The Definitive Guide to Virtual Agents

In 2020 during the height of the COVID-19 pandemic, McKinsey and Company found that 62% of employees were working from home. While 41% felt more productive, one in three were struggling. That’s not surprising when you consider that, according to HR Executive, nearly 7 out of 10 workers claimed the pandemic was the most stressful time of their professional careers.

What contributed to that level of employee stress? It was a combination of concerns around both health and income, as well as trying to be productive in a work from home world.

Now that we are entering a phase where vaccines are becoming available, Gartner asked HR leaders about their plans to allow work from home to continue:

• 90% will allow work from home at least part of the time
• 65% will continue to allow flexibility for when employees work
• 50% of the workforce wants to return to the office at least part of the time

The phrase to pay attention to is “at least part of the time.” Organizations are going to have to support a workforce with a mix of remote, in the office, and hybrid.

When working from home, employees can no longer shoulder tap the person sitting next to them to get help. Flex hours makes getting help even more difficult. Because of that, calls to the IT help desk after the work from home mandate went into effect were up approximately 35%.

Even before the pandemic, a Pulse Report showed that employees reported they were struggling to get help from HR, with 57% finding it too hard to use the HR portal, and 53% not knowing where to go for an answer. When they call or email the help desk, nearly half (48%) complain it takes too long to get an answer. And while this is an HR example, a Pulse Report focused on IT service management showed very similar results.

Imagine if that same report were done during COVID-19. These numbers would be heightened even further. And why do employees get frustrated when they can’t get the help they need? In addition to just wanting to be productive, according to PwC, “employees want their work organizations to provide a workplace experience that matches what they’ve come to expect as customers and in other areas of their lives.”

By May of 2020, Gartner reported that 55% of CEOs were not prepared for the economic downturn although 43% had reported they were worried about economic recession impacting their revenue growth in the next 12 months.

Despite that concern, many delayed taking action to prepare for this eventuality. That meant CIOs had to move into emergency cost optimization. As a result, investments were minimized and prioritized on operations that kept the business running. Initiatives that provided immediate, tangible ROI were put to the top of the list.

As we move into 2021, cost containment remains a top CIO priority since their budgets have only increased an average of 2% this year. Now in 2021 a new Gartner survey revealed that optimizing costs has become increasingly important to HR leaders as well. When asked about the top business priorities, cost optimization was #4 and had risen in importance 13% versus the prior year.

This demonstrates that optimizing costs is not just an IT concern, it’s an HR concern too. And it’s a concern for the entire year and likely into the near future.

For a virtual agent to “deliver on the promise,” it must be accurate. If an employee asks, “My guest needs internet access” and the assistant responds with instructions to reset a password, the employee will give up.

The accuracy of consumer virtual assistants, such as Amazon Alexa and Google Home, has improved over time thanks to millions of user interactions each day, along with an army of data scientists and computational linguists behind the scenes who go through these interactions to improve the experience.

That is what’s required to achieve high accuracy – an immense amount of data. Unfortunately, no single enterprise has the data and resources to replicate that, which is why many VSAs have had limited success. With over 80% of employee interactions nearly identical across most organizations, it doesn’t make sense to build in-house.

Many natural language processing (NLP) tools build language models based on the knowledge articles they are fed. This means the language model can only be as expansive as the language used within these knowledge articles. But knowledge articles are written by subject matter experts and don’t represent the myriad of ways that non-technical employees express themselves when interacting with a virtual agent. The result? The accuracy rate of the virtual agent would end up low unless employees knew what keywords to use.

So how can you know whether a solution will work for you? Ask vendors to analyze your last quarters worth of tickets against their language model to determine the accuracy rate. In other words, what percent of phrases would be understood by the virtual agent out of the gate. You should expect 70 to 90 percent, with a strong language model at the higher end.

You cannot expect employees to carry around a decoder ring to know what words to use when interacting with your virtual agent. The virtual agent should recognize a billion or more phrases on day one and have the ability to continue to learn and grow in recognition over time with no effort on your part. Without that, you will need to hire expensive AI talent such as data scientists and professional linguists to build a tool from scratch and then to continuously adapt the language model to changes in language patterns.

In addition, the virtual agent needs to also be able to understand the specific language of an enterprise such as proprietary applications, acronyms, buzzwords, etc. And, because you cannot expect employees to constrain their questions to only one department such as IT, the virtual agent should recognize phrases across all enterprise service teams such as IT, HR, Facilities, Payroll, as well as terms that are specific to specific industries such as retail and healthcare. Finally, if your organization is global, you should expect the virtual agent to understand and respond to inquiries in the native language of each employee.

Many virtual agents fall into the lightweight integration category. Their approach is to integrate with as many ITSM tools to maximize their available market using the lowest common denominator.

What’s wrong with that? Lightweight integration has limited value and creates an environment where employees need to use multiple tools in order to get the help they need, causing a break in the experience. For example, employees will start with a modern, engaging experience by asking the virtual agent their question. But when a ticket needs to be created, they will need to leave the virtual agent and revert back to a legacy ticketing experience, usually email based.

Another example can be seen when employees need to order items, such as hardware or computer accessories, from an ITSM service catalog. A virtual agent with a lightweight integration will take the employee to the legacy service catalog in order to browse and order what they need. This again causes a break in the employee experience.

A virtual agent that is deeply integrated with the ITSM platform should provide a single place for employees to leverage all of the capabilities of the ITSM tool, without having to deal with the complexity of the tool. Employees should be able to interact with the virtual agent for a single, unified experience, limiting the number of tools they need to consume in order to resolve their issue or request.

