Writing
Start where the work gets stuck.
Before choosing software, look closely at the task that has become difficult.
A business can become very good at working around a problem. Someone knows which spreadsheet is current. Someone else remembers the exception the software cannot handle. A third person copies the result into the system that actually matters.
By the time the business asks for new software, the request may contain a diagnosis: replace the platform, automate the process, add an assistant. The first useful step is to understand the work underneath that request.
Follow one task.
Choose an ordinary piece of work and follow it from beginning to end. Who starts it? What information do they need? Where does it wait? Who checks it? What happens when the usual case does not apply?
The people doing the work can show you things a diagram will miss. Their workarounds may point to a broken handoff. They may also preserve an important distinction that a replacement system would accidentally erase.
Make the uncertainty explicit.
Separate what you observed from what you believe is causing it. Then choose a small way to test the explanation. A useful result might be a prototype, a changed handoff or evidence that the proposed project is solving the wrong problem.
This is also where a conversation about AI belongs. What part of the task would change? How would someone check the result? What would happen to the person currently doing the work? A tool choice becomes easier to discuss when those questions have concrete answers.
Keep the next step proportional.
Sometimes the answer is a new system. Sometimes it is a small integration or a clearer responsibility. The scale of the remedy should follow what you find.
That is a useful place for a technical partner to begin: close enough to understand the work, and able to help change it once the next step is clear.
Tell us what you are working on.
A conversation is a good place to start.