Conversational UX for chatbots
CONTENT • MARCH 22, 2019
Let us speak the truth, the reason why most customers contact your company is to get answers to their questions. Fast. Regardless of the channel they pick, their expectation is to live an efficient and low effort experience.
This is even more true with phone calls. Most of your customers will call only when they couldn’t find answers to their questions themselves. They have searched the web, their mobile apps, and are now able to formulate their questions even more explicitly. Get into their skin for a second, and put words on how they feel when they call your company. Impatient, focused. They have high expectations.
Now, let us suppose your contact center is in a call peak situation. Your customers have no reason to feel responsible for this peak — this is a situation your company has to deal with, not them. They are not aware nor care that they are one of many and rightly expect a high-quality service. Truth be told, few companies can take pride in providing an acceptable customer experience during call peaks, even fewer companies manage to make it a competitive advantage on the market. One of the main reasons is that companies need a contact center platform which can both seamlessly scale up to handle any kind of spike (think hundreds or thousands times more interactions that the usual volume) and provide high-end IVR and routing capabilities.
Amazon Connect is a self-service, cloud-based contact center that makes it easy for any business to deliver high quality customer service at low costs. By leveraging the power of the whole Amazon Web Services ecosystem (AWS), Amazon Connect offers your company the empowerment to focus on your customer’s experience when dealing with peaks. Many built-in capabilities echo to the specific needs your CC encounters when managing this demanding situation.
While most content published on Amazon Connect highlight the obvious technical reasons why Amazon Connect is a good fit for peak management (scalability, stability, etc.), we decided to take another angle and focus on the customer experience the platform can deliver.
Peaks, do they really happen?
Theoretically, a peak is a higher than usual volume of interactions entering your contact center. Two families of events may cause peaks: expected and unexpected events.
Expected events can be:
- Intensive sales periods (Black Friday, Cyber Week, Boxing Day, etc.)
- Holiday periods (Christmas, Easter, etc.)
- Planned marketing campaigns (new customer offers, etc.)
- Changes to the service (price increases, changes to general conditions, etc.)
- Major events (Soccer World Cup, Olympic Games, etc.)
- Due dates or events specific to certain industries and services (RRSPs, etc.)
Unexpected events can be:
- Breakdowns, technical issues
- Products or services issues generating an increase in complaints
- Product Recalls
- Exposure to harmful content on the Internet or press
- Non-announced marketing campaigns
When reading this non-exhaustive list of events, you immediately realize that peaks actually happen. More often than you might think. Depending on your industry, products and services, and the capacity of your contact center, you encounter peaks in multiple forms.
How do CC typically deal with peaks?
Let us state it first, dealing with peaks is demanding and costly for a company. Each company has its own way to deal with peaks. Each one makes their own choices. Typically, this comes down to planning and technological choices.
Planning starts with a business choice: the way you manage the balance between capacity (employees) and demand (workload). Every company or department can have its own way of managing that balance (chooses to align the two, or not), according to its business priorities. Accurate planning is key in peak management, many companies fail to deliver a quality customer experience because of poor planning.
If the company decides to act, its most common levers are:
To temporarily increase the capacity:
- Hire part-time employees, seasonal or temporary staff
- Use the support of subcontractors
- Use another skill / department
- Use overtime.
To distribute the demand more evenly:
- Smooth out the demand over time by promoting self-service resources to communicate with the company (chatbot, website, transactional IVR, etc.)
- Rely on the different communication channels — more on that approach in the next section.
From a technological perspective, the main concern is about the capacity of the systems. Peak periods imply a stress on the system infrastructures (telephone and other), it is essential for the company to ensure that they have the capacity to support peaks of customer contacts. More on that in a next blog post.
Get your focus on your customer by leveraging the power of Amazon Connect
In Amazon Connect, customer interactions are enabled and controlled in ‘contact flows’, from answer to termination. Contact flows are designed in the Amazon Connect administration console, and consist of a series of interrelated building blocks. Are you a Lego fan? You will love it!
The contact flow includes the logic for your IVR and for your routing strategies. It leverages powerful speech and computing tools from the AWS platform. For example, Amazon Lex is integrated to provide Automatic Speech Recognition (ASR) and Natural Language Understanding (NLU), and Amazon Polly for Text To Speech (TTS). Plus, you can benefit from the superpowers of Amazon Lambda to obtain data from virtually any existing backend application and use this data as inputs into the contact flows to dynamically control your customer’s interaction.
A well thought out contact flow design strategy that leverages the surrounding AWS tools makes it possible to build an advanced self-service speech application and benefits from advanced routing capabilities. Functionally, the main reason why Amazon Connect is a great tool for dealing with peaks lies in the fact that the scope of most of the questions the callers will have is limited. By being able to anticipate the majority of your customer’s needs, you can (pro)actively leverage on the ‘automated, dynamic, personalized’ customer experience promised by Amazon Connect. Now, let us go back to the customer experience, and check how Amazon Connect can help you smoothly manage your customer’s expectations.
