1. Does this task happen often enough?
List tasks repeated over a week: copying customer details, preparing documents, chasing approvals or consolidating tracking. Measure frequency and actual time spent. A visible but rare task may not be the best candidate. A small operation repeated by several people may represent a larger problem. Include corrections and follow-up checks, not just entry time.
2. Is the information reliable?
Automation also accelerates errors when input data is inconsistent. Identify the reference source for each piece of information, required fields and possible duplicates. Define what happens when a value is missing. Before connecting two tools, check formats and access permissions. Workflow quality depends as much on this preparation as on the chosen technology.
3. Are the rules explicit?
Notifications after approval or prefilled documents can often use conventional rules. If the task involves interpreting variable text, AI assistance may be worth assessing. In that case, prepare representative examples and output checks. The useful question is not 'can we add AI?' but 'which mechanism handles this need correctly with the required control?'.
4. Who handles exceptions?
A customer is missing, a connection fails or an invoice needs correcting. These cases need an owner and a recovery path. Decide when to stop automation, how to inform the user and how to avoid duplicate actions during retries. Keep appropriate approval for sensitive decisions. A system that works only when everything is perfect may shift the burden to people handling incidents.
5. How will you check the outcome?
Before the change, observe time spent, errors and clarification requests. After an initial version, compare the same workflow with comparable volumes. Expected outcomes should remain concrete: one fewer entry, a document found or visible approval. Mawydo can help scope an initial automation from your examples, including the data and exceptions to assess.