In order for there to be deep integration, the virtual assistant should be able to:

1. Provide a seamless handoff from virtual agent to live agent.

   The virtual agent should give employees immediate, personalized answers to their questions. If the virtual agent doesn’t have an answer, a ticket should be automatically generated and routed to the appropriate service team, providing a seamless handoff from virtual agent to live agent. From the agent side, deep integration means they are able to continue to update tickets and communicate with the employee using the ITSM tool, while the employee can continue to respond through the virtual agent. This promotes a better collaboration between the employee and the live agent.

2. Distinguish between service requests vs. incidents:

   When tickets come in, an ITSM tool needs to distinguish if the ticket is a service request or an incident in order to accurately route them. A virtual agent that deeply integrates will know the difference based on the employee question and will be able to gather specific data from the employee in order to correctly classify the ticket.

3. Consume an existing knowledge base:

   ITSM tools often come with their own knowledge base, such as Confluence for Jira Service Desk. For organizations that have a strong knowledge management practice in place, the virtual agent should be able to consume and improve upon the work that is already done—not let it go to waste. A deeply integrated virtual agent will also leverage knowledge articles across multiple sources, even non-ITSM ones, such as from Confluence and Microsoft SharePoint.

4. Eliminate ticket ping-pong:

   In traditional ITSM tools, 30–40% of IT tickets are not created correctly or routed to the right team, resulting in a ticket ping ponging across teams. A smart, deeply integrated virtual agent can use machine learning to look at historical ticket data from the ITSM tool, along with current tickets and agent corrections, in order to accurately populate and route tickets the first time.

5. Trigger service catalog automations and approvals:

   If your organization has invested in building workflows into your ITSM catalog, you should be able to seamlessly integrate those into your virtual agent, maximizing value. Deep integration means your virtual agent would gather the right data from the employee and trigger the automation you have already built. In addition, some employee requests will require approvals, such as ordering new equipment. Your virtual agent should take corporate protocols into consideration, making sure approvals happen before the request can be completed.

IT departments have sophisticated ITSM tools such as ServiceNow, Jira Service Desk, and Ivanti that work for them, but might be too expensive and complex to deploy across the enterprise. Other internal service providers, such as HR, Payroll, Facilities, etc., need a case management system (i.e., ticketing tool) to improve the efficiency, experience, and quality of support they offer to employees, but need something less expensive that non-technical staff can use.

It is important for a virtual agent vendor to also offer a case management system to fulfill this need. The IT team can continue using their ITSM tool, while other departments can stop using Microsoft Excel spreadsheets and shared voice mailboxes to facilitate service management.

When selecting a virtual agent, you should make sure it can be accessed by employees where and how they already work. Here are some questions to ask a vendor to verify a virtual agent’s accessibility.

1. Collaboration Tools: Is the virtual agent available as an app for popular collaboration tools like Slack or Teams? Can the virtual agent be configured to monitor public channels in order to answer common employee questions? Can employees check the status of their tickets within a single view?

2. Desktop Icon: Can the virtual agent be pushed to employee desktops for both Windows and Mac?

3. Mobile App: Is the virtual agent available as a mobile app for your employees to download? Is the app available for both Android and iOS?

4. Browser: Can your employees access the virtual agent through their preferred browser (e.g., Chrome, Safari, etc.)?

5. Service Portal/Intranet: Can the virtual agent be featured as a widget on a service portal or intranet page? If you have already encouraged employees to go to your portal on these pages, you don’t want to lose that adoption.

6. QR Code: Can your employees scan a QR code and automatically launch the virtual agent? This is especially useful for reporting issues with shared resources, such as a conference room or an office printer.

7. Email: Can the virtual agent intercept an email going to your help desk, read it, and either answer it or create a ticket?

8. Phone: Can the virtual agent be integrated with an interactive voice response (IVR) system in order to intercept calls going to your help desk?

1. Support for Private Channels: An employee should be able to send a direct message to the virtual agent, and the virtual agent should be able to answer or create a ticket when needed. The ability to enable direct, private message is particularly important for departments that manage highly sensitive content (e.g., HR, Payroll) where employees should not ask questions in a public forum.

2. Support for Public Channels: Help desks often have designated public channels within collaboration tools where employees can ask questions and get support. However, agents can become monopolized monitoring these channels and there is a loss of visibility to metrics. A virtual agent should be able to monitor these channels and answer employee questions with metrics, so help desk agents don’t have to.

A virtual agent should also have a designated spot within the app, such as a homepage, that employees can go to and see their pending requests. If they need to provide more information or reopen a request, they should be able to do so from the homepage, without ever having to leave the collaboration tool. A homepage can also be a great way to communicate to employees about important updates, such as announcements and approvals.

A virtual agent that can integrate with an existing IVR system can help deflect calls going to a help desk, freeing up time for both help desk agents and employees. Employees calling the help desk can receive a response from the virtual agent in seconds, instead of waiting minutes to hours on hold to speak to an agent, resulting in higher productivity and less downtime.

Integration with an IVR system should enable the virtual agent to be triggered by the employee simply being placed on hold, or through a deeper integration that enables voice prompts to take specific actions.

For example, if the employee hears, “Push 5 for password reset” and selects this option, the virtual agent could send a notification via push, text, Slack, Teams, or email. The employee could then click the notification to be taken to the virtual agent interface or even into a specific workflow based on their selection.

A virtual agent that can be launched via a QR code can enable quick and easy problem reporting. This is especially important for shared resources, where problems often go unreported since there is no designated owner. Here are three use cases: