In researching for our next venture, we dove deep into the exploding world of voice AI to develop criteria for high-potential conversational AI use cases.
Conversational voice AI’s ability to understand complex queries and respond in a more human-like way is now far, far beyond the smart voice systems of Apple's Siri and Amazon's Alexa. However, these technologies are still in their relative infancy. They sometimes stumble, and they can be duped.
Given the current state of latency and intelligence, some people might see talking to AI as a bug. Is it too time-consuming to speak out loud? Is this format more faulty than chat? Is it more aggravating to repeat oneself verbally than to type a new prompt? Do we question a conversationalist's intelligence when they speak awkwardly, differently, or without our familiar social cues?
These are all valid hesitancies, but there are times when the bug is actually the feature, where it is preferable and better to use AI than to talk to humans. These will be the best first use cases for conversational AI, because there is already a bias towards the technology.
Our team went looking for these 'bugs turned uses cases' as opportunities for our next ventures, diving deep into the fundamental elements of conversation and experimenting with scenarios in which voice AI could shine. In doing so, we arrived at a set of criteria for evaluating the potential of different conversational AI use cases.
How to Evaluate Opportunities for Conversational AI
To be a good use case for an AI voice tool, it must be a conversation where...
1. Verbal communication > written communication
We’re looking for scenarios where verbal conversation is preferable to text or other options because it directly increases the quality, quantity, or usefulness of information captured. Specifically, scenarios where a verbal response is valuable because it:
- Improves the accuracy of information (vs. a more crafted, curated response via text)
- Improves the authenticity of the response (emotions, and spontaneous feedback is better)
- Increases the quantity of information that someone will share (vs. a more succinct written response)
- Increases the likelihood of response because it is significantly more convenient
An example: Sourcing feedback
Speaking allows people to express themselves more freely and candidly, leading to more candid and richer feedback. This format could potentially increase the quantity of information captured and also improve quality by capturing nuances and emotions that might be lost in written responses.
2. Structure is the norm / is welcome
It must be a scenario where a more contrived conversation with established rules and pre-defined context is reasonable, expected, and even leads to a better outcome. At this point, AI isn’t going to pass as a real human, so it needs to be a scenario where people are willing to engage in a structured, non-organic way. Specifically, scenarios where:
- Consistent delivery of questions or information actually adds value
- Humans are willing to or even expect to engage in a structured, non-organic way
- “Rapport” isn’t required / appreciated / expected
- Reducing human bias can be valuable
An example: Job candidate screening interviews
An interviewer asks candidates a series of predetermined questions to assess their qualifications for a position. The goal is to determine whether the candidate is a good fit for the role in terms of prerequisites, skills, and indicators of commitment/interest.
.png)
3. There is a preference for non-human
This is a scenario where someone would prefer to engage with AI vs. a human. When there’s a reason to talk to AI, a person is less likely to have a negative reaction to it. In some cases, knowing that you are not talking to a real person could actually lead to better outcomes. Specifically scenarios where:
- The information is sensitive or potentially embarrassing to talk to a person about
- A person is demonstrating a weakness or vulnerability
- AI would get more honest feedback from people.
- AI could give more accurate or objective feedback
- The personalization possible through AI adds value
- People worry that a human might judge them or be uninterested in what they have to say
An example: Practicing or learning something new
Learning something new can feel vulnerable. You could practice and make mistakes without the fear of judgment or embarrassment. Conversational AI could potentially provide instant, honest feedback. (e.g., music lessons, foreign language lessons, etc.)
.png)
4. The conversation is outcome-focused vs. relational
We’re looking for a conversation with a defined goal or specific desired outcome that both the AI and the human are aligned on (vs. just a chat or an interaction that’s designed to replace a more meaningful human conversation.) Specifically, scenarios where the conversation is:
- More transactional, or where the conversation isn’t providing connection or relational value.
- Difficult, costly, impractical, or not meaningful to have with a real person
- Facilitating or enabling real human connection vs. trying to replace them entirely
An example: IT, tech support, or troubleshooting
There’s a clear goal: to diagnose and resolve the user’s issue. AI could potentially streamline problem-solving processes by guiding users through troubleshooting steps, asking targeted questions, and providing relevant information in a more dynamic way than a menu of options.
.png)
Wheels Turning?
Our team is actively exploring ideas (and building) around a number of different use cases for conversational AI. Whether you have an area that fits the criteria above or your interest is piqued to explore a use case in your area of expertise, we’d love to hear from you, share what we’re learning, and talk about how to hone in on a business model and user experience.
Drop us a line at joey@19days.com
