Start with an activity familiar from your daily work
Business process automation can cover many areas. Here, I focus on scripts that run on a computer: processing files, transferring data and preparing reports using agreed rules. This describes the scope of my automation service, rather than every possible type of automation.
For a few days, record activities you perform repeatedly in a similar manner. Downloading attachments, renaming files, or merging exports are all candidates for analysis. However, repetition alone isn't enough to make a decision.
Check whether you can explain the task to someone else and describe a correct result. If your instructions end with “then assess the situation”, clarify that step. The program needs a rule or a point at which it refers the decision to a person.
How do you recognise a suitable process for your first implementation?
Look for a task with a narrow scope that recurs frequently enough to warrant streamlining. It's easier to evaluate a script that combines a single export type than a program that claims to "manage company reporting." The initial release doesn't necessarily support all departments and formats.
Consider the cost of error. A report prepared for verification allows for comparison of results before use. Automatic data changes in the system may require additional confirmation and a method for reversing the operation.
A process should have an owner: someone who explains the rules, accepts the results, and reports changes to the data format. Without this, organisational problems can easily be transferred to the program.
- The data has a recognisable format and a known source.
- The rules can be written down and checked with examples.
- The result can be compared with a known correct example.
- There is an agreed way to handle exceptions.
Example: report from several sales files
Let's say you download exports weekly, organise columns, and combine the data into a report showing sales by product and period. This is a sample scenario for planning, not a description of a confirmed implementation.
First, define what each field means. Does the date refer to an order being placed, paid for or dispatched? Does the amount account for returns and delivery? How do you identify the same order in two files? A script can add numbers correctly while still producing a report that misrepresents the business.
Then, describe the flow: reading the folder, checking columns, combining data, calculating, and saving. Rows requiring explanation can be placed in a separate list. For a similar task, review reporting automation —both files and calculation definitions will be important.
Exceptions are part of the process
An empty file, a missing identifier or a renamed column should not cause data to be skipped silently. Agree when the program should continue and when it should stop. Error messages should describe the problem clearly enough for the user to know what to correct.
Restarting is important. If you click a second time, the program should behave as specified: overwrite the report, create a new version, or skip detected items. When saving to another system, it's important to prevent unintentional duplication of operations.
Start with copies and a small sample. Compare the result with a manually prepared reference, then increase the amount of data. Include files containing errors, not just perfectly prepared examples.
Calculate the total time spent on the task
Consider a manual task that takes 45 minutes and is performed eight times a month: 360 minutes in total. If running the script and checking the result takes 10 minutes each time, that becomes 80 minutes a month. The difference is 280 minutes, or 4 hours and 40 minutes.
This illustrates a calculation, rather than promising savings. Your assessment should include actual measurements, development costs, exception handling and maintenance when data changes. The value of the time saved depends on how you can use it.
Also record corrections and delays. A consistent report format or fewer errors may be valuable, but these benefits need to be verified. Faster execution does not guarantee a better result.
Computer, schedule and integrations
The script runs in a specific environment. Information about the operating system, programs, permissions, and execution method is required. If it's scheduled, determine whether the computer will be on and have access to the required resources.
When integrating systems, there are further considerations, including data flow direction, identifiers and behaviour during service outages. This goes beyond reading a file stored on disk.
If the data and rules are clear, start with a rule-based solution. If the task involves interpreting free-form text, consider AI and the checks on its responses separately.
What should you prepare for a quote?
Select one process and describe its current flow. Include anonymised input data and an example of a valid output. Note the frequency, number of files, common exceptions, and who will be using the program.
Add acceptance conditions: what examples must pass and what the user should see if they fail. This ensures that the quote is specific to the job. If you need help organising this information, describe the task in a request for work automation using a script.
Questions and Answers
Do you need to automate the entire process?
No. You can automate data preparation while leaving review and submission to an employee. What matters is that the automated part produces a useful result.
What if the file format changes?
The program should detect significant incompatibilities. Support for the new format may require script changes, so it's worth establishing update policies before implementation.
From knowledge to action
See how I can help with your project.
