That's presupposing that there is a flow that's repeatable - which may, or may not, be the case.
If you're meaning a 'time & motion' study, or 'workflow analysis', there are a few different ways of looking at what needs to be done.
Time & Motion studies were all the rage in the '70s, my brother had a vac job doing them for Tongaat sugar company. They've now been subsumed under what's now called 'LEAN'. They're renamed 'method and time' studies. It's all, essentially, observation using clipboards and stop watches, or the modern equivalent, and ignoring findings in physics that suggest that measurement of something alters the thing measured.
There certainly is value in establishing what the flow is, if only to establish where there are wasteful loops that establish nothing, but have been traditional - like the finding, one day, in the British Army, that one of the chaps in an artillery battery was there to hold the horses, decades after the horses had ceased to be there.
One way of speeding up the discovery is to get the names and positions of all the people who authorise things (you'll be doing this for a bureaucratic company, sans doubt) and then interview them about their jobs - you can ask them about all the things they authorise, but the only important piece of intelligence is whether they've ever rejected anything - if they haven't then their authorisation is redundant and you've found a wasteful loop.
If you want to then document the flow of work that's evolved in an organisation over, perhaps, several decades, your best tool is probably something like SPICE - an electronic circuit diagram tool, very good at documenting rat's nests.
The tool that I use for a first approximation to such diagrams, in real life, is the most excellent 'dot' language, used to produce directed graphs - it's free (of course, as good stuff so often is) and you can learn about it and get to a download site from here:
https://en.wikipedia.org/wiki/Graphviz
Everybody, I assumed, knew about this magnificently easy to use and powerful tool - but I've encountered lots of people who've only been exposed to the famous WeaklingGesture product from the evil empire and cannot believe that I can produce a good looking, fairly complex diagram in three or four minutes.
I often use my 'dot' files with Omnigraffle, because it's very good for the later stages.
If you're wanting to be more analytical and have flasher, more 'from the expert' looking pictures, then lots of people use UML that, like COBOL is meant to be as easy for managers to understand as techies - [hollow laugh]
I'd recommend, instead, using BPMN, which is less like a circuit diagram and much easier to work with - as well as being considerably more expressive. You can get free tools for it, but I happen to use something called 'Visual Paradigm' that's quite expensive, but works pretty well - as long as you don't expect any miracles from the support team in Hong Kong - http://www.visual-paradigm.com/whats-new/ get the free download and try it for a few days to see if it's your thing. The free tools are just as good, in many ways, but I didn't get on well with the X11 interface.
If you want to really get down to the nitty gritty - and I'd recommend leaving this until phase III when you're finally getting to understand things and have made some improvements, then the Open Group's Archi is free and very good (as long as you're not using it for anything commercial, in which case it's furiously expensive). http://www.archimatetool.com
I've attached a picture of the Adaptive Service Model (takingserviceforward.org/wiki ) that was produced in Archi to give the idea.
Oh, and don't believe any documents or diagrams that you're given that are claimed to explain the process, but smell like something ready for the British Museum rare documents team.
