⚠️ Unsupported Browser

Your browser is not supported.

The latest version of Safari, Chrome, Firefox, Internet Explorer or Microsoft Edge is required to use this website.

Click the button below to update and we look forward to seeing you soon.

Update now

A Complete Guide to Conversational AI for Customer Service

1. Overview

What is Conversational AI?

Conversational AI refers to a suite of technologies that use artificial intelligence to enable machines to communicate with humans. 

At a very basic level, it works by deciphering the meaning behind a given written or spoken utterance and responding in a natural manner to continue the conversation. Modern advances in conversational AI technology allow for machine-human conversations that are intelligent and feel natural. 

How Conversational AI Can Be Used

Chatbots


Customer support chatbots found in apps or on websites

Smart Assistants


Smart assistants like Google Assistant or Amazon Alexa

Voicebots


Customer support voicebots that handle customer queries over the phone

Conversational AI for Customer Service

Conversational AI can be used to power voice assistants or chatbots that fully or partially automate common customer queries or transactions. 

By automating common common customer service calls, companies can expect to:

  • Free up agents to focus on more complex, empathy-requiring tasks
  • Serve customers quicker
  • Gather in-depth insights into customer behavior 
  • Deliver personalized experiences based on data
Conversational AI example between a PolyAI voice assistant and customer, showing intent and value detection.

Conversational AI example between a PolyAI voice assistant and customer, showing intent and value detection.

2. Voice

Voice-Based Conversational AI

Conversational AI has developed in leaps and bounds in recent years, allowing voice-based conversations between humans and machines that feel natural and human. 

Voice-based conversational AI has been widely accepted by consumers in the years following the launch of Siri in 2011.  Now, research shows that over 56% of US adults use a voice assistant on their smartphones and over 38% of UK adults own a smart speaker. 

Enterprises have been slower to adopt this technology, but in recent years companies from BP to Landry’s Golden Nugget Hotels & Casinos have found real world applications for conversational AI in their customer service departments. 

Voice automation for customer service

Many companies turned to text-based chatbots to reduce the cost of handling customer calls, but studies have shown that only 37% of organisations successfully steer customer support call volumes to new digital channels like chatbots. Furthermore, 11% of contact centres have actually seen an increase in calls after implementing new digital channels! (Source: 2019 Global Contact Center Survey, Deloitte Digital [PDF])

The truth is, customers are still calling – and they’ll continue to do so. This is particularly true during times of uncertainty, as we’ve seen during the COVID-19 outbreak where 57% of customers ranked call support as their preferred contact method.

Companies that want to continue to serve customers on their preferred channels are now turning to voice automation to deliver great customer experiences at lower costs.

Your guide to voice assistants for customer service

Learn more about voice assistants for customer service

Building Conversational AI for Voice vs Text 

Humans inherently know how to hold successful conversations, but replicating natural speech through voice interfaces is incredibly difficult. The complexity of this challenge increases exponentially when you consider the way that humans regularly interrupt, change topics and use slang over the course of a single conversation.

In order for voice assistants to hold conversations with multiple turns, they need to both understand what users are saying and give a relevant response that moves the conversation one step closer to resolution. 

Until very recently, it has been near impossible for companies to automate voice transactions for the following reasons: 

1. People don’t talk how they type

People are typically more direct in writing, but in verbal conversations they chop and change unpredictably. Conversations don’t follow linear paths, which makes them more difficult to design and pre-empt. 

2. People have different accents and cadences of speech.

While advancements in speech recognition have reached impressive levels of performance, it’s not perfect. Which can cause all types of problems, for example, when a customer says ‘for’ but the computer recognises it as ‘4’.

3. There’s no graphical user interface for voice

Chatbots often make use of buttons to show customers their options, or link off to longer pieces of content where customers can self-serve, but voicebots can only rely on spoken utterances. 

Addressing these challenges is only possible using technology that is specifically designed for voice interactions.  

3. Technology

The Technology Behind Conversational AI

Conversational AI is built on a number of components to enable it to work, from speech recognition to intent detection right through to a spoken or written response.

The following are the most common components in the conversational AI tech stack.

Speech-to-text

Speech-to-text technology transforms spoken utterances into text transcriptions.

