Artificial Intelligence (AI) is going to be a game changer in automating support functions and the customer journey, but automation doesn’t always require AI. 

There are different levels of automation, and different types of AI.  The costs to set up and maintain each are vastly different, and it’s important to use them strategically so you’re not adding cost or effort where it isn’t needed.  You don’t need a Ferrari to drive next door.

In this week’s post we’ll explore the different types of automation and AI most applicable to the contact center and the customer journey. 

Scripted Automation in the Contact Center

A scripted automation is program code that executes a series of steps and actions.  For example, a click-to-dial function is a simple scripted automation that saves a user from keying in a phone number. 

Pros  

Scripted automations are very effective at reducing effort and workload by automating common, basic tasks.  They may also be developed for complex decision trees with conditional elements.  The net benefit is the same as other types of automation: cost savings by reducing the demand for human capital.  It can also help ESAT by giving workers tools to make their jobs easier.

Cons

Scripted automations require a software engineer for design, development, and maintenance.  This increased both the cost to deploy the automation, and it extends the timeline for implementation and for support or moves/adds/changes.

Where to Use It

• Gather information in an IVR
• Chatbots answering FAQs
• Data dips, such as order status lookups or product price and availability
• Streamline processes, e.g. the click-to-call

Workflow Automation in the Contact Center

Workflows are built using a graphical user interface (GUI), which makes the build-out and maintenance accessible to more users.  It’s not necessary to know a programming language, the workflows that drive the bot actions are built using drag-and-drop components and simple commands.

Pros

Workflows save time and money by moving the bot development and automation from the higher-cost software engineers to operational leaders.  The workflows can be built and tested faster, and more users are able to understand and support them.  Moves/Adds/Changes may be completed same-day, instead of waiting days or weeks in a developer’s ticket queue.

Cons

Sometimes allowing more people to manage a process opens up opportunities for error.  For example, there could be situations where there are “too many cooks in the kitchen” or where a ball gets dropped because each person thought the other had it.  These can be mitigated through formal SOP and training, but those processes may need to be created if a function that used be to done by IT is moved to Operations.

Where to Use It

• Add conditional elements and context to IVR
• Upgrade IVR to IVA
• Useful in programs where there are frequent changes
  • Growth / scaling
  • Seasonality
• After-Hours Support

Conversational AI: Natural Language (NL) and Large Language Models (LLM) in the Contact Center

Natural Language (NL) and Large Language Models (LLM) are used to transform chatbots from a workflow or scripted automation to a more powerful tool that can hold dynamic back-and-forth conversations. 

Pros

Chatbots powered by workflows or scripts are limited to the programmed question and answer sets.  Chatbots driven by Conversational AI are able to leverage the LLM to decipher free speech and discern meaning.  If it can’t understand the input, it is able to engage in clarifying questions and drill down to confirm customer intent.

Cons

LLMs and NL engines still need to be trained, particularly where a brand’s nomenclature and culture is concerned.  These tools are very powerful, but they are also much more costly.  Conversational AI is not able to generate something new from the conversational input, however, that can be provided with Generative AI.

Where to Use It

• Chatbots and IVA
  • Clarify and capture Intent
  • Caller Verification
• Internal Assistant, Virtual SME for Agents

Generative AI in the Contact Center

Generative AI is able to create things based on input, that is, it generates text, images, music, and audio.  This form of AI leverages LLM and NL, building on those models and also incorporating other transformer algorithms, Variational Autoencoders (VAEs), and Generative Adversarial Networks (GANs).  Generative AI has gained popularity in 2023, since it is the underlying model behind Chat GPT.

Pros

Generative AI can create simple summaries based on complex data sets.  For example, it can take dozens of rows of a customer’s order and interaction history and summarize them in a few concise statements.  It can also answer conversational inquiries, and is even able to respond with a picture, a song, or a poem.  While those elements may not have a direct application on today’s contact center, you can specify parameters, like “summarize this order history in 25 words” which is very useful in a tool to assist your agents and help expedite the CX journey.

Cons

Generative AI is creative, sometimes to a fault.  It is known to “hallucinate” and respond with content that is partially or wholly inaccurate.  It is also prone to encoding the biases or fallacies contained in the training data.

Where to Use It

• Chatbots and IVA
  • Generate summaries for agents based on order and interaction histories
  • Create interaction summaries and notes to reduce or remove agent ACW
  • Full interaction handling and customer service automation

Next week, we’ll dive into the costs and ROI for adding automation and AI to the contact center operations and customer experience (CX).

Can’t wait to talk about it? Click here to schedule a discovery call with the MotionCX strategic solutions team!