Most organizations have a lot more manual work than they realize.

The reason nobody notices it is that the work has become normal. For example:

• Team members copy information from one system into another.
• Someone builds the same report every Friday.
• Someone sends the same type of follow-up email over and over again.

None of those tasks may seem particularly important on their own, but across an entire team they can consume a surprising amount of time and create a lot of unnecessary friction.

That is what an automation audit is designed to uncover. It uncovers repetitive work, quantifies how much it is costing the business, determines what can realistically be automated, and then focuses on the opportunities that will create the most value.

Here is a practical way to do it.

Step 1: Find the Repetitive Work

Start by asking employees to identify the manual tasks they perform repeatedly. We do not need to make this overly complicated. Just look for work that happens every day, every week, every month, or every time a particular event occurs.

A useful question to ask is: What do you do repeatedly that follows roughly the same process each time?

Some examples/inspirations include:

• Copying information between systems
• Sending the same type of email
• Building reports from raw data
• Routing requests to the appropriate person
• Sending reminders for recurring deadlines
• Creating documents from templates
• Updating spreadsheets
• Renaming or organizing files
• Entering information from forms
• Checking whether something has happened and taking the same follow-up action

Have the team track these activities for a week. For each one, capture how often it happens, roughly how long it takes, which systems are involved, and what usually triggers the work.

At this stage, do not worry about whether the task should actually be automated. The first objective is simply to make the invisible manual work visible.

Step 2: Calculate What the Manual Work Is Costing

Once you have an inventory, start quantifying it.

For each task, estimate how often it happens, how long it takes each time, how many people are involved, and the approximate labour cost.

For example, suppose five employees each spend 30 minutes every week preparing the same type of report. That is roughly 10 hours of labour every month. At a fully loaded cost of $50 per hour, that one process costs approximately $500 per month, or $6,000 per year.

That may not sound like an enormous number by itself, but it becomes much more interesting when you find 20 or 30 similar processes across the business.

This is why the automation audit should not just produce a list of annoying tasks. It should translate those tasks into time and money so you can compare the value of fixing them against the cost of building the automation.

Step 3: Decide What Is Actually Automatable

Not every repetitive task should be automated.

Some processes are relatively easy because they follow a predictable sequence with consistent inputs and outputs. Others depend heavily on judgment, exceptions, context, or human interaction.

A few questions usually tell you a lot.

1. Is the process consistent? Does it follow roughly the same steps every time, or is every case completely different?
2. Are the inputs structured? Does the information arrive through a form, database, spreadsheet, API, email template, or some other reasonably predictable source?
3. Are the rules clear? Can you explain what should happen in a series of clear steps or conditions?
4. How many exceptions are there? A process that works the same way 95% of the time is usually much easier to automate than one where every second case is different.
5. Does it require professional judgment? Some tasks can be automated completely, some can be partially automated, and some should remain human.

The goal is not to force every process into automation. The goal is to identify where automation can reliably remove work without creating more problems than it solves.

Step 4: Measure the Value of Each Opportunity

Time savings are usually the easiest benefit to calculate, but they are not the only reason to automate something.

Automation may also reduce errors, improve response times, make the customer experience more consistent, reduce delays between departments, improve reporting, reduce dependence on individual employees, help the business handle more volume without adding people, or make sure important steps do not get forgotten.

For each opportunity, look at both the amount of work being removed and the broader operational impact.

A task that saves only two hours per month may still be worth automating if missing it could delay a customer order, create a compliance problem, or cause something important to fall through the cracks.

Step 5: Prioritize the Opportunities

By this point, you should understand two things about each task: how valuable it would be to automate and how practical the automation is to implement.

Those two dimensions are usually enough to sort the opportunities.

a) High Value + High Automation Potential

These are generally the best places to start. The process consumes meaningful time or creates meaningful operational problems, and the automation itself is relatively straightforward. These are usually your quick wins.

b) High Value + Lower Automation Potential

These may still be very worthwhile, but they are likely to require more integration, process redesign, custom development, or AI. They belong on the longer-term roadmap rather than being treated as an easy first project.

c) Lower Value + High Automation Potential

These can still make sense if they are extremely easy to implement, but they should not distract the team from higher-value opportunities.

d) Lower Value + Lower Automation Potential

These should usually be left alone. There is no benefit in maximizing the number of automations. The objective is to improve the business, not automate things for the sake of saying they are automated.

Step 6: Fix / Simplify the Process Before You Automate It

This is one of the most important parts of the audit, and it is often skipped. Sometimes a manual process is inefficient because the process itself is badly designed. If you automate it without questioning the workflow, all you have done is make a bad process happen faster.

For example, if employees are manually copying information between two systems, automation may be the right answer. But it is also possible that one of those systems should not exist, the information should only be entered once, or the overall workflow should be simplified.

Before building anything, ask whether the process still makes sense in its current form. The best automation projects often simplify the workflow first and automate it second.

Step 7: Decide How Much Should Be AI

One of the questions that should come out of the automation audit is whether a process is better handled with traditional deterministic automation, AI, or a combination of the two.

Deterministic automation works best when the rules are clear and the inputs are predictable. If a process can be described as “when X happens, check Y and then do Z,” traditional code, APIs, workflows, and integrations are usually the better choice. They are easier to test, more predictable, and generally more reliable.

AI becomes more useful when the process involves ambiguity or unstructured information. That may include reading emails or documents, interpreting natural-language requests, classifying information, summarizing content, extracting meaning from text or images, or drafting a response.

In many cases, the best solution uses both. AI may interpret an incoming email and determine what it is about, while deterministic code handles the actual workflow, updates the correct system, applies the business rules, and records what happened.

In a nutshell, if a process can be handled reliably with straightforward code or an existing integration, that is often the better solution. AI should be used where judgment, interpretation, or unstructured information makes deterministic automation impractical.

The automation audit can identify what type of automation (AI or deterministic) is most appropriate for each process.

Step 8: Build a 90-Day Automation Roadmap

Once the opportunities are prioritized, turn the best ones into a short implementation plan. Pick a small number of high-value opportunities and prove that the process works. For each automation, define the business problem being solved, the process being automated, who owns it, the expected implementation timeframe, the expected time or cost savings, and how success will be measured.

A 90-day roadmap may contain three to five automations, depending on their complexity. The exact number is less important than keeping the plan realistic and making sure the first projects actually get finished.

One automation that reliably saves 40 hours per month is far more valuable than five half-built projects that never quite make it into production.

Step 9: Measure Whether the Automation Actually Worked

Once the automation goes live, go back and compare the result with the original process. Did it actually save the expected time? Did it reduce errors? Did employees stop doing the manual work, or did they continue doing it alongside the automation? Did the automation introduce new problems? Did the business get the benefit you expected?

This matters because it is very easy to build something that technically works without actually improving the process very much. Measuring the result helps prove the value of the first projects and makes it easier to decide what to automate next. Over time, the automation audit can become an ongoing process rather than a one-time exercise.

How Smartt Helps

At Smartt, we help businesses find the manual work that is worth fixing and turn the best opportunities into working automations through our FlexHours program. If you are interested in learning more, please get in touch.