Automate first the process that repeats often, can be described as rules, and is expensive when it goes wrong. A rare process, or one that is different every time, is not worth automating.
Frequency: how many times a week does it happen. Twice - wait. Fifty times - useful. Regularity: can you explain it so a new employee gets it right? If not, the process is unclear, not the tooling. Cost of error: what happens when someone gets it wrong. If the answer is "a fine" or "we lose the customer", it is a priority.
Two places: moving data by hand from one system to another, and producing documents by hand. Both are frequent, both are rule-based, and errors in both are expensive. In Georgia, waybills and invoices sit on top of that. They are mandatory for most VAT payers and repeat many times a day - which satisfies all three filters at once.
Decisions that depend on context. Exceptions. Processes that are still changing. If a process changed three times in six months, automating it locks in a shape that is probably wrong. Stabilise it first.
Not a large system, but one chain end to end: event → action → result, with a log and a rule for what happens on failure. Once that one chain works and saves time, the second is easy. The opposite - a large project covering everything at once - frequently never ships.
Spend one week noting where data gets moved by hand. That list almost always contains the first process worth automating.
8 Cristoforo Castelli St, Tbilisi 0159, Georgia · +995 596 16 66 88 · info@zenlumex.com