Virtual agents have proven themselves valuable in customer experience settings by saving costs and increasing customer satisfaction through faster service. The emergence of COVID-19 has put virtual agents in the spotlight as a way to manage the impact on contact centers and keep the customer service workforce safe. But not all virtual agents are created equal, and the quality of interactions continues to be a real differentiator. Here are 10 ways to ensure you're building a standout experience.
1. Provide a top-notch user experience
Consider what an amazing experience looks like for your users. Today's virtual agents are intuitive to use and incorporate visual communication, voice and text communication, and rich responses. If the user interface (UI) does not look amazing and the user experience (UX) flows are not carefully thought out, adoption may suffer, derailing the entire project.
2. Build for speed
There is nothing more frustrating to the user than asking the agent a question and experiencing a long lag time before getting a response. This is where a solid technical architecture becomes essential. If you are unsure which technology stack is best for you, consult a senior system architect or do technology spikes to quickly test and evaluate different options before committing to a solution.
3. All technology is not equal
I often hear sales teams tell customers "we can build your virtual agent on any technology stack." That may be true, but the final user experience will vary greatly depending on the technology that you choose. I like to compare creating a virtual agent to building a car. We can build a car out of a variety of components, but there's a big difference between a Toyota Yaris and a Porsche 911. Both will take you from point A to B, but the experience will be very different. If customers expect they're buying a Porsche, this requires an appropriate technology stack for the Porsche experience. These first three points of technology -- stack, speed, user experience -- are interrelated and extremely important to creating a great virtual agent.
4. Risk management
You can take several steps to mitigate the risk of your project. First, conduct technical spikes with your team on any new technologies and ideate different solutions with estimated effort and technical tradeoffs. Second, don't take on too many new technologies at once; spread them throughout your release plan. Third, if a specific feature or intent could be greatly enhanced by a technology, build that feature iteratively starting with a simple solution and incorporate more advanced technology over time to avoid development delays.
5. Where's the data?
Like most AI products, virtual agent features are highly dependent on data. Having data siloed in different parts of an organization makes it technically challenging to implement certain features. It's important to map where different data lives within an organization at the beginning of the project to understand feature feasibility and technical risk. In some cases, a data migration project may be necessary before taking on an AI project.
6. Prioritize intents
Many teams struggle with deciding what intents, or scenarios, that virtual agents should assist customers with. Intent selection can be prioritized by intent frequency, feasibility, viability, business value or customer value. If cost savings is the main goal, business value is a good metric to determine what intents to start with. Using call center data, you can calculate the cost of each intent:
Intent Cost Savings = [Intent frequency] x [average time (s) to resolve intent] x [customer service agent cost / second]
This will help rank the relative value of the intents and prioritize them in terms of business value, as well as legal viability and technical complexity. There also may be instances where intents with less business value might provide a more immersive user experience and delight customers.
7. Create a roadmap
Creating an outcome-based feature roadmap is essential to lay the technical foundation that will enable near- and long-term features. You want to spread technical risk throughout agent solution releases and balance that with the features that align to desired business outcomes, while delivering the most customer and business value. Once in production, you can collect data to help determine your future feature roadmap and potential channel expansion.
8. Figure out channels to focus around
Virtual agents have to provide an omnichannel experience that meets customers' expectations. There are three key points to think about when deciding what channels to focus on.
First, know your customers, what their pain points are, and what digital channels they prefer. You don't want to invest in channels that customers don't actively use. Second, create a technical architecture that allows for easy channel expansion. Many companies start with two primary channels and design a flexible back end for future secondary channels. Third, understand technical limitations and how user experience will vary in the different channels. Be aware of the trade-offs as you make your feature and channel selections.
9. Create master agent and sub-agents
Depending on how many intents your virtual agent will handle, it might make sense to create a master agent/subagent architecture, where the master agent controls subagents that handle one distinct intent. This architecture can be beneficial if you have multiple engineering pods working on different intents, as it allows them to work independently without overlap.
10. Set the right tone and brand
A virtual agent's personality must be in line with your tone and brand. I usually research the client's Twitter, Facebook and Instagram pages to understand how the company communicates with customers. This helps me determine the correct tone of voice, UI, animations and that each feature aligns with the company's brand.
Use these tips to create great virtual agents
Creating an amazing, omnichannel virtual agent is a challenging task that requires balancing design, user experience, technical feasibility and business value. It's important to have empathy for your users and understand their needs. By following these steps, you will have a smoother delivery process, keep your client happy, and most importantly create an amazing experience for their customers.
About the author
Erik Widman, Ph.D., is a senior manager at Accenture. He is specifically part of their AI business and works with voice AI, computer vision and machine learning.