why virtual assistants fail blog header

Table of Contents

Why Virtual Assistants Fail (and How to Avoid Them)

Peiru Teo

Peiru Teo

If not done rightly, most virtual assistants can fail to deliver customer satisfaction. Learn the top reasons why and how you can avoid them.

It is well known that virtual assistants (commonly known as chatbots) are seeing increasing adoption among businesses today. Everyone from e-commerce stores to SaaS startups to large healthcare institutions and insurance firms is experimenting with the technology. 

In fact, there was an increase of more than 160% in interest in implementing virtual assistants and associated technologies in 2018 from previous years, according to Gartner’s analysis in 2019. The main drivers of this increase were customer service, knowledge management and user support, all of which are relevant use cases in any modern organization. 

There was an increase of more than 160% in interest in implementing virtual assistants and associated technologies in 2018 from previous years  

The fundamental benefit of virtual assistants is that automating repetitive conversations and transactions allows companies to focus on more strategic and revenue-generating activities. 

But not all virtual assistant implementation projects in organizations come to fruition. Many of them begin as proof-of-concept trials and get abandoned after that. In other cases, they move to the next stage to cover more complex tasks but end up being seen as failures and are just kept dormant without further improvements. Another Gartner report in 2018 found that around 40 per cent of virtual assistant applications will have been abandoned by 2020. 

Why do such virtual assistant implementation projects fail or get abandoned? Here are 7 common reasons based on our experience.

Evaluation purely on technology related results

It is not uncommon that the goal for a virtual assistant is set to be around the technology itself during the vendor selection phase, with what the customer and business want being sidelined till the project launch.   

For example, it is common for businesses to request the “smartest” virtual assistant without defining what is smart. In some cases, the goal for the virtual assistant is set to try and be as funny and natural sounding as possible. While these may be ideal end goals to strive towards, they may not be universally effective from the get-go. Besides, even if they are requested by decision-makers in the organization, they may not be serving your customers well. 

At the end of the day, a virtual assistant is as good as the value it provides for the end users. It is therefore important to think about the goal of this project from the customers’ perspective. What do you want the virtual assistant to do for the customer? Is it expected to respond to every query quickly, or is customer satisfaction a better business goal to aim for? 

There is no clear alignment on success metrics

Similar to business goals, there could be a misalignment of the success metrics planned for the virtual assistant. We have discussed the importance of looking beyond vanity metrics, especially when progressing from a “Phase 1” FAQ-type virtual assistant to a more comprehensive implementation.  

Additionally, the metrics planned should not be limited to a short-term campaign, such as a promotion drive to accelerate lead generation during a peak season. Think of virtual assistants as serving a much bigger purpose and one that needs to be approached with a long-term engagement strategy. This often requires a dedicated team, either within the organization or from a vendor, to be in charge of monitoring the results of trials and improving performance over time in a test-and-learn approach.

Unrealistic expectations breed confusion

Compared to other recent technologies, virtual assistants are still relatively new. Organizations may be familiar with traditional virtual assistants that can produce responses based on pre-set rules but they may not fully grasp the capabilities of more modern incarnations of virtual assistants. 

Specifically, virtual assistants that employ artificial intelligence and machine learning can still seem a bit abstract in their technologies. A number of related terms such as self-learning, reinforcement learning and deep learning can get mixed up and create confusion. 

There is also a tendency to expect virtual assistants to be a magic wand that can answer any questions asked of them, like a personal butler or a science fiction character. And that too without giving adequate training. Despite the advancements in artificial intelligence, we are not at a stage where machines can fully simulate an intelligent sentient being.  

Therefore, business users and management need to be realistic in their expectations on what the technology can do. This requires a bit of education and basic training that could be provided by a vendor involved in the implementation project.

There is a lack of continued commitment from business users

Another common reason why virtual assistant projects fail relates to change management. It’s never a case of “build it and they will come”. Humans are creatures of habit and it is much harder for us to change our ways to adapt to a new tool when we are used to an existing method.  

Quite often business decision-makers are driven by the hype at the start but never fully make the shift. They either revert back to their old ways of communication such as email, or in the case of some insurance firms, are compelled to use phones as a channel to accommodate users.   

