AI Automation for Small Business
Prev Next

Finding the work worth automating · Lesson 1

The two-hour audit

A short, practical walk through the two-hour audit.

This lesson covers the two-hour audit and the judgement calls that come with it.

What this is for

Most people meet the two-hour audit as a checklist item and never ask what it is actually buying them. By the end of this lesson you should be able to explain, to a colleague who was not here, why it matters on AI Automation for Small Business and when it does not.

The three things that matter

  • Start from the decision. If nothing you do changes as a result, you are decorating rather than working.
  • Do the smallest version first. A rough pass you finish beats a thorough one you abandon halfway.
  • Write down what you assumed. Future you will not remember, and it is usually the assumption that was wrong.

Walk through it

Follow these in order the first time. Once the shape is familiar you can skip straight to step three.

  1. Write the question you are trying to answer in one sentence.
  2. Find the smallest piece of evidence that would settle it.
  3. Do the work, then check the result against what you expected.
  4. If the two disagree, work out which one is wrong before moving on.

The expensive mistake is not getting it wrong. It is getting it wrong and not noticing for three months.

A worked example

Here is the shape of it, written out so you can adapt it rather than copy it:

# 1. state the question
question = "which step loses the most people?"

# 2. smallest evidence that settles it
funnel = load_events(last_days=30)
drop   = funnel.step_completion().diff()

# 3. check against what you expected
assert drop.idxmin() in expected_suspects, drop

Common mistakes

  • Measuring something because it is easy to measure rather than because it decides anything.
  • Changing two things at once, then not knowing which one worked.
  • Treating a result from one week as a trend.

Before you move on

Try the exercise on your own data, not the sample. The sample is clean and yours will not be, and the mess is where the lesson actually lands. When you are happy, mark the lesson complete.