Natural Language Understanding (NLU)

Widely referred to as NLU, Natural Language Understanding is the process by which technology is able to understand natural human language. 


NLU is especially critical in voice interactions where speakers may not use specific keywords, or may tell longer stories to get to the point.

Intent

Within conversational AI, intents define what actions should be triggered based on conversational inputs.

Intent detection

Intent detection is the process by which the bot correctly picks up the intent behind an utterance.


Intent detection is typically more challenging in voice than text, due to our natural proclivity for telling longer stories.

Value extraction

Value extraction is the process by which AI agents extract the relevant information from customer queries and store them against the relevant ‘slots’.  


In speech, callers will often give multiple values in one utterance (e.g. I want to book a table for 2 people for tonight). Great voice assistants must be able to extract multiple values in order for a conversation to feel natural.

Text-To-Speech (TTS)

Text-to-Speech technology transforms written text into spoken utterances. 


Off-the-shelf text-to-speech solutions often sound robotic. Voice actors can be used to record on-brand, natural sounding responses.

Multi-turn conversations

A ‘turn’ in a conversation is marked by one back-and-forth interaction: the user speaks and the bot follows, or vice-versa. If more than one turn occurs in an interaction, this becomes a multi-turn conversation.

Context

In order to hold natural-feeling conversations across multiple turns, conversational bots must be able to carry context throughout a conversation. 


Context is particularly important in the voice channel, where chat history is not displayed to the customer.

Dialogue policy

Dialogue policy informs the flow of a conversation, allowing the bot to intelligently guide a caller through a transaction.


A great dialogue policy is what will allow a caller to interrupt the flow of a conversation, for example by asking a clarifying question (where do I find my order number?) that may not have been predicted by the bot developer. 

4. Capabilities

Conversational AI Use Cases in Customer Service

Conversational AI can be used to handle many common customer service transactions, providing instant support for customers while freeing up agents to focus on more complex tasks. 

The most common use cases for conversational AI in customer service are as follows. 

Intent Detection & Call Routing

Voice assistants use conversational AI to determine why a customer is calling and route them to the right department. This is similar to a conversational Interactive Voice Response (IVR), but using conversational AI technologies like NLU allows customers to use natural language instead of keywords. 

Conversational AI can also be used to take important details – like customer names, account numbers and related products or services – and forward this information directly to the agent, along with the call, to save time for both the agent and the customer.  Voice assistants can even start troubleshooting before transferring, increasing self-service to prevent simple requests from reaching an agent altogether.

Learn more about intent detection and call routing, and listen to a call with a voice assistant. 

Bookings and reservations

Voice assistants and chatbots can use conversational AI to take bookings and reservations over the phone or via live chat. These bots integrate with booking systems to fully automate the reservation process, and are especially useful where bookings are typically taken by front-of-house staff who are juggling multiple tasks. 

Voice assistants and chatbots can also offer personalized service to repeat customers or members, and intelligently upsell products and services in the booking process.

Learn more about bookings and reservations, and listen to a call with a voice assistant. 

Authentication

Voice assistants can use conversational AI to securely identify and verify customers through natural conversations. Call center agents can spend up to 60 seconds per phone call authenticating customers, so automating this process with a voice assistant saves a considerable amount of time and money.

Authenticating users through natural conversation is a cheaper way to automate the process than voice biometrics, and doesn’t come with the same set of risks that voice biometrics presents.

Learn more about identification & verification through conversational AI, and listen to a call with a voice assistant.

Troubleshooting

Troubleshooting can be a very repetitive and lengthy process for agents. Voice assistants and chatbots can fully or partially handle the troubleshooting process, sometimes even resolving the issue without ever having to transfer to an agent.

If the issue cannot be resolved by the bot, it can transfer the customer to an agent along with a record of what has already been tried, and what the results were.

Learn more about automated troubleshooting over the phone, and listen to a call with a voice assistant.

FAQs

Automating FAQs with a voice assistant or chatbot allows customers to get answers to their questions fast and frees up live agents from answering repetitive questions. The best conversational AI technologies allow customers to ask questions naturally, instead of having to use keywords to get the answer they need. 

Good conversational AI vendors will be able to add new content or make changes to FAQs quickly and easily to deal with emergencies, new releases or updated offerings.

