A business buys an inventory system. Three months later everyone is back on the spreadsheet and the system is open in a tab nobody looks at. This is common enough to be predictable, and the software is rarely the reason.
The reason is that it was configured around how the work was assumed to happen, not how it actually happens. Fixing that costs nothing and has to happen before you spend anything.
Follow one job from beginning to end
Take a single real order, delivery or client engagement from the moment it arrives to the moment the money lands. Walk it through physically. Talk to the people who touch it.
Write down every step, including the ones nobody would put in a formal process document: the WhatsApp message to check stock, the notebook behind the counter, the call to the supplier, the photograph of a receipt sent to the accountant, the thing that only one person knows how to do.
Those informal steps are the important ones. They exist because the official process did not cover something. Software that ignores them does not remove them -- it just forces people to keep doing them on the side.
Note every time information is entered twice
As you walk through, mark every point where the same fact gets written down again somewhere new. A customer name typed into a delivery note, then into an invoice, then into a spreadsheet. A stock count written on paper, then keyed in later.
Every one of those is both a cost and a place where the numbers begin to disagree. It is also the clearest measure of what a system would actually be worth to you. A business with fourteen re-entry points has far more to gain than one with two.
Find where the work waits
Mark every point where a job sits still waiting for someone. Waiting for approval. Waiting for the one person with the password. Waiting for stock confirmation. Waiting for someone to be back in the office.
Delays are usually where customers get lost, and they are usually caused by process rather than effort. Some of them software fixes. Some of them only a decision fixes, and it is worth knowing which is which before you pay for the wrong solution.
Ask what people work around
Ask the staff directly: what is the annoying part of your day? What do you do twice? What do you keep your own record of, because you do not trust the official one?
That last question is the most revealing in the whole exercise. A private spreadsheet somebody maintains on the side is a precise map of where the current system fails them. If a new system does not address it, they will keep the spreadsheet, and you will have two sources of truth instead of one.
What you now have
An honest description of how your business actually runs. It usually differs from what the owner believed in at least two significant ways.
With it you can judge software on whether it fits the real process, insist that configuration matches your sequence rather than the vendor default, brief whoever implements it properly, and measure afterwards whether anything actually improved.
The signs you are not ready yet
If nobody agrees on what the current process is, settle that first. Software will not arbitrate a disagreement between two departments.
If the process changes depending on who is working, decide which version is correct before automating either.
If the real problem is that one person is overloaded, a system may help or may simply give them another thing to maintain. Be honest about which.
A note on migrating your records
When you do buy, insist your existing records come across -- customers, stock, history. A system that starts empty is one your staff will run a spreadsheet alongside, because the spreadsheet has the information they need today.
That migration is usually the least glamorous and most important part of the whole project. It is also the part most often quoted as an optional extra. Get it in the scope in writing.