Compared to traditional enterprise software experiences, voice technology is still in its infancy. But already, applications for the workplace are proving intriguing, exciting, and in some ways, very different from the consumer context. If you’ve been charged with designing a voice-based conversational experience for employees, here are some key tips to keep in mind.
Help employees get their work done
There’s nothing more irritating than a chatty colleague when you’re shooting for a hard 5pm deadline. When employees are trying to complete tasks critical to their business or imperative to their paycheque, they aren’t likely to take the time to pull up an experience that adds overhead or hurts productivity. Understand the workflows and tools that employees already use, and find ways to piggyback your experience to maximize utility.
This logic applies for both voice and chat-based conversational experiences. For example, don’t force an employee to launch Google Assistant to give feedback — instead, integrate a few choice questions into the natural flow of team chat. Likewise, support voice-based multitasking — don’t expect an employee to stop what they’re doing to focus on following a lengthy conversation.
Finally, a bit of wordplay to communicate personality is fine, but as in traditional user interface work, you’ll want to ensure these flourishes and adornments are contextually appropriate and tastefully deployed. In general, for the workplace, it’s best to focus most on fulfilling the agent’s functional purpose to demonstrate its value.
Make security and confidentiality a priority
For an at-home voice-based user experience, it’s important to remain conscious of who might be listening in. Rattling off things like bank balances when eavesdropping ears (or young children) are near can be a recipe for disaster, so good conversational interfaces deliver sensitive information with the right line of prompting.
In a workplace environment, the stakes are higher. If you’re building a voice experience that communicates confidential information of any sort, be particularly mindful of where this information might be retrieved, and prompt the user accordingly. These security measures can also be a perfect opportunity to set user expectations.
For example, a contrived system string such as, “I’m going to summarize your performance ratings aloud — is that ok?” can feel like an irritating extra step in a conversation flow. Instead, add value to your preventative prompt by revealing functionality or clarifying scope of available information: “I’m going to summarize your performance ratings, but only for last quarter. Is that ok?”
Keep the audio environment in mind
Open office plans — love them or hate them — are widespread in today’s workplaces. Many voice-based devices employ far-field microphones to cut through ambient noise and capture clear user utterances as an input. But in highly-trafficked environments, background noise can hamper interactions and frustrate employees. To help prevent frustration, there are a few things to do:
- Be sure to assess and test for the fundamental feasibility of a voice-based interaction before you invest time in designing a robust conversation flow.
- Be deliberate about your error-handling strategies to minimize frustration: help users understand when your virtual agent literally can’t make out requests, due to noise and interference, and where in other cases the agent doesn’t have the functionality they’re requesting.
- Remember that conversation is inherently multi-modal. When humans have a conversation, they use much more than voice to communicate. Look for opportunities to integrate both verbal and visual communication in contexts where only one or the other just wouldn’t suffice.
First impressions are everything
If Alexa doesn’t play the right genre on Spotify because she doesn’t know how, you’ll probably give her another chance later. Perhaps you want the experience to work, or simply don’t mind the failed attempt because it doesn’t have dire consequences. In the workplace, where the impact of change is greater, resistance to new modes of interaction can be more significant and as a result, the bar for first impressions is high. Setting expectations and supporting solid feature discovery for an at-work conversational experience is imperative.
For example, imagine you’ve delivered an early beta for a scheduling agent that only supports one-on-one meetings. You can reveal your agent’s capabilities through explicit means (consider a feature description: “I can book one-on-one meetings for you”) or implicit means (consider a leading question: “Which person (singular) would you like me to book a meeting with?”). Given that a conversational interaction path is less constrained by the interface than its graphical counterpart, we need to foreground the guard-rails. Make it known to users what your experience does or does not support, or risk being written off the moment your virtual agent can’t tackle a role-specific request.
Remember that one voice doesn’t fit all
As designers, we know that colour has different connotations in different parts of the world. But conversation — particularly verbal conversation — is perhaps even more context-dependent. Culturally-dependent idioms, jargon, and conversation etiquette all influence conversational interactions. Multiply that by the range of individual personality types and corresponding communication styles (not to mention various languages), and it becomes clear that getting conversational interfaces right requires a lot of effective targeting.
For a global workforce, everything from the utterances your voice agent recognizes to the system response strings it returns may only serve a slice of the employee population. The directness of Germanic communication might not gel with a Canadian population when the subject matter is personal or controversial in nature. Casual commentary on a topic of professional importance might engage someone in the US, but alienate someone in Japan. To get this right, you’ll need to research across regions and cultures, and then test, iterate, and collect feedback continuously.
Start small and bring employees along for the ride
By targeting a limited set of functions for a defined workplace audience, you can limit the blast radius of early failures and pivots, but also create a great opportunity for employee involvement.
For example, you might facilitate co-creation sessions so employees can have a hand in defining the capabilities and designing the personality of your agent. You might also engage with colleagues to understand the types of questions they frequently have, or tasks they’d love assistance with. And if you’re just starting out, you can secure buy-in and generate excitement one-on-one as you investigate the places where voice technology could seamlessly integrate into an employee’s day.
Finally, once your voice experience is live, you can leverage your conversational agent to facilitate feedback. Conversational intercept surveys can supplement analytics to reveal how employees feel about the agent as they’re using it — a great method for collecting feedback in context, rather than peppering users with surveys.
Voice technologies — and conversation in general — present new questions for user experience designers. A traditional software environment might be limited to a few, reasonably consistent and binary modes of input, even where touchscreens are involved. But in the case of voice, the range of inputs is seemingly infinite. Getting the experience right in a consumer context is a feat in itself; the employee context only adds to this challenge. These tips represent our early experiences with the medium in a workplace context, but we hope you’ll expand on and refine these core ideas in your own practice as you explore applications of voice technology at work.
Interested in finding out how voice technology and other smart devices could improve your workplace? Connect with one of our team members today.