How to Get Your Customers to Stop Bypassing Your Bot

A virtual agent icon beside symbols for travel, ecommerce, banking, health, and weather.

With a few simple tweaks to your dialogue flows, you can optimize your conversational AI to satisfy customers, support your human agents, and increase your deflection rate.

A highway to our human agents

When our Principle Solution Architect, Pekka Stenlund, opened a recent customers’ automation dashboard, what he saw gave him pause. Direct escalations, meaning instances where customers had tried to bypass the bot by typing in messages like “human”, “talk to human agent,” had skyrocketed.

“It was like a highway to our human agents,” recalls Pekka, meaning that one certain intents triggered thousands of immediate escalations without the virtual agent having collected any upfront information first. The result? Even higher average handling times and longer queues.

Picking a lane

Pekka’s customer operates in the travel industry, but every industry has its peak season. Customers will tend to bypass bots whenever they feel like their problem is so urgent that it requires immediate attention.

In these cases, a customer-focused dialogue flow can keep customers happy, human agents calm, and deflection rates up. Think of it as equipping that highway to your human agents with a system of lanes, roundabouts, and traffic lights to get everyone where they need to go as smoothly and efficiently as possible.

Here’s what you should keep in mind when creating a customer-focused dialogue flow:

Know your brand

Automation should never come with an at-all-costs agenda, and deflection should never come at the price of upsetting customers. Think of your dialogue flows as an extension of your overall communications philosophy. Does your business stress a no-frills approach? Are you known for your premium service?

While one business may decide to escalate immediately, even if this means that customers may need to queue for a while longer, another may decide to build one - or several - loops into their conversational structure. This can decrease average handling times by letting human agents step in at a later phase of the query.

“When I worked as a customer service agent, instructing customers over the phone where they might find their customer number to proceed with their case could take ages.”

- Pekka Stenlund

Building on this experience, he ended up installing a loop for his customer in which the virtual agent would kindly ask for upfront information several times before eventually escalating queries, freeing up human agents’ time in the process but leaving the door open to escalation all the while.

Know your customers

Researchers have found that customers value two elements of conversations with virtual agents. The first involves pragmatic aspects, such as efficient assistance and accuracy in interpreting requests. The other is related to tone of voice.

Depending on customer demographics, one or the other may carry more weight. Younger customers are more likely to warm up to automated customer service in the first place; they’re also more likely to be affected by your virtual agent’s tone of voice. Those users more skeptical towards tech in general may place more emphasis on functionality and clear instructions, and simply won’t be swayed by charm.

Here’s where training your AI model comes in — the more accurate your interpretations of incoming messages, the more useful your conversations, the happier your customers.

Align expectations for customer-focused dialogue flows

Based on your key demographics, you should ask yourself which conversational flow best aligns your brand’s style of communication with your customers’ expectations. Then, you can complement your communication strategy and tone of voice by:

  • Creating different flows for logged in vs. not logged in customers.
  • Building in a loop (or two, or three) asking customers to rephrase their question or to specify the information they are looking for if possible.
  • Training certain intents or expressions that will end the conversation, just as a human agent would. No customer service agent should have to deal with expletives, human or virtual.

Research on consumer trust in virtual agents has also revealed that the more positive their experience with a bot, the more likely they are to engage with it in the future.

Building customer-focused flows will not only help your business in the short term, but also make conversational AI and automation-enhanced customer experiences more popular across the board. A win-win for us all.

Learn more best practices from our collaborations with customers