Adopting voice bots or virtual agents makes a lot of sense for a wide variety of companies, particularly those with high and varying call volumes. Rather than trying to hire and train temporary staff for peak periods, virtual agents act as a permanent pressure valve, deflecting high call volumes away from existing staff so they can focus on the high-impact calls.

“Used properly, these virtual agents save time, money and resources,” says Ryan Falkenberg, CEO of CLEVVA. “Increasingly, they’re capable of meeting customer demands to the same standard as human agents and can be scaled up and down almost instantly. You can see why that might be appealing to a financial service provider, for example, which has experienced a public data breach. The breach itself would have already resulted in a loss of customer confidence. Leaving irate customers to deal with hours-long queues would only make that worse.”

Where should companies looking to embrace voice-powered virtual agents start?

What if it does not work and infuriates customers? A practical and low-risk way to get started is to channel specific calls to virtual agents to handle. Falkenberg suggests starting with one then add another. “As you and your customers build confidence in your virtual agent, you can assign more call types for full voice automation.”

The trick is to start small and scale quickly

Voice automation is improving daily. Just try talking to ChatGPT 4-O to see how far it has already come. The challenge, however, is to ensure every conversation sticks to your guardrails, answers correctly and can offer a detailed record to prove it. To achieve this, you need to work with specialised technologies designed for contact centre engagements.

Then start with the simplest queries and expand from there. The financial services data breach example cited above could include queries like “Is my data impacted by the breach?” or “How do I reset my password?”

Falkenberg says this is not because of any technical limitations with these specialist virtual agents. Today’s offerings are capable of handling increasingly complex queries because of the rapid advances in artificial intelligence and conversational process automation technologies.

Instead, it’s because these simple queries are often the most frequently asked and where the conversation requires you to walk the customer through a specific process in context. Human agents can add very little value to these conversations other than following their scripts. Allowing a virtual agent to handle these conversations frees up human agents to deal with the conversations that require a more complex, nuanced and sensitive touch.

Falkenberg says another advantage of starting with simple queries is that companies can review the data on how well the virtual agent is performing and adjust quickly. If the call logs show that customers are hanging up at a certain point in the conversation or they are asking to speak to a human agent, the team can analyse why this is happening and adjust the logic to improve the outcomes. They might realise the virtual agent is not understanding certain accents or is failing to respond to negative sentiment. Within a few months, you can perfect the patterns and then apply these to new call types.

Voice automation is gaining pace

A year ago, very few companies could boast about having a virtual agent capable of a voice conversation with customers at the level of human agents. Most voice bots struggled to understand or “hear”. And when they did, they had a very formulaic, menu-driven way of conversing. It felt like you were talking to a digital IVR – the technology used to channel your call with prompts like “press 1 for accounts; press 2 for sales”.

Today with huge advancements in language understanding and speech automation, accelerated by generative AI, we are seeing virtual agents capable of human-like conversations. Importantly, these conversations stick to the business rules and ensure the customer’s query is resolved without transferring to a human agent.

“Starting the virtual agent process sooner, rather than later, will help your teams learn how to perfect their conversations and stay true to your business rules,” adds Falkenberg. “By initially channelling a single call type to a virtual agent, and learning off the data, you can learn quickly and scale fast.”

Virtual agents are already used to talk through the Ts and Cs in sales calls, to do the follow-up calls required for document gathering and data collection and the calls required to resolve many of the high-volume, low-complexity queries that flood the centre and consume human capacity.