Blog

Try before you bot: why a conversational AI POC is a must-have in 2020

July 7, 2020

Starting off with a trial or proof-of-concept is the best way to ensure you find the right virtual agent for your business and avoid complications along the way.

As the business world continues to adapt to the new state-of-play brought about by the impact of COVID-19, one thing remains certain: the need for automation in customer service is in high demand. Recent data from McKinsey has shown that digital adoption in both the consumer and business sectors has been propelled five years into the future in a matter of just eight short weeks since the pandemic hit. With things moving at a breakneck pace, companies are looking to balance their bottom line while maintaining the safety of their customers and employees.

Automating customer service, support and sales with conversational AI presents a unique opportunity to achieve this goal. With many contact centers and support desks either transitioning to remote working or being shut down entirely, an artificially-intelligent virtual agent can act as a business’ first responder. Available 24/7, this digital colleague can assist customers with common questions and transactions that would otherwise take up time for human support reps to respond to, reducing call queues and leaving staff free to attend to customers with more sensitive requests.

Of course, for those businesses that haven’t already adopted a conversational AI solution pre-COVID-19, it presents the challenge of how to transition to this new digital customer service strategy quickly while still delivering a genuine return on investment. Even before the current pandemic, the model of selling an expensive conversational AI project - complete with a pricey implementation cost - and hoping to see a decent ROI over the course of several years was already beginning to look tired. In 2019, Gartner predicted that 40% of the chatbot projects launched in 2018 would be abandoned by 2020, which is meant not as a slight against the potential of the technology, but more so against the fact that most vendors in the space tend to overpromise and underdeliver. Anecdotally, we have encountered a large number organizations that have spent years building with a solution only to find that, come launch, they are unable to effectively scale it to their needs. Combine this with the cautious attitude that the current climate encourages and it’s no surprise that businesses are less inclined than ever to invest money and time into a virtual agent before they can ensure that they will start to see its benefits from day zero.

Try before you bot

So, how can a business interested in deploying a virtual agent get started? And, importantly, how can they do so without having to invest large amounts of time and money? The answer is simple, approach a vendor and have them develop a proof-of-concept (POC) for a full-featured virtual agent tailored to your requirements. Be sure to outline clear goals and KPIs, and insist that their conversational AI can at least offer the following:

  • Natural language understanding (NLU) that can scale to handle thousands of user questions simultaneously
  • Self-learning capabilities that can be leveraged to do most of the heavy lifting in developing and maintaining the virtual agent
  • A turnaround of a few weeks (not months or years), from project start to go-live

If your criteria are not met, you should be under no obligation to move forward with the project. The onus should be on the vendor to prove that their solution will benefit you in the short and long term. This idea of a ‘free trial’ is nothing new. In fact, it’s something that has been a standard in the enterprise SaaS space for decades but has, thus far, eluded most conversational AI vendors.

Some platforms offer access to a sandbox environment where you can try to build a chatbot from scratch within a limited time period. This tends to be unrealistic because you are expected to learn how to use completely new technologies and scope out a comprehensive implementation plan in only 30 days. It equates to having all the materials necessary to build a house but no blueprint to follow - the results will naturally be subpar, at best. Other vendors offer limited solutions unable to handle high levels of complexity that can only answer questions on a few hundred topics. Useful for small businesses that have basic customer interactions, but unsuitable for larger organizations who’s automation requirements quickly scale up as the needs of their customers evolve.

If a trial is to be genuinely useful (and not just a sales tactic) in helping you decide whether or not to implement a virtual agent, it needs to be bespoke to your business and demonstrate the true value that automation can offer. You want a solution that meets your needs not just today, but one that is agile enough to allow you to successfully navigate whatever challenges may arise (be they a pandemic or otherwise) well into the future.