Learn more about automated FAQs over the phone, and listen to a call with a voice assistant.

Account Management

Chatbots and voice assistants can be used to help customers change their details, upgrade their accounts and learn more about account services. 

Voice deployments will require advanced technologies to enhance speech-to-text capabilities, ensuring that details like addresses and phone numbers are recorded accurately. 

Learn more about account management, and listen to a call with a voice assistant.

Order Management 

Voice assistants and chatbots can be used for customers wanting to place, track, edit or cancel their orders. These bots integrate with your technology to identify callers, pull real-time updates and make edits directly within your backend. 

Learn more about order management, and listen to a call with a voice assistant.

Billing & Payment

Voice assistants and chatbots can use conversational AI to guide customers through billing and payments in the same way a live agent would. These bots can take payments, check refund statuses, balances and more, all while identifying smart and natural upselling opportunities.

Learn more about automated payments over the phone, and listen to a call with a voice assistant.

Personalized Service 

Voice assistants and chatbots can integrate with your technology stack to access customer data to deliver personalized customer experiences. They can identify returning callers and greet the customer by name, recall previous orders and preferences, and identify opportunities to up-sell additional products or services based on the customers’ needs. 

Learn more about personalized customer service over the phone, and listen to a call with a voice assistant.

Data & Insights

Voice assistants and chatbots automatically record data on each conversation that can be used for meaningful insights. The data will show why customers are getting in touch, and reveal how operations can be optimized for maximum efficiency and best-in-class customer satisfaction. 

Learn more about real-time conversational data

5. Industries

Conversational AI Industries

Conversational AI can be used to deliver enhanced customer service across a broad range of industries. 

Any company with a high volume of repetitive customer queries can benefit from implementing conversational AI – and their customers will benefit too with faster access to support and shorter wait times to talk to a live agent.

Conversational AI for Banks

Some banks have already implemented conversational AI with great results. Conversational AI  in banking is often referred to as “conversational banking” – an overarching term that applies to both text-based and voice-based assistants. 

Voice assistants and chatbots offer an affordable way to deliver always-on customer service, and are a great way for forward-thinking, digital-first banks to disrupt legacy banks.

Banking Use Cases

In the banking industry, voice assistants and chatbots can be used for:

  • Intent capture & intelligent call routing
  • Account queries including balance inquiries, changing account details, opening accounts and password resets
  • Money transfers
  • Personalized service based on customer history
  • Intelligent sales & upselling based on customer data and needs 
  • Authenticating customers through natural conversation
  • FAQs such as branch opening times or specific product questions
  • Reporting lost or stolen cards
  • Reminders
  • PIN resets

Learn more about conversational AI for banking.

Conversational AI for Telco

For telcos, robust and speedy customer service is essential. The pressure to offer customers what they want, when they want is a crucial competitive differentiator. 

Voice assistants and chatbots are becoming an integral part of the telco customer experience. Instead of long wait times often associated with calling these companies, customers can speak to digital assistants immediately to troubleshoot issues, ask question, change account details, or even book engineer appointments. 

Conversational AI for Telco - Example of troubleshooting with a PolyAI Voice Assistant

Troubleshooting with conversational AI

Telco Use Cases

In the telco industry, conversational AI can be used for:

  • Intent capture & intelligent call routing
  • Authenticating customers through natural conversation
  • Account services including activating plans, paying bills, bill queries, updating details, checking data usage and more
  • Personalized service based on customer history
  • Intelligent sales & upselling based on customer data and needs 
  • Troubleshooting for technical issues including network and coverage, routers, modems and devices
  • Answering FAQs such as roaming charges or information on upgrading plans
  • Booking and managing engineer appointments, including confirming appointments

Learn more about conversational AI for telecommunications companies, or watch a demo from one of PolyAI’s live deployments with a major telco provider in the UK.

Conversational AI for Hotels

Conversational AI takes the pressure off hotel staff allowing them to focus on the customers in front of them.

As a real-world example, Landry’s Golden Nugget Hotel & Casinos deployed a PolyAI voice assistant to solve staffing issues and deliver better CX. 

Learn more about how Golden Nugget were able to automate 87% of calls from day one, and other ways they’re planning to use the voice assistant in the future.

