Something people can refer to, such as "Supplier name clean-up".
Exactly what is done, in order, with settings.
A person who looks after it.
A short list of changes, with dates and reasons.
Rebuild and compare
Learn
To test a hidden step, rebuild it from the written description and compare the result with the existing numbers.
Where they match, you have confidence. Where they differ, you have found something to explain, and sometimes an old error.
Do this for a handful of important figures first, not everything at once.
RULE OF THUMB
Reproduce before you rely.Keep, change or retire
Learn
Keep
The step works, is understood and is documented. Give it a review date.
Change
The step is useful but unclear or risky. Fix and document it, then test again.
Retire
The step is no longer needed or cannot be explained. Replace it with something simpler.
Keeping it alive
Going Deeper
Documentation goes stale. Put a review date on each step and keep to it.
A growing list of unexplained exceptions is a warning sign. Track it and read it for patterns.
If reopening a number is awkward, people stop asking. Make it easy to ask where a figure came from.
New staff learn how things are really done. Teach them to name and note a shortcut when they make one.
A short example
See It
Back at Mei's manufacturer: she writes down the supplier clean-up step, names it, adds herself as owner and starts a change log.
She rebuilds the step from her description and compares it with the old file. Most figures match. Two do not, and one turns out to come from two suppliers merged by mistake.
She fixes it, notes it in the log and sets a review for three months ahead.
THE LESSON
Writing it down is how you find what was hiding.This is a made-up example for teaching.
Document one step
Try It
YOUR TOOLS
Use your own data and tools. The written description is the main output.GUIDANCE
What to look for is in the Guidance section at the end.Check yourself
Check It
Why make a shortcut visible before removing it?
What does a documented step need?
What does a growing exceptions list tell you?
Answer in your own words before you read the Guidance section at the end.
Chapter 4 in four sentences
Keep It
KEEP THIS
Make a hidden shortcut visible before you change it: give it a name, write the steps, name an owner and keep a change log. Rebuild it and compare the numbers. Then decide to keep, change or retire it, and set a review date. Make it cheap to ask where a figure comes from.
Guidance and answers
At The Back
Questions and exercises stay in the chapters, on their own slides. Guidance lives here, so you can try first and look afterwards.
Try the task or question in your own words, then compare. Where your answer differs, that is worth a note, not a correction.
WHERE TO FIND IT
The Guidance section below: answers and exercise guidance, one part for each chapter.Chapter 1: answers and guidance
Guidance
Someone reuses it, then it becomes normal. At no point does anyone decide it should be permanent.
Any two of: nobody remembers the origin, the steps are not written down, checking is awkward, or people reuse the cleaned file instead of the source.
Looking sure is a claim. Evidence is knowing where the figures come from and being able to reproduce them.
Most people find at least one step they cannot explain. That step is the shortcut. Do not fix it yet. First write down what it does.
Chapter 2: answers and guidance
Guidance
Accurate, complete, consistent, current and traceable.
The tool may change things in ways you cannot see. The original lets you compare before and after, and go back.
The cost of a mistake differs. A rough count may do for a roster but not for pricing or payroll.
Most small data sets have at least one duplicate or odd value. The useful part is your note on what the data is used for, which tells you how much more checking it needs.
Chapter 3: answers and guidance
Guidance
Owner, steward or fixer, and user.
What changed, when, who did it and why. If a tool was used, it says so.
Fewer people who can edit means fewer silent changes, and it is easier to see who did what.
Most people cannot fill in the Steps or Last checked lines at first. Those gaps are the work to do next.
Chapter 4: answers and guidance
Guidance
You cannot judge or fix what you cannot see. Writing it down shows what it really does and lets others test it.
A name, the steps with settings, an owner and a change log.
People are arguing about figures that the process cannot explain. It is a warning sign to read for patterns.
If the rebuilt numbers match, you have confidence in the step. If they differ, you have found something worth explaining, which is a good result.
About Ekalavya Academy
Find Out More
Ekalavya Academy is a learning series from Almost Magic Tech Lab. The books are written for people who teach themselves, using the tools they already have.
Almost Magic builds tools that help people check and govern their AI use. The books never depend on those tools.
LINKS
Almost Magic: almostmagic.net.au
Case studies: ai-casestudies.almostmagic.net.au
Page for this book on the Academy site: link to be added when the page exists.
Ekalavya Academy by Almost Magic | almostmagic.net.au