Digital agents are creating a new level of user experience for first-line support that’s fast, easy and — get this — useful. It wasn’t always that way. Like many things digital, asking for help from early generations of answerbots was like asking for directions using a foreign language dictionary. You could never be sure whether the directions you asked for would get you to a pizzeria, the plaza or a periodontist.
Today’s digital agents are far more engaging and helpful, for several reasons. Improved natural language processing enables digital agents to interact in a way that’s more conversational and less artificial. That improved processing also makes it more likely the agent can discern a user’s needs with fewer queries. Cognitive abilities gained through artificial intelligence and enhanced through machine learning help agents more effectively grasp the intent of what users are asking. And they can make small talk with responses that are more contextual because agents can infer what a user wants based not only on the current query but a history of interactions as well.
As an example, a roadside assistance app could employ a digital agent as a first line of support for customers who need help, perhaps for a flat tire or dead battery. Through a mobile phone, the customer can call for help and then engage a digital agent for additional information such as where the car might be towed or how to use the emergency triangles so they can be visible on the roadside. In another example, a credit card customer could engage a digital agent to report a stolen card and provide critical information about when and where the card was stolen, when and where any fraudulent charges were made, and other information. The key in each of these examples is that the digital agent is easy to use, conversational, insightful and intelligent.
In all situations, opening up an API is key to meaningful conversations with digital agents. Anywhere users have logged in through an app or a website, agents can access a broad range of personalised user information in systems of record and other sources. That kind of access enables them to go much deeper than presenting a few generic FAQs and allows them to address a wide range of issues with highly personalised responses.
The obvious, and not-so-obvious, digital agent at work
Digital agents are showing up in a growing number of locations. Some are obvious, like the voice-driven personas that inhabit Amazon Echo and Google Home. Agents can also be added to platforms in less obvious ways that eliminate the need to build special-purpose applications. A railway system might be inclined to build an app for passengers to provide current information about routes, schedules and delays. Riders on the New South Wales transit system in Australia use digital agents to assess their morning commute and routing options based on their history and trips. But there's no need to bring up another app because the riders can access that information through Facebook Messenger.
Agents can be deployed throughout an enterprise for internal and external applications, and in situations where users are authenticated (customers, employees, suppliers) or non-authenticated (guests, prospects, visitors). Many companies view agents as a mechanism to reduce call centre volumes, but that’s only half the picture. Like a digital veneer, agents can act as an orchestrator to pull information and answers from multiple legacy systems and give users a first-contact resolution, whether they’re employees looking for quick tech help or customers with an account question. That can eliminate one of the most frustrating aspects of asking for help — getting bounced from one department to another and restating your case over and over.
Can digital agents effectively replace live chat?
Many companies hesitate to make the swap, fearing that digital agents won’t be up to the task and can’t deliver the kind of experience users expect. But there are steps you can take to make sure the digital agents you deploy are delivering value and satisfaction right out of the gate.
Get organisational buy-in by demonstrating how digital agents work with a small selection of use cases in your environment that feature real questions and answers. Stakeholders will gain an appreciation for the advanced capabilities that today’s digital agent technology delivers, and they’ll more easily envision the role agents can play in the enterprise. For example, if your plan is to use an agent to support human resources (HR), start out by focusing only on a specific subset of HR activities. Or, if you plan to use a digital agent to be the first-line support for product inquiries, centre the use case around specific interactions related to one model of a product.
If possible, limit a digital agent’s initial audience. That can demonstrate the bot’s effectiveness and ease concerns about replacing live chat functions. Access might be filtered by geographic region, business unit, language or any other criterion that can be discerned in a user session.
Pick the right use cases
Before you launch, make sure to do your research. For example, in initial deployments that deliver first-line support for call centres, companies often choose to target the call types that dominate call centre volume. But the way call types are recorded can differ from what customers are actually asking about. Picking the wrong call type when you are just getting started could erode an agent’s effectiveness and ultimately turn off those you are trying to help.
Instead, invest in automated tools and conduct assessments that dive into “utterances” (chat logs and call logs) to help identify what customers are really saying and what can be answered by a bot. This will help define what use cases to attack first, ensuring greater success.
Prove the value
Access to personalised customer data is critical to the success of an interactive, engaging and relevant digital agent. Generally, this is the difference between a good cognitive agent and a simple chatbot. In an initial proof-of-value (POV) process, work with your IT partners and third parties to make sure APIs are exposed so the digital agent can pull in personalised information. You’ll need a robust set of APIs that enable agents to reach deep and wide — into account details, logs, contact and purchase histories, and other data. More access means more context, which in turn means more personalised service. One DXC Technology customer running a digital agent POV was able to handle 50 percent of the inquiries, using an interface for account and card inquiries, to create a seamless, personalised banking experience.
Finally, use a digital agent platform to accelerate the process and improve your odds of success. A POV process built on a digital agent platform should include all the steps, services and infrastructure needed for evaluating use cases, setting up the agent, creating content, building APIs and training the bot.
DXC’s digital agent services are able to provide quick POV. DXC can assess what can be moved to the chatbot in about 10 days and implement a digital agent POV in as little as 60 days. Using an assessment tool that’s part of the platform, we can analyse a range of use cases to determine which can be successfully handled by the agent. Our services feature more than 40 FAQ-type use cases, and we have built FAQ, Q&A and semantic-matching libraries, all of which can be leveraged when running a POV process to help illustrate what will work best for you and make implementation easier and quicker. DXC’s POVs generally run 90 days, and much of the infrastructure, content development and API and UI testing can be done concurrently, which speeds up the process.
Armed with intelligence and backed with data, digital agents are ready to help customers and employees get useful answers to their questions and speedy fixes for their problems. They’re friendly operators who, at any time of the day or night, are always standing by.
Learn more about DXC's digital agent services.