Provide your customer with the feeling that YOU have control over his experience
Your customer calls you. Your CC lives a peak situation. You want to be in control and prove it through the customer experience you offer — here is how to achieve this with Amazon Connect:
1. Do not miss an opportunity to leverage on the available data by designing dynamic and personal interactions using the contact flow engine
Data directed processes play a key role in showcasing your ability to build from your current knowledge of your customer and from the context of the ongoing interaction. This may simply start by showing that you identified your customer — and not asking for another pointless ID procedure. With Amazon Connect on AWS your can perform live digging into virtually any type of application, database, or analytics data source to tailor a personalized experience. Here are a few possibilities:
- Use Lambda functions to retrieve data from any backend application that will prove useful in proposing your customer an effortless interaction — ie contact attributes to identify the customer, categorize him, know his latest navigation with the company self-services, historical actions that may be linked to this particular call, his past contact history, his preferences, etc.
- Use Lambda functions to extract live queue statistics and limit the number of interactions waiting in the queue in order to maximize your distribution among the multiple queues and get the quickest agent response possible
- Use dynamic prompts capabilities to adapt the IVR experience based on contact attributes out of Lambda or Lex
- Use dynamic queue/agent selection capabilities to adapt the routing based on contact attributes out of Lambda or Lex.
2. Make the most out of artificial intelligence
The good news in peak situation is that most customers probably call your company for the exact same known reasons (which is true in particular for peaks caused by expected events). AI will prove especially efficient (and less pricey) when dealing with a limited number of intents, within a narrow scope.
Feel free to use Lex capabilities to answer the highest number of questions possible.
3. Anticipate by making use of your past peak experience
A clever idea would be to develop/train/deploy Lex robots for each known peak scenario — these could prove very useful in enhancing your customer’s experience and your own understanding of how you should deal with peaks.
Another suggestion would be to prepare a library of contact flows that will allow you to potentially adapt your customer’s experience for each known peak scenario.
To make the most out of your past peak experience, you can for example start the call by proposing a short selection of dynamic choices for all currently expected peak reasons (ie you will know when expected marketing campaigns will generate calls), then use the prepared contact flows and bots for each of these choices.
4. Do not be afraid to announce that you currently live a peak
If you offer a qualitative experience do not miss an opportunity to inform your customer that you did so in the context of a peak. A simple, honest statement will always prove valuable to your customer, plus it will highlight your ability to be in full control of his experience whatever the situation.
5. Do not mess with audio quality
It may sound straightforward, however audio quality plays a key role in the customer experience. In high peaks your company will potentially enlarge the agent pool (internally or by outsourcing) and agents may work remotely, from home or on a different network (or country). The Amazon Connect softphone leverages WebRTC and the Opus audio codec to allow your customers and agents to enjoy a high-quality audio discussion.
Provide your customer with the feeling that HE has the control over his experience
As a customer, your impatience and focus will quickly shift into frustration if the only way out of your call is to undergo another passive waiting experience… or to hang up. You want your questions answered and be in control of your interaction — here is how to achieve that with Amazon Connect:
1. Do not let your customer down
While typical peak experiences end up in asking the customer to call back later (or get him to understand that he would better do so after 30 minutes waiting in queue), please provide your customer with multiple options. An informed and active customer is a happy customer. Here are a few possibilities:
- Use Lambda services to extract live queue statistics and warn your customer with a rough estimated waiting time (ie current number of customers in the queue, approximate waiting time, etc.)
- Use built-in callback capabilities to offer your customer to keep his place in the queue and allow an agent to call him back once his call sits at the top of the queue
- Use Lex conversational capabilities to keep your customer active (if he wants to): if the IVR can answer your customer’s questions before he needs to speak to an agent, then you should offer him to use it. Even if his questions were not completely answered, gathering natural language information will prove useful to dynamically adapt the routing to the agent and further enhance your Amazon Connect application (ie by training the Lex bot and making the appropriate changes). How ever AI capabilities are used, always provide the customer with the choice to speak to an agent.
2. Do not stuff his waiting experience with information that he already has
Most callers have already used all self-service resources available before calling (or will do so while waiting in the queue). Your best strategy is to make sure you do not repeat him the same information again and again… and again. Here are a few possibilities:
- Use dynamic prompts capabilities in your IVR to provide information specific to the peak situation (ie some services may not be available during the peak, information may be updated accordingly in the other channels)
- Use Lambda services to extract past navigation data and adapt prompt play accordingly in the waiting queue (ie suggest him to switch to one or more less charged channels for a faster response (chatbot, chat, etc.)).
Think about it as a market opportunity
Peak management is heavy and hard to deal with — it implies strategic planning, deploying trained agents (potentially new or external agents who need to be trained), lots of logistics and stress. High peaks will test the flexibility of your CC to the limit. Through the storm, it is easy to ignore how it will feel to the caller. But isn’t his experience what it is all about? I think it is.
While peak management usually involves taking low value, high volume calls and diverting them to a lower cost resource either within the company or externally, Amazon Connect takes the debate to another level by making it a potential market differentiator. Making your customer call not a call like another one, not a customer experience like another one. Making it one to build on.
Nu Echo now offers Amazon Connect customization services to enterprise clients across North America. As a conversational innovation solution provider, we ensure that when a customer calls your company, they are presented with a high-quality experience every step of the way.