At risk of incurring wrath, I actually design & implement those things (and the rest of what goes into a PBX) for a living.
The ones that are driving you nuts are classic examples of crappy design and crappy user interface.
---
I have a few rules about designing voicemail / auto-attendant systems:
1) ALWAYS have an option to reach an Operator via the first menu a caller hears.
2) Never have more than two levels deep, or five choices plus Operator at each level. By "level," I mean a menu, after which you press a button to make a selection. After you reach the main menu, you should be able to make one selection to get to where you want, or at most get to one more menu before you get the option you're looking for. On each menu you shouldn't have more than five options plus Operator. Ideally you shouldn't have more than three options plus Operator.
Typically in my designs, I set up a backup operator group. If the main receptionist is busy and doesn't answer in three rings, another group of employees' phones ring. If no one can answer within six rings, the caller gets a mailbox and can leave a message. The point of this is to increase the probability of a caller reaching a live person instead of having to leave a message.
For most types of businesses, the following is all that's sufficient:
"Welcome to ABC Company. If you know your party's extension please dial it now. To dial by first or last name press Pound. To reach an operator press 0 or stay on the line."
The purpose of the above is to take a good bit of load off the receptionist, who is then free to speak with callers who really do need assistance.
For a wide range of consumer product companies, the following should be added:
"For sales, press 1. For product support, press 2. For billing issues, press 3." And each of those should take you directly to a group of people who can answer your call and provide direct assistance.
In cases where call volume fluctuates and can be high at times, a simple queue such as "All of our agents are busy at the moment. Your call will be answered in the order received. If you want to leave a message and have us to call you back, press 1. To return to the main menu press Star." (The feature that anticipates how long until your call will be answered is a predictive algorithm, and when it makes a mistake, the result is major frustration.)
---
Re. overseas employees and "distributed call centers." When you reach someone in e.g. India and can't understand them, it's not their accent, it's the crappy phone system their company has installed. If you spoke with the same person live or on a decent phone, you'd understand them quite well; in fact people speak with people having different accents all the time with little trouble. The problem is the crappy phone system: companies that implement VOIP (internet-protocol telephony) and use the G.729 compression algorithm, which sounds like a bad cellphone. What they should be doing (and what we do wherever possible) is use the G.711 algorithm. The difference between crappy sound and good sound is: crappy sound uses 35K of digital bandwidth, whereas good sound uses 83K of digital bandwidth.
So with crappy sound you can squeeze more people into a carrier circuit. This creates the illusion that your company is saving money on carrier circuits. In fact what happens is, due to the "Huh?" factor ("Huh? What did you say? I can't hear you, could you say that again..?"), each call takes longer, so the number of calls that can be handled per hour is not significantly larger than would occur with better sound (and therefore shorter calls w/o the "Huh?" factor).
---
Here's an example of user interface design. Back in the days when the telephone system was "manual," i.e. all calls connected by operators, the Bell System made an interesting discovery. If the operator said, "I'm sorry they won't answer," the caller would be more likely to get frustrated and ask the Operator to ring them longer or louder. Bell implemented a standard by which the Operators would say "I'm sorry they don't answer," and that solved the problem. Reason why: "won't" implies (and in fact literally is) "will not." Saying "won't" implies the Operator knows the called party is in but refusing to answer the phone. Saying "don't" means "does not answer," and is simply the reporting of an empirical fact. Callers accept that and try their calls again later.
So something as simple as a "won't" or a "don't" can make a significant difference in the degree of satisfaction or frustration for the user. This is a good lesson for modern times with respect to the design of voicemail & auto-attendant systems.
---
AT&T didn't buy out SBC. SBC bought out AT&T. To get the brand as well as some assets in other related lines of business. Yes there was some chaos when the change first occurred. However a good bit of that was also due to storm damage all over CA: floods up north, wind damage down south. People were working 60-70 hours a week. Turns out I got caught in this one with a client who had a complex repair issue and thought it was my fault when in fact it was water damage to outside cables. Finally we had a meet-me with the client and myself and the SBC tech, who was great. He explained the problem to the client and verified that it was outside plant, not the PBX.
---
What really frustrates the hell out of me is "voice recognition." For example, "Say Sales, Support, or Billing." I don't mind pressing buttons, but I do object strongly to being forced to conduct a "conversation" with a machine. OTOH it's a useful reminder of the fact that "certain agencies" are also using voice recognition; for example NSA is 10 years ahead of the civilian sector in this area (as well as most other areas of math, computer science, and associated technologies). "To reach a live Intercept Operator say Anthrax. To have your call transcripts read by an Intel Analyst, say Plutonium. For an in-home visit from Homeland Security, say Assassination..."