Hotel Use Cases

In the hotel industry, conversational AI can be used for:

  • Intent capture & intelligent call routing to departments, specific services or individuals
  • Bookings for hotel stays, restaurants, spas or other services 
  • Automated concierge services 
  • Answering FAQs with up-to-date information 
  • Personalized service based on customer history
  • Intelligent sales & upselling based on customer data and needs 

Learn more about conversational AI for hotels.

Conversational AI for Restaurants

Restaurants typically miss 30% of phone calls, including a large proportion of reservations and orders. These missed calls are often the result of understaffing.

Voice assistants can answer every call immediately, ensuring no missed opportunity for revenue and freeing up restaurant staff to focus on the guests and tasks at hand.

Restaurant Use Cases

In restaurants, conversational AI can be used for:

  • Taking and managing bookings, including cancellations and outgoing confirmations
  • Taking orders over the phone
  • Recording full call transcripts in case of issues with orders
  • Handling order queries and complaints
  • Answering FAQs with up-to-date information 
  • Personalized service based on customer history

Restaurant Demos

Conversational AI for Insurance

Insurance companies can use voice assistants or chatbots to deliver fast, always-on customer service at a fraction of the cost of previous call center solutions. Voice assistants allow call center staff to focus on the tasks that require their help the most. They use customer and business data to deliver fast, personalized service, and recommend offers that suit individual customers needs.

Insurance Use Cases

  • Intent capture & intelligent call routing
  • Authenticating customers through natural conversation
  • Claim management such as status updates on current claims
  • Policy management including adding or removing beneficiaries, modifying the policy subject, changing details or renewing a policy
  • Payments and early arrears
  • Personalized service based on customer history
  • Intelligent sales & upselling based on customer data and needs 
  • Answering FAQs such as policy inclusions and premium amounts

Learn more about how conversational AI is transforming the insurance industry.

Conversational AI for Retail

Retail customers expect always-on service and answers to their questions fast, or they will go to a competitor. Chatbots are best suited for simple tasks like answering FAQs or checking order statuses. Voice assistants can offer the next level of customer service, handling more complex use cases and delivering personalized service.

Retail Use Cases

  • Intent capture & intelligent call routing
  • Handling where is my order (WISMO) calls
  • Checking inventory and putting items on hold
  • Answering FAQs including store opening hours, product information and delivery and return information
  • Intelligent sales & upselling
  • Taking payments over the phone

Learn more about conversational AI for retail or listen to a voice assistant answer a WISMO query.

6. Benefits

Benefits of Conversational AI for Customer Service

Customers now expect access to support outside of standard business hours. Conversational AI allows businesses to offer 24/7 support to their customers, handling more complex use cases than ever before and delivering exceptional customer service. 

Voice assistants solve staffing issues in call centers by handling repetitive tasks and automating common queries to take the pressure off live call center agents. 

Benefits of voice-based conversational AI

Answer every single call immediately, 24/7

Voice assistants delight customers by answering their calls 24/7 – without the additional costs incurred by staffing a 24/7 call center. Voice assistants can answer every single call, day or night, resolving queries or taking down information to pass to agents when they’re back in. Customers are able to get answers to their questions when it suits them, instead of only during business hours.

Free up staff to deal with other tasks

Voice assistants automate repetitive tasks over the phone, freeing up staff to deal with more value-adding calls, or to handle other tasks. 

Minimise risk of overstaffing

Overstaffing during peaks is often unavoidable to accommodate for staff absenteeism and unforeseen demand. Voice assistants are able to handle thousands of calls simultaneously, making them a great solution for peak times. During quieter periods they can also sit quietly and wait for calls to come in, without costing a penny.

Consistent & personalized customer service

Voice assistants provide the same high level of customer service for every single call. PolyAI voice assistants only need customers to provide information once and can recall it quickly and accurately throughout the course of a conversation. They can even provide personalised service, identifying previous callers and recalling past conversations to increase efficiency and delight your customers. Voice assistants are able to identify when a customer is vulnerable or asking a question that requires an agent, and hand them off with the information collected so far. 

Voice assistants also provide transparent insight into conversational flows, making it easy for third parties to audit them to ensure conversations are aligned with business policy and brand voice.

