Want to see how we do chatbot analytics? Join us for a product demo webinar.

Why are there so many terrible virtual agents

You can’t go too far into a major brand’s Twitter feed without people complaining about their virtual agent. 

The expectation gap between what a company aims to achieve and what users want from bots remains wide, but you can avoid making common virtual agent mistakes.

This story is all too familiar. A giant brand launches a shiny new virtual agent after much testing and development to meet a growing business need and key customer service goals.

Within minutes of launch, real customers are complaining about how useless it is. While it might achieve some of the company’s goals, the builders failed to address the key needs of customers. And, for many firms, when they signed off on the project, the resources moved on to other things, making fixing the issues more complex than necessary.

The reality is there are many limited, poor, if not terrible, virtual agents out there, often due to companies jumping on a bandwagon without understanding what they’re getting into. Poor design choices, expecting non-experts to become virtual agent masters and real-world testing remain thorny issues for many.

On the plus side, there are a minority of better virtual agents demonstrating strong customer experience and high levels of customer success.  These can serve as a guide to others.

Why do so many virtual agents suck?

There are a number of reasons:

1. Decision-makers think that the virtual agents platform is all they need

Wave the term AI at many people in business and they think it is an easy solution to a pressing problem. But, as with all technology, it is generally more complex than talking to Siri and getting an answer. Similarly, a virtual agents that duplicates an established simple process is not the same as duplicating and presenting the knowledge of an experienced customer service agent.

A virtual agent platform is a good place to start with its basic building blocks like natural language processing (NLP), but may not be strong at supporting virtual agent business operations. All the witty banter in the world won’t help when it comes to results analysis, prioritization, testing, and reporting. These are all key elements and business basics that have nothing to do with swanky live translation or an NLP feature.

NLP vendors touting their smart virtual agent platform often lack experience delivering business results. While they might expect you to integrate the bot with other IT business services, they likely lack the experience or cannot afford to spend time helping you integrate.

Back in 2018, Gartner suggested that “around 40 percent of chatbot applications will have been abandoned by 2020” and there’s little to suggest they were wrong. Among the vanished are eBay’s Shopbot and Microsoft’s Tay. Yet, many well-designed bots thrived. Take KLM’s BlueBot that saw query handling rocket at the start of the COVID lockdowns from 8.1% to 22%, supporting or solving 23.5% of customer queries.

The key difference here is the team supporting it.  There is no substitute for a dedicated and experienced team managing a virtual agent full time.

2. Focusing on a quick win and not planning for the long run

Many virtual agent projects start as a pilot to deliver a quick win for the business. Typically a bot can reduce support costs by 50% or more. The focus on speed often means not establishing key processes for the long-term maintenance of the virtual agent which will negatively affect the customer experience.

As with customer service, IT teams and virtual agent developers often speak slightly different versions of the same language. There can be confusion about what the bot should do and who is going to do it. These potential issues could arise from design decisions or technical uses over intent, which are the use of unclassified terms, as well as other problems.

Feature creep will eventually show up, with leaders trying to add new capabilities to their bots as new ideas show up. While these might make it smarter, they can impact the customer experience and create other unforeseen issues.

Planning for the long run from the outset of a bot program is they key to avoiding these pitfalls.

3. Leaders think bots are simple projects suitable for an inexperienced team

Since the perception that virtual agents and AI are simple tools to deploy persists, digital, customer service or IT leaders assume they can leave it to a small or inexperienced team. As with any technical project, experience matters. People need to understand the processes and tools as well as what customers expect.

That approach will create problems as virtual agents fail to deliver high-quality customer experiences, project teams fail to interpret the data correctly, and part-timers walk away to their next project at the company.

A small part-time team cannot support bots that are supposed to deliver millions of successful conversations a year. It requires a long-term commitment with expert input and knowledge at all stages, plus management oversight and clear goals to deliver success.

4. Bots don’t have the connectivity to back office systems they need

The modern virtual agent can deliver a complete range of services if connected to the right APIs accessing back office systems. They can perform transactions, present customer-specific data, personalize content and show products and solutions that might be relevant to a specific customer.

However, making these more advanced capabilities a reality requires integration to back office systems.  Without these connections, a virtual agent will be left to its most basic functions, acting as a place for frequently asked questions.

Planning for these connections when rolling out a virtual agent program are critical for long-term success.

Get the professionals for the best virtual agent experience

When a company puts out a superlative-laden press release about a new virtual agent, the first thing a certain type of person will do is try to break it.  Unfortunately, it’s usually all too easy to find holes in the new system.  If you are worried about being embarrassed by your virtual agent, then ensure it gets the planning and expertise a customer-facing product deserves.

Outsource the most difficult parts of AI operations

And we’ll guarantee you that your virtual agent will meet or exceed the mutually agreed KPIs. That’s true Wysdom.

Get started today.

Subscribe to our newsletter

Get inspired

Read more insights on the latest in virtual agent analytics and performance management

Subscribe to our newsletter

We use cookies to ensure that we give you the best experience on our website. By clicking “I Accept” or if you continue to this site, we will assume that you consent to the use of cookies unless you have disabled them.