A useful starting point for an automation project is a plain-language description of the work. Put the current steps on a page before deciding what software to build.
1. What starts the work?
Name the trigger: a file arrives, a customer submits a request, or someone reaches a scheduled review date. Record who owns that first step and what information is available.
2. What information goes in?
List the inputs and where they come from. Include the awkward cases: missing fields, inconsistent names, duplicate records, and files that arrive in a different format.
For a supplier-data workflow, you might distinguish a supplier’s item code from your own internal reference. Whether those codes can be matched reliably is a question to investigate.
3. Where does a person make a decision?
Separate straightforward steps from judgement calls. A fixed calculation may have a clear rule; an ambiguous match may need someone to review the source material.
If you are considering AI or machine learning, identify the specific decision it would support and how a person would check the result. Avoid treating “add AI” as a complete requirement.
4. What should happen when something goes wrong?
Define an exception path. Who should see a missing input or uncertain result? Can the work pause safely? What information would help someone understand and correct a mistake?
Include a way to trace an output back to its input. Discuss what needs to be retained and who should have access before connecting real business data.
5. How will you evaluate the first version?
Choose a small, representative set of examples, including difficult cases. Define the expected result for each one and the situations that require human review.
The result of this exercise is a clearer project brief: what to automate, what to investigate, and where people stay involved. That brief can guide the technology discussion.
Discuss the next step
Explore ZipLogic’s AI/ML solution planning or share your project goals.