The Top 5 AI Pitfalls in the Contact Center and How to Avoid Them

Organizations across industries are leveraging artificial intelligence to both improve customer and agent experience and reduce labor costs. Today, AI is becoming embedded into business processes and driving the growth in enterprise automation, helping teams deliver speed, quality, and resilience.

I'm excited about what AI is doing for the contact center. Yet, it's important to remember that how you deploy AI to meet your needs is not just about the technology. To be successful you must think about the full picture, from developing the models to training the AI, having enough data, and more.

Customer experience professionals have quickly embraced a new lexicon. Machine learning (ML), natural language processing (NLP), text-to-speech, intelligent virtual agents (IVAs), and agent assistance are all components of AI's foray into customer service. These technologies can drive true impact when deployed with a clear eye toward the customer experience.

We're seeing some of the biggest improvements in call-routing accuracy; solving routine customer requests without human intervention; and analyzing the content of conversations to provide agents with real-time assistance during interactions. Conversational AI delivers value via self-service IVAs and with agent-assist tools, where bots provide agents with real-time information, recommendations, and guidance during complex calls. IVAs directly help the customer, while agent assistance tools help the agent help the customer.

These things have demonstrable ROI. For example, if you can help guide your agents in real time with meaningful information, you can reduce average handle time and hopefully improve your customer satisfaction score.

AI itself does not inherently improve CX, and not every vendor with an AI offering is equally mature and positioned to deliver practical solutions to improve CX. To really get the most out of AI, the deployment must be practical. The improvements need to be tangible, measurable, and deliver actual business outcomes that correlate to improved CX. They can't just be based on the hope that AI will improve things.

Wherever you are in your AI journey, ensuring maximum value relies on avoiding common pitfalls. Here are the top five I see in enterprises today:

Pitfall #1: Looking at AI as an add-on technology.

It's easy to see AI as just another tool to add to your contact center toolkit. But AI is more complex than that. AI fundamentally transforms how your contact center works and alters the customer experience journey.

Adding conversational IVAs to your contact center is, in practical terms, adding a new &digital workforce that can collaborate with your agents and deliver service alongside them. Would you just randomly add more untrained live agents to your contact center without a plan to train them, get them ramped up and make sure they are successful? No. So why would you expect to add AI and just have it work?

AI shifts the way work is done and creates the opportunity to re-think the purpose and role of the contact center. For instance: If AI handles the bulk of repetitive, transactional calls, are your live agents ready to switch focus and provide a proactive, outbound service? Do they become sales agents as well as service agents? Start with the big picture and think through the ripple effect of offloading huge swaths of work to AI.

At the same time, AI itself needs a plan. Walk through exactly what the customer will experience. Does it feel easier? Or does it feel like one more layer to get through before they can get a real answer? Are you really deploying the AI where it will matter most?

Pitfall #2: Implementing AI without practical, tangible, measurable business outcomes.

AI arrives with a lot of bells and whistles and futuristic idealism. However, if it's not clear how the AI will provide a better CX and deliver measurable business outcomes, don't buy it. Insist that any AI solution you deploy can demonstrate value and help you improve key metrics.

For example, can the AI do the following:

  • More accurately route calls to the right agent the first time?
  • Improve call handle time, first-call resolution, or call abandonment?
  • Completely automate certain tasks that will allow you to reallocate agents to focus on higher-value interactions that require more empathy or drive more revenue?

Make sure your vendor can provide proof points and customer references to back up their claims.

Pitfall #3: Considering AI as an infallible, one-and-done implementation.

AI isn't magic. The reality is that AI makes mistakes—it's the only way AI can learn. And the models upon which you build the AI must be continually trained to learn and improve outcomes.

When deploying AI, consider questions such as the following:

  • How does the technology handle mistakes?
  • How is it trained, and who is responsible for training it?
  • Can it work in real time, and does it provide agents oversight to ensure accuracy?

If AI is creating automatic call summaries, human agents need the ability to quickly review a summary for accuracy before it's placed in the CRM. This simple step will ensure accurate information and help the AI learn from any corrections made so that it can continually improve.

Pitfall #4: Cutting corners on change management.

AI changes a lot of things for both agents and customers, and it's not always an easy transition. A key thing to emphasize is that AI is not designed to replace humans entirely but to assist them and free them to engage in more valuable customer interactions (this is true even in the case of IVAs; there are very few if any CX practices that are run on AI alone). Rolling out an AI initiative and just telling agents how great it will be won't quell their anxiety of being replaced by robots. If they view AI as a competitor, they will be less likely to adopt it fully.

Walk people through the changes AI creates and bring them in on the process. AI might be easier to manage and easier to scale once it gets going, but it can never replace the empathy and kindness that people have to offer your customers.

Likewise, include your customers in change management. Let them know you're creating new ways of engaging with them and give them the opportunity to provide feedback. When a call is transferred from an IVA or a bot to an agent, ask the customer if the AI was helpful. Acknowledge that AI isn't perfect and let customers know that you are working to continually improve it.

Pitfall #5: Neglecting to ask AI vendors key questions.

AI is baked into almost all cloud contact-center solutions. But the AI maturity level varies by vendor. Not all AI is built the same nor offers the same flexibility needed to truly scale and grow as the AI itself develops over time.

Ask vendors these questions:

  • Which conversational AI technologies do you offer? Is there flexibility to switch between the underlying platforms at any time—Google Dialogflow, IBM Watson, Amazon Lex, etc.? You don't want to be locked into one AI vendor.
  • Can your platform easily integrate with my back-end systems? It's critical that it can and do so out of the box. There should be documented APIs or SDKs to support a quick start.
  • Does your underlying platform enable shared components with one common workflow and one set of integrations? With agent assistance tools, agents should be able to use one screen to access everything they need to help the customer. They shouldn't have to jump around into a new app (unless you want them to).
  • Does your platform allow non-technical users to make simple changes to the applications? You shouldn't have to wait on a highly skilled developer or your IT department to make quick updates.

AI has the potential to offer some solid CX improvements, but only when you do it well and keep the human experience front and center. A thoughtful, strategic, and practical approach to choosing and implementing AI will yield the best results.

AI lets us automate, respond, and move faster than human capacities, and when put into the service of helping customers and agents get accurate answers faster, it positions companies to excel at customer experience. Avoid the common pitfalls and design and deploy an exceptional AI experience for your customers.

Jonathan Rosenberg is chief technology officer and head of artificial intelligence at Five9.