In simple terms, Conversational AI is a technology that uses machine learning to to talk with users in a way that feels natural and personalized. Conversational AI is a combination of three technologies – Artificial Intelligence, Messaging and Speech Recognition.

Conversational AI is designed to help software understand and interact with people in the most intuitive way possible – using natural language. It enables businesses to deliver automated and personalized communication experiences using voice assistants, chatbots and messaging apps.

With the advent of social media and messaging mediums, enterprises are on the lookout for solutions that help them in providing delightful and personalized customer experiences that can also reduce the cost of maintaining Customer Service teams. Conversational AI holds the key in achieving these objectives. It can change every aspect of how, when and where enterprises engage with people. Conversations can be short on-off requests or can be part of longer-running customer engagements. Conversational AI empowers enterprises to deliver intelligent, superior and customized Customer Experience.

Here are some of the reasons why enterprises should consider implementing Conversational AI into their business.

• Time Saved – Conversational AI provides quick responses and thus faster customer service. Customers get answers to their questions instantly and in case of a complex case, are transferred to the suitable customer service agent.
• 24×7 Availability – Conversational AI is always present to serve customers, be it day or night.
• Ease of communication – Customers can access Conversational AI from the channel of their choice. Also, the conversations are maintained in sync across channels to maintain conversation continuity,
• Increased Efficiency – Conversational AI provides operational and customer service efficiency. As Conversational AI handles most of the common customer queries, only few complex queries are forwarded to the customer support team, thus increasing their efficiency. With this automation in place, enterprises can concentrate on other business functions.
• Consistent Branding – As Conversational AI remains consistent in all its interactions with the customers, the enterprise’s image and branding are maintained and are always consistent.
• Delightful CX and High CSAT– Conversational AI can provide highly personalized interactions by identifying individual users, their profiles and previous conversations. This is extremely helpful for interacting with not only existing customers, but also potential customers.
• Reduced Costs – Conversational AI solutions considerably reduce the customer service costs as the technology has the capacity to handle multiple queries at once and is also capable of solving most of the queries itself without any human intervention.

Conversational AI is powered by the following technologies:

1. Natural Language Understanding – for intent identification and entity extraction
2. Automated Speech Recognition (ASR)
3. Sentiment Analysis
4. Reinforcement learning for state prediction

Combination of the above technologies is used as per the use-case. For example, ASR is used for a voice based virtual assistant, which is not required on a chat channel.

Conversational AI can be deployed to a wide range of channels. Conversational AI can be configured to stay consistent across all channels and the conversations are synchronized across channels. Some of the common channels include –

• Mobile App
• Facebook Messenger
• WhatsApp
• Skype
• Slack
• Website page

… and more

Conversational AI leverages the power of classification, prediction and entity extraction models that are developed using advanced Machine Learning techniques. A user input is transcribed, understood and converted into an action based on the predefined workflows.

Conversational AI is suitable for a wide range of use-cases. But these are the most prominent ones

• Customer Support – Conversational AI can answer customer FAQs and common queries. In case of a complex customer query, it can gracefully transfer the customer interaction to a human agent. Conversational AI thus reduces wait times, increases employee efficiency and provides quick and delightful customer experience.
• Sales – Conversational AI can act as a sales executive by providing the suitable products to the customer according to their requirements. Conversational AI can also help in effective lead capture. It receives  important information from the potential customer and depending on their level of interest, it captures the lead and forwards it to the appropriate human agent.
• Data Aggregation – Conversational AI is capable of collecting information from users in an interactive conversational manner, which is otherwise a tedious and time consuming process.
• Employee/Agent Assistant – Conversational AI can be employed to assist employees in their day to day activities. Covering a wide range of areas like HR, Finance, Performance, Learning and more, Conversational AI can act as a personal assistant to employees. For Agents, Conversational AI can assist them during calls and meetings, giving them suitable information and insights in real-time.

Yes, it is possible. However, many of the companies fail because of the following reasons:

• Lack of right talent
• Lack of awareness on the  development process as most are first-timers
• Lack of awareness about  the right set of tools

Conversational AI is still evolving and it might take another 3 to 5 years to stabilize.. Till we get there, it is recommended to take external help from domain experts.

Conversational AI solutions vary in complexity. A simple Q&A bot (1st generation) might cost in the range of 10-30K USD. A more complex (3rd generation) bot might cost in the range of $100K to $500K.

In case the solution is implemented internally, the initial learning and experimentation cost could be in the range of $30-50K. However, this doesn’t guarantee project success because of the challenges involved in building the bot internally.

The following skills are required for building a conversational AI solution.

1. Project manager
2. Conversation Solution Architect
3. Conversation designer
4. Language model developer
5. System integration engineer
6. Backend developer
7. UI developer for channel design
8. Conversations testing engineer

Some  organizations tend to ignore the conversation design part which results in too much of to and fro during the implementation phase and the project might end up as a failure because of the poor performance of the bot.

While few of the skill sets like project manager, backend developer might be easy to find within the organization, the majority of the skill sets are difficult to find not only within the organization, even outside. Generally, getting the complete skill set within the organization is a challenge..

Microsoft, Amazon, Google, and IBM have provided services and tools to build conversational AI solutions. These tools and services are states of the art. However, if not used in the right way, it results in overall failure.

For building a simple bot, with the ability to handle some QnAs and few intents, the quantum of customization is low. However, if you are looking at a goal-oriented complex bot with the ability to handle QnAs, multiple intents, seamless context handling ability, and complex integrations, the amount of customization needed increases significantly.

If you are looking to do a POC and if you have access to a good tech team, it is not a bad idea to try in-house. However, have a cap on the number of resources and time, as these kind of experimentation projects might go on forever. The ideal duration for a POC is 1 to 2 months. In-house development has the following advantages:

1. You will get a closer look at the problem you are trying to solve, and the associated complexities.
2. You will get a better idea on the approximate time and cost estimate for the project implementation
3. In-house POCs are lower on cost (in case you are tight on time and resources allocated)

However, be very clear on the expectations of the in-house team. In general, the quality of the output would be low, because of the team’s in-experience and limited information available online.

For a full-scale project, the best way to go about is through an external vendor in collaboration with the in-house team. The advantages of going  with the external vendor are:

1. Experience doing multiple projects
2. Access to the right set of tools

Conversational AI bots are generally not out of the box solutions. They involve integrations with multiple enterprise systems. Involving the in-house IT team as a key stakeholder in the project is very important. The concerns which the IT teams have are about data security and data duplication.  If the IT team works on the POC, they would get a better idea of the challenges and can help the vendor better.

Here are the steps in implementing a Conversational AI solution

• Gathering business requirements & use case development
• Conversation design & workflow development
• Language model development
• Business logic development and system integrations
• Channel setup

While these are the stand steps, there could be variations based on specific requirements part of the use case. For example, there could be an RPA integration in the place of system integration in some use cases. While the overall implementation is a very long and resource intensive process, having a set of pre-built tools (channel setup, system integration frameworks, QA tools etc) and skills (pre-built set of intents and entities), help in speeding up the development process as well as ensuring quality.

While the enterprises have realised conversational AI can address this problem, they often find it challenging to set up and run conversational AI initiatives.

Here are few of the challenges which enterprises face:

1. Where to start?
2. Finding the right set of tools
3. In house development team with right skill set
4. Finding the right vendor
5. Building an enterprise grade bot with ROI
6. Compliance
7. Solution life cycle management challenges
   • Testing
   • Deployment
   • Training
   • Analysis