Identify issues as they arise to deal with them proactively

Customers pick up the phone when other digital alternatives fail. PolyAI’s voice assistants collect conversational data which can be accessed on live dashboards, so that managers are able to see a rising trend in calls about specific topics. The ability to analyse conversational data in real-time can help organisations react faster to unforeseen risks and issues such as website problems, current affairs, or a rise in specific product enquiries. You can access full call transcripts to understand the issue and react quickly and proactively to it, before it gets out of hand.

PolyAI’s Live Conversational AI Dashboard

Collect detailed conversational data for specialised training purposes

Agents may be unable to take detailed notes during peaks, whereas voice assistants collect notes automatically on every call. This detailed data will show the most common requests and conversations your customers are having, so that call center managers can pinpoint what training needs to be delivered to agents to make them better at responding to customers.

7. Getting Started

Approaches to deploying conversational AI for Voice

Once you’re ready to get started with voice AI, you’ll find that there are a number of different ways to design, build and deploy a voice assistant. Some routes to building voice assistants start cheap, but quickly get expensive, others tend to hit roadblocks that prevent deployment or cause users to flee.

At a glance, the 4 main options for deploying conversational AI for voice are:

1. DIY, from scratch

For the DIY purists, there are open source Github projects such as DeepPavlov, MindMeld, or ChatterBot that will give you a starting point. These algorithms supply some core components for a chatbot, but if you want to handle voice, you will also need processors for automatic speech recognition (ASR), natural language generation (NLG), and text to speech

This allows for purpose-built technology that the company has full control over. But it is an extremely expensive and time consuming process, requiring a full team and often over a year to deploy, making it accessible only to very large companies and very slow to show any ROI.

2. DIY, with a third party platform

A more practical DIY approach is to build off the voice platforms developed by the tech giants. Google DialogFlow and Amazon Lex are the two most popular. These platforms handle the underlying natural language processing and speech algorithms for you. However, this comes at the cost of control. If the ASR or NLG do not work well for your project or in the language you need, then you have little recourse to fix that roadblock. These platforms also skew toward text-chat based use cases and tend to underperform in voice centered applications. 

3. Porting chatbot technology into voice

Many companies begin their conversational AI journey with a chatbot. It is a reasonable place to start; chatbots can be deployed on established channels, and they can help to deflect call volume. However, chatbots are built for text, not voice, so converting a chatbot into a voicebot yields mixed results, often failing at stages such as speech recognition and intent detection.

4. Partner with a voice-first vendor (such as PolyAI)

Partnering with a vendor who specializes in voice, using speech technology specifically developed for spoken interactions, is the best way to reduce risk when deploying a voice assistant for your business. A specialized vendor can deploy a highly accurate voice assistant in as little as two weeks, with very fast ROI.

For more information, read the blog post on how to deploy conversational AI for voice. 

8. PolyAI

PolyAI and Conversational AI

PolyAI’s Multilingual Conversational AI

70% of customers feel more loyal to brands that offer support in their native language. However, most customer service automation solutions have limited options for language support.

PolyAI has developed some of the best multilingual technology in the conversational AI industry. Our voice assistants can be ported into any language in under two weeks, without the need for additional training or dialogue design.

Our voice assistants will follow the same behaviour across all languages, giving customers around the world a consistent experience. Or customise prompts in each language to provide a true reflection of your brand.

In the video below, PolyAI co-founder Eddy Su calls up a live PolyAI voice assistant and speaks to it in Spanish, Mandarin Chinese and English (and back to Mandarin Chinese). The voice assistant identifies each language spoken and responds in the same language. 

 

Getting started with PolyAI

Talk to us about how PolyAI can help your company launch new customer experiences at scale, improving loyalty and retention, reducing call center costs and proving ROI within months. We’ll give you a live demo of our voice assistant, personalized to your industry.

In an initial meeting with you, we might discuss: 

  1. How voice automation fits into your customer service program
  2. A comparison of the different platforms available for building a voice bot
  3. How to build a successful voice bot with minimal training data
  4. How to capture your brand’s identity in the voice channel
  5. How to build a voice bot that understands a variety of accents
  6. How to port your voice bot into all the languages spoken by your customers