AA just doesn't have the heat to do the elements we analyze but FAAS is just fine for Au/Ag/Cu/Zn etc. Marginal for Pt and Pd w/ C2H2/N2O. Virtually worthless Re or Rh or Ta and W and a ROYAL pain if you had to analyze the chalcogenides/hydride formers. We had an 8 lamp turret so we'd have Ag/Cu/Pd/Zn/Ni as our nitrate suite when we would do all HNO3-containing sample days. Then for Au/PGMs we'd do in 3% HCl. The AA was convenient...just walk up, aspirate your sample and then you could get your readings and crunch them easily. I think for people doing cyanide work or silver cell work, an AA, a good analytical balance, 1 fume hood, and a burette and dispensette are ample.
Alas, that AA's still sitting in storage waiting for a home :/
I do enjoy being able to prep 200 different samples/dilutions (we bracket our concentrations) and let the robot do the work. Prep takes the longest amount of time, as you have to take the temperature of the room, weigh the aliquot from the pipette, enter it into the spreadsheet and program and go down the list. You select a method, let's say you wanted to know Pd--then you do a Pd method specific to the XRF trace of the impurity profile. If it's high purity Pd where matrix/impurities effects are minimal and IECs/models aren't needed, then you use a method by difference.
Of course, and what no one tells you when they sell you one, and what hasn't been mentioned here: anyone who does this for a living knows (or darn well better) that you don't go straight to ICP for settlement results--you go straight to gravimetric pre concentration THEN ICP. This means fire assay (gold or NiS collector for the PGMs), dimethylglyoxime, certain analytical resins, etc. ICP is good for house keeping solutions but it does not compare to tried and true protocols. Marry them both, like fire assay for Au, impurities by difference, silver by titration/Pd by DMG...then you have something.
Unfortunately, as my business seems to process really oddball/difficult matrices that either the big refiners won't do or can't, often times I'm spending effort and lots of time on method development where I'm figuring out which elements interfere with which others and correcting for that. In some instances, ICP-OES, even with a 150K instrument is nigh on worthless for Pt so a different technique needs picked out. Nice thing about ICP-MS for Pt is no barometric interferences; not nice thing is the stupid dilution factors which throw in absurd error bars, so you're stuck doing it on a Cahn microbalance if you want to keep RSD down. I'm a big fan of mass/mass make up for samples and standards.