Don't announce the availability of an agent transfer right away
For most applications, the current leading practice is to NOT announce the availability of agents when a call starts, reserving transfer prompts or automatic transfers to the call center only when there is clear evidence that a caller is in trouble (multiple requests for help, noinputs, and/or nomatch events). Balentine (2006) found that for an application his company was redesigning, briefly delaying the announcement of the availability of agents until after the main menu had played led to a 2% rise in call containment.

Acknowledge every request for an agent
Even unprompted, callers sometimes request an agent or press 0 early in the call, either before or during the initial prompt. For some it's habitual, but some callers are willing to use self-service in general but are confident that their current situation requires an agent. In either case, if they request an agent, it's important to recognize this and respond appropriately (Balentine, 2007; Larson, 2005) – otherwise you waste time for both the enterprise and the caller. As Balentine (2007) puts it, “Users who want to speak to an agent will find a way to do it. They will either experiment with different keys (starting with zero), or they will hang up and call back on a different number. Once they reach an agent, they will vent their frustration. If you don’t believe me, go ask your agents. Then stand back.” If the caller finds it excessively difficult or impossible to reach an agent, then another strategy is to call competitors to see if they offer better customer service – the likelihood of this depends on the difficulty of switching service (Cunningham et al., 2008).

The current leading practice in handling immediate requests for agents (including zero-outs) is to respond immediately to the caller's request. Responding immediately doesn't necessarily mean transferring immediately. Some design options are:

• Transfer the caller as requested
• Tell the caller that no agents are available (e.g., after hours, holidays) and, ideally, offer a call back option
• If early in the call, negotiate with the caller to stay in the IVR
• Compromise with a single question to help you route them better

For effective negotiation, it's important to be able to provide callers with a value proposition (e.g., “For fastest service, please choose one of the following.” or “OK, to get you to the right person, please let me know why you're calling” <followed by a menu designed to provide routing that is as effective as possible with as few choices as possible>). Note that the second version implies a transfer rather than self-service, so you should only use a message like that if agents are available – if not, consider using “To better direct your call …”. If a caller rejects an attempt at negotiation (e.g., by pressing 0 again), it's best to accept that and transfer them immediately to a general agent skill group.

Hura (2008) reported better success at getting callers to make an initial selection with “OK, I'll get you to an agent, but first please tell me if you need help with A, B, C, or D” than with “OK, I can transfer you to an agent after you make a selection”. She interpreted the relative success of the first prompt as, “Users in this case successfully and happily made appropriate selections and were almost always routed correctly; this prompt obviously motivates users to make a good choice because there is a direct benefit to them. The same benefit exists for the other prompt, but the wording makes the selection seem like just another hoop they must jump through for the sake of the automated system” (p. 7).

A different system used this prompt after an early zero-out:

• I'd be happy to transfer you. One question to get you to the right agent. Are you calling to ___?

The researchers estimated that over 90% of callers would answer this one yes/no question. If your system has a clear split between two caller groups for routing, an approach like this can be highly effective.

Research in the consequences of forced self-service indicates that companies which simply ignore attempts by callers to reach an agent in support of the goal of maximizing containment (percentage of tasks completed in the IVR with no agent involvement) are very likely doing damage to their brand Larson, 2005). In a survey of more than 1000 railway customers on having choices in selecting self-service channels, Reinders et al. (2008) found:

• Forcing self-service had adverse effects on attitudes toward the self-service technology and the service provider
• These negative attitudes led to a reduction in positive word of mouth and an increase in the intention to switch to a different service provider
• Offering access to employees to assist in the service offset the negative consequences of forcing self-service

From an analysis of 60 IVR studies across mobile service, airline, and financial service industries, Leppik (2005) found that making it harder to reach an agent:

• Slightly increases the number of self-service transactions
• Causes a strong increase in customer ratings of frustration
• Causes a dramatic drop in customer satisfaction
• The simple fact of automation was not the problem – comparisons of satisfaction and completion with automation showed no statistically significant relationship

The Leppik (2005) data indicate that it is possible to create self-service that works as well as live service, as long as the service is well designed with an understanding of the limits of automation. “The data are very clear: making it hard to reach an agent has only a slight effect on automation rate when you take multiple calls into account, but the price in terms of lower satisfaction and reduced single-call completion is very high” (Leppik, 2005, p. 3).

Consider the balance when playing the agent option
The recommended best practice is to always enable the Operator option, but at what point do you actually tell the caller that the Operator option is available? On your initial prompt of the main menu? The first reprompt? The second? Be aware that although some callers are savvy enough to try going for 0, some callers will consider the absence of the prompt to mean the same thing as lack of the functionality. In other words, if they don't hear the option to choose Operator, they will believe that it is not available, even if it is. This could have an impact on overall customer satisfaction if the CSAT metric you are using takes that into account. On the other hand, featuring the operator option too prominently could lead to increasing queue times, which could potentially also cause a drop in customer satisfaction. Ultimately it is a balancing act for the call center. Consider deploying different versions of your design in an A/B configuration and measuring queue times vs. satisfaction, if possible.

Balentine, B. (2006). The power of the pause. In W. Meisel (Ed.), VUI Visions: Expert Views on Effective Voice User Interface Design (pp. 89-91). Victoria, Canada: TMA Associates.

Balentine, B. (2007). It’s better to be a good machine than a bad person. Annapolis, MD: ICMI Press.

Cunningham, L. F., Young, C. E., & Gerladina, J. H. (2008). Consumer views of self-service technologies. The Service Industries Journal, 28(6), 719-732.

Hura, S. L. (2008). What counts as VUI? Speech Technology, 13(9), 7.

Larson, J. A. (2005). Ten guidelines for designing a successful voice user interface. Speech Technology, 10(1), 51-53.

Leppik, P. (2005). Does forcing callers to use self-service work? Quality Times, 22, 1-3.

Reinders, M., Dabholkar, P. A., & Frambach, R. T. (2008). Consequences of forcing consumers to use technology-based self-service. Journal of Service Research, 11(2), 107-123.