Within the product team at Ufonia, when we don’t like a new conversation feature, one common criticism is “it’s too IVR-y”.
I previously worked on traditional IVR (interactive voice response) systems, so it made me wonder, what does it mean to be “too IVR-y”? We categorise IVRs as being bad, but why is that?
“Please now select from the list of 12 possible options….” 😫
Ufonia’s flagship product is Dora, an autonomous clinical assistant that can conduct medical phone calls to reduce the load on doctors and nurses.
I work as a conversation designer, writing the prompts and flows for Dora. Prior to this, I worked on designing those automated phone systems that you may have encountered when calling your bank or utilities company. These phone systems are called Interactive Voice Response systems (IVRs), and they tend to get a bad rap.
As a phone-based system, Dora is technically an IVR. But we try really hard to avoid branding Dora as one. This is for two reasons. One is that Dora uses AI to be smarter and more effective than conventional IVRs. The second is that we strive to always put the user first and “IVRs” are notorious for being user un-friendly.
So, why do IVRs have such a bad reputation?
The first thing to bear in mind is the context in which people call IVRs. Typically, people only call companies when they have a problem.
Maybe you think there’s an error with your utilities bill. You go online to see if you can solve the issue on their website. But you can’t, so you try calling their hotline and hope to get someone to fix it as soon as possible. You’re already feeling kind of angry right off the bat.
Next, you’re confronted with a long set of options, none of which really fit why you’re calling. Suddenly, it’s your job to try and figure out how to navigate this annoying system that keeps misunderstanding you (“I think you said ‘Billing’, is that right?” no, it’s NOT 😠).
After some navigation (and time on hold), you finally get to a human and suddenly everything becomes easier. You can articulate your problem to someone who understands the nuance of the situation and can help solve the issue.
Why was that so hard? Shouldn’t the IVR have been designed to make things better rather than worse?
How do we avoid being “too IVRy”? IVR anti-patterns!
At Ufonia we cultivate a list of “IVR anti-patterns”. These are common design approaches that conventional IVR companies have adopted because they make IVR implementation easy, but they’re at the expense of user experience.
We constantly iterate on this list of anti-patterns, updating and adding to them as we learn more, examples include:
Forcing a particular response: Something like “If this is correct, say ‘yes’” is common in IVRs, but never in human conversation! Simply asking “Does that sound right?” is much more natural. Dora’s natural language processing (NLP) allows it to understand the different variations of “yes” or “no” that could be given in response.
Repeating things verbatim: Often if the call flow goes awry in an IVR system, the default fallback is to start over from where things went wrong. Typically this means the user sitting through the same information twice, even when, more often than not, the user knows everything at this point! No human would ever repeat themselves verbatim in this situation. Instead, it’s helpful to paraphrase and think about the potential reasons why things have gone wrong.
Listing options: We’ve all had to sit through something like “Press one for Accounts, press two for Billing, press three for…”. By the time you get to seven you’ve probably forgotten which number was right for you! Instead, it’s better to use open ended questions like “What can I help you with today?” This is another situation where NLP can reliably map users’ responses to intents.
For Dora, we avoid IVR anti-patterns at all costs. Each prompt is different, and whilst there are times when avoiding an anti-pattern can make the conversation longer, we always put our users first.
Automated phone systems have the power to save time and help users. In the case of Dora, we are helping to manage the huge demands on healthcare services, while providing more reliable and timely care for patients. The key to unlocking this power is to design them to help the user first, then the organisation 😇!
