Enterprise Conversational AI versus Chatbots
Demand for enterprise conversational AI solutions has increased rapidly over the past 18 months but what is conversational AI vs chatbots? The terminology alone is confusing, not to mention the different chatbot and conversational AI tools and platforms that support the building and management of these natural language solutions.
Impact of COVID-19 and Artificial Intelligence (AI)
The impact of the pandemic forced businesses to urgently seek digital solutions that would tackle the sudden challenge they faced in interacting with customers and employees in virtual rather than physical worlds. Voice assistants and text or messaging interfaces backed by AI have opened up new opportunities for more automated digital engagement models that came into their own during the pandemic.
Virtual assistants powered by AI and natural language understanding NLU that could be built and launched in a matter of days and weeks became highly attractive business propositions as companies struggled to manage remote working models without impacting the customer experience.
As conversational AI technology and maturity levels have advanced, more strategic approaches to growing and managing these solutions are needed. Growth and sophistication have also revealed some challenges that hadn’t surfaced with simpler bot solutions.
An Enterprise Conversational AI Platform has become an extremely attractive proposition to overcome these challenges and enable greater agility and cost-efficiency in maintaining and managing multiple conversational AI projects.
Conversational AI vs Chatbots
Chatbots and conversational AI are two terms that are often used interchangeably. Most people can visualize what a chatbot is whereas conversational AI sounds more technical or complicated.
“A chatbot or virtual assistant is a form of a robot that understands human language and can respond to it, using either voice or text.”
The term conversational AI gained popularity as a means of distinguishing more advanced chatbot implementations from simple rule based chatbots. Early chatbots focused mainly on simple question-and-answer-type scenarios that the natural language processing (NLP) engines could support but stopped short of executing anything more complex, often handing off to human agents to continue processing the request. In doing so, they gained a poor enough reputation that has lingered in some user perceptions.
Conversational AI, on the other hand, combines natural language processing (NLP) with other technologies (e.g. machine learning, secure integration with data, process workflows, etc.) so that the modern chatbot can now understand more complex multi-turn requests, integrate securely with business data, automate underlying workflows in a journey, and hand over to a human only when necessary.
The Tools for Building Chatbots and Conversational AI Solutions
There are multiple developer tools and platforms that support the building of chatbot and conversational AI solutions. Here are some examples:
Natural Language Processing NLP Development Tools
NLP development tools generally provide a set of natural language processing (NLP/NLU) tools to enable bot developers to create AI conversational experiences. The biggest and most common NLP engines are Google DialogFlow, Amazon Lex, IBM Watson, Microsoft Luis, Rasa, and Facebook’s WIT.ai. Although these tools can be a good starting point for a business to build its first chatbot, they are developer tools so rely on skilled bot builders, conversational designers, and sometimes even data scientists to create the chatbot solutions and AI models that address the business need.
Building more advanced conversational AI experiences by just leveraging the NLP tools gets very complex very fast as they are fully reliant on building intricate conversational flows to support the use case. For example, Rasa’s open-source framework developed the notion of stories to structure conversation flows for creating more enhanced experiences while DialogFlow CX takes a state machine approach to designing bots. While both of these tools and approaches make conversations easier to design, enhance, and maintain, the AI bot solution is dependent on managing the flow purely using conversation. Process flows are not integrated into these solutions. So imagine building a chatbot for a complex use case or journey that has many different turns and tasks. These solutions become unwieldy to build, difficult to operationalize, and costly to maintain.
Low Code Development Platforms
For business analysts or other business users, these bot development tools are less suitable. Hence some of these users have turned more toward low-code bot development tools that sit on top of these NLP engines (or sometimes leverage their own proprietary NLU) and make the bot building exercise more intuitive and easier. These have been instrumental in enabling business users to spin up their own chatbots with little need for IT involvement. Some of these low-code platforms come with some out-of-the-box connectors to common ticketing or CRM systems and can support deployment to either single or multiple digital channels, such as Facebook Messenger, Amazon Alexa, WhatsApp, or web. Usually, they come with pre-built templates that tend to be very specific to domain-specific solutions such as for customer service, or IT helpdesk use cases.
The advantage of low-code platforms is that they enable the business to get chatbots launched more easily but often this means circumventing the IT department which can lead to different departments using different vendor solutions, spawning a diverse NLP and vendor landscape of bots that are not governed by the IT organization. Managing and scaling these diverse chatbot solutions can become problematic and costly over time. Organizations that have reached a higher maturity level with their bot deployments tend to turn to enterprise conversational AI platforms to fulfill their requirements.
Enterprise Conversational AI Platforms
An enterprise conversational AI platform is more suited to the rigors of enterprise IT requirements while still offering low-code bot-building capabilities to business users. The technology stack that these platforms support is much more typical of what enterprise architects look for to ensure good governance, control, and flexibility and is in line with how they make technology investments. However, this enterprise approach is often less suited to a small business that just wants to deploy a single digital assistant with few integrations or plans for extra use cases or capabilities.
Both low-code and enterprise conversational AI platforms leverage NLP technologies without the business users needing to be in the weeds of how that all works under the hood. In many cases, these platforms also offer workflow design tools where bot builders can integrate both conversational and process flows into chatbot solutions. These are often easier to build and manage than similar solutions created using one of the NLP developer tools.