We encounter similar problems when we deploy insurance virtual assistants where customers and agents circumvent the virtual assistant interface and request to talk to a live agent. The same is often true in a healthcare setting. A McKinsey survey in 2020 found that 44 per cent of healthcare professionals in the EU who engage in healthcare innovation had never been involved in the development or deployment of an AI solution in their organization. 

A McKinsey survey in 2020 found that 44 percent of healthcare professionals in the EU who engage in healthcare innovation had never been involved in the development or deployment of an AI solution in their organization. 

This problem highlights the importance of involving the actual end users in the development of a virtual assistant at an early stage. It helps to at least conduct an assessment to understand whether they are experienced and willing to use it in their work.   

A related consequence of the diminishing enthusiasm from the top management and end users is that the technology teams in charge of support and maintenance also cease their work on the virtual assistant So once the build is done, they move on to another project without investing enough time maintaining and improving its content, just like one would update a website or undergo training for the company’s latest products. One way to avoid this particular scenario is to explore managed service models with key points of contact within the company, where a vendor can first help build the platform and subsequently manage it for ongoing improvements.   

PRO TIP: Engage a vendor who can first work with your organization’s management and end users to build the virtual assistant platform, and who can also maintain and improve it in a managed service model.

Vendors are chosen for single use case

The low barrier to entry in the virtual assistant landscape has given rise to a number of vendors. They range from those providing simple out-of-the-box solutions to more advanced conversational platforms and enterprise-grade expert virtual assistants that leverage artificial intelligence.  

The smaller players may promise expertise in narrow domains and single-use cases. The tools they use are often free and licensed from providers such as Microsoft, Google, Amazon and IBM in pay-as-you-go APIs. But ultimately, these individual bot applications may not be capable of delivering significant business benefits by expanding into other areas and uses cases within the organization. For example, they may deploy an IT support or HR virtual assistant for an insurance firm but could miss out on a number of other insurance use cases of conversational AI.    

This may even create a situation where virtual assistants proliferate within an organization, tackling multiple instances of single-use cases but end up delivering inconsistent and often poor performance, wasting resources and overlapping functionalities. One example is a billion-dollar enterprise from Europe that had to manage 30 virtual assistants from 10 different vendors at one point. They had to undergo a major overhaul to consolidate. 

PRO TIP: Look for virtual assistant vendors with proven expertise in not just single-use cases but also expansion to benefit other areas of their business. 

Vendor versus partner relationship for long term growth

The expertise to deploy and maintain the virtual assistant is a good criterion when selecting the vendor. But the best providers go the extra mile and provide the right guidance in the implementation process and understand the industry’s priorities and trends – from strategic planning to developing the tools and post-launch support.  

An ideal implementation process starts with strategic planning and business assessment to map out the process flows with various stakeholders along with sharing of best practices. This is followed by a data gathering and testing phase involving internal stakeholders, tuning of the NLP and ML algorithms and continuous re-training and improvement of the virtual assistant.  

Then comes the configuration of the virtual assistance flow, design and logic flows and integration with existing systems. Finally, once the testing of the virtual assistant and the user training is completed, the implementation is considered to be done. But there’s still more to be done once it goes live. Post-launch activities should include documentation, support and ongoing maintenance.   

The best virtual assistant providers are those who bring their expertise to act as advisors who can help you chart the implementation, deploy and scale on multiple channels and also provide long-term support after launch.

There is a lack of platform extensibility and connectivity

Virtual assistants on their own may be able to produce responses from an FAQ database with a high level of accuracy if it is well trained. But this is only scratching the surface of what they are capable of and the true value they can bring to a business. This happens when organizations implement virtual assistants into their existing technology portfolio without ensuring that they can integrate with the workflows, communication channels and existing technologies. 

For example, they could be embedded on a website, a mobile app or any of their existing channels such as their WhatsApp Business account, Facebook Messenger, SMS or other platforms like Slack, Facebook Workplace and Line. It’s only when they are properly deployed on these channels can the true benefits be reaped for the business.  

It is also important for it to be connected to a live chat platform for seamless handover from the virtual assistant to a live chat agent. Besides that, it is also important for it to be integrated with the key systems within the company to perform tasks to help users end-to-end without redirecting them to another site. 

Thus, it is important to look for a platform that can be customized and provide an extensible architecture to be able to integrate into these other channels without requiring additional time and resources.

Learn more about KeyReply

To see our Conversational AI in action, sign-up for a guided walk-through and see how our platform can be tailored for your organization’s needs