Observe the work that already happens
Choose a recurring task. Record its frequency, time spent and errors. Ask the person doing it what information they need and which decisions they make; a diagram without that conversation often misses important work.
Keep observed figures separate from estimates. If you have not measured time yet, do not turn an impression into a promise of savings.
Define one verifiable improvement
Choose a small outcome: enter an enquiry once, show its owner, or generate a draft from reviewed data. Specify inputs, outputs and what happens when information is missing.
An existing tool may be enough. Custom software makes sense when an important need is not adequately covered and the total cost is reasonable.
Compare benefit with total cost
Include implementation, licences, maintenance, training and team time. Use conservative scenarios rather than assuming every freed hour becomes cash.
Before expanding, test with a small group, retain a way to roll back and compare the result with the starting baseline.