Key Elements of an Enterprise Conversational AI Platform
Several vendors coming from the enterprise technology space saw the need to provide an enterprise-grade platform that satisfies the needs of business users in building their own conversational AI solutions that reflect their goals while also meeting the needs for IT in areas like security, scaling, and management of more complex or diverse environments. The core elements of these platforms are:
1.Build Chatbots for Multiple Use Cases
Unlike point solution providers that focus on a specific use case (such as sales, marketing, customer service, IT helpdesk, etc.) or a particular industry sector (insurance, healthcare, banking, contact center, etc.), enterprise conversational AI platforms support the building of conversational bot solutions no matter what the use case is. If a business invests in the technology they can build all their chatbot solutions using this single platform, with a simpler chatbot architecture.
This allows for creating more consistent bot experiences for customers and employees, a more standardized approach to creating, training, and managing all chatbots, and more efficient and cost-effective management of multiple bots in production. There is no need for an organization to use a different chatbot platform to build customer service bots than for building an internal IT helpdesk, quotation, collections, or other use case bots. Yet many have done this, partly because of the initial allure of subject matter expertise in the form of intent libraries and other domain knowledge. But is this a sufficient trade-off as the organization adopts the technology across other parts of the business?
2. Support all Roles in Bot Building, Training & Management
These platforms fulfill the needs of business users as well as IT groups. Business analysts, conversational designers, or general business users get the low-code tools that give them the freedom to build their own chatbot projects, even with the ability to use available APIs to connect to common 3rd party systems running in the enterprise (e.g. live chat, CRM, content systems, ITSM, or HR systems, etc.). These platforms can even make it easy for users like customer service agents or product managers to train and update the bot based on conversational history and missed intents that are tracked and monitored.
Enterprise developers can use other platform tools or the command-line interface (CLI) to build more complex bot solutions while IT gets visibility and control over security, APIs, management, and deployment. Even enterprise architect needs are met as some of these platforms offer architectures that support multi-bot or diverse NLP environments. These platforms are designed to demystify the underlying AI which means that the need for data scientists to continuously tune AI models is no longer necessary. However, the platform provides the tools for data scientists to optimize training datasets to achieve optimum outcomes.
3. Bot Security, Governance, and Control
Security and control over software technologies and applications that run in a business are essential requirements for any enterprise IT organization. It is no different when it comes to chatbot deployments. Enterprise-class security features ensure that customer data is always protected, never aggregated, always isolated, and under the control of IT. It supports the need of organizations that are regulated in meeting compliance standards and supporting business continuity plans. Protecting sensitive customer or employee data in conversations and connecting securely to data sources is of paramount importance. An enterprise platform takes a holistic approach to security across the complete lifecycle, touching on all aspects of this such as physical, application, data, and network security.
For an enterprise, the above requirements are table-stakes but without a platform these need to be handled on a case-by-case basis, making it hard to govern and control at an individual bot level.
4. Bot Lifecycle Management
As a business creates and deploys multiple chatbot projects, bot lifecycle management can become more challenging as this has to be done at each individual bot level. Managing a growing set of intents and utterances, tracking conversation history, and managing missed or overlapping intents gets trickier the more complex a bot experience gets and as bot projects multiply.
Bot Lifecycle Management not only includes the smooth deployment of a bot from staging to production but extends to the complete lifecycle of managing updates, versioning, and rollbacks.
5. Bot Orchestration
Advanced organizations that are growing the capabilities of their chatbot experience or that have multiple bots in production, potentially running on different NLP engines, are often faced with challenges around scaling and management. An initial chatbot solution is usually designed with a limited scope. Over time its capabilities may be expanded to handle things like the addition of more products or services, increased language capabilities, different target customer bases, and/or extending functionality to encompass other use cases.
For most NLP engines, there is a point at which the addition of more intents causes the experience to degrade or the bot to completely fail. This can be overcome by unifying multiple bots that act like subject-matter experts into a single experience. This is a very efficient and scalable way to architect advanced solutions but it then requires the concept of a bot orchestrator or conversation manager that is at the forefront of the conversation and can route to the right bot to act on the intent.
The vendor landscape for chatbot and conversational AI technology is diverse and complex. Chatbot building tools require the skills of developers and data scientists to create solutions and AI models that are truly enterprise-grade. These can be difficult and costly to maintain and manage over time. Then there are point solutions that focus on a very specific use case (customer service) or industry (banking, travel, etc.). These may limit an organization that wants to have a centralized approach for all use cases across the business, both customer- and employee-facing. For businesses that want to more easily scale, monitor, and manage their AI bot solutions and reduce the skills and cost requirements to do so, an enterprise conversational AI platform provides the structure and tools to help them.
During the pandemic, chatbots and conversational AI came of age. Now as they mature, enterprises are seeking more advanced capabilities, architectures, and governance structures from vendor solutions. The argument is less about chatbots vs conversational AI solutions and more about how enterprises can approach the building and management of these solutions more effectively, securing their investments, lowering costs, and accelerating speed to market.
Take a moment to download our eBook: A conversational AI Journey Guide to learn more about the steps to get started and advance your chatbots and conversational AI maturity.