Answering routine questions, assisting contact center agents to handle calls, creating support tickets, checking statuses—conversational AI technologies have transformed all of these contact center tasks. And by doing so, this advanced call center technology has undoubtedly enabled enterprises to create a range of differentiated user experiences and increase productivity in specific workflows.
Whether they’re called chatbots, virtual assistants, or digital assistants, it’s clear that for margin-squeezed and increasingly commoditized industries, any reduction in contact center call volume and human-administered support greatly benefits a business’ bottom line. This is especially true for financial services and, in particular, wealth, insurance, and retail banking—
Stuck in the silos of POCs
In 2019, a survey by Forrester, 69% of global data and analytics decision-makers whose firms were adopting automation said they implemented or planned to implement chatbots in the next 12 months.
Yet most businesses struggle to scale past proof-of-concept (POC) or pilot deployments and reach the promised land of reduced contacts and optimized costs.
Whiles businesses and their contact centers face many challenges to scale chatbots, the most-cited hurdles are:
- Inadequate contact volume
- Chatbots only programmed for information delivery
- Substantial effort to maintain and update chatbots
- Insufficient Return on Investment (ROI)
And without scaling the use of chatbots, it’s impossible to accurately measure the total impact they have on a company’s bottom line, This is because the true success of any enterprise-wide deployment of chatbots can only be assesses by tracking metrics such as
- Adoption
- User satisfaction
- Use case scope
However, the sad reality is that once the initial enthusiasm of a chatbot launch dies down, the initiative loses momentum. Firms often find that it’s hard to scale the deployment to cater to the diversity of users and products their business supports. Some contact centers even end up with hundreds of staff working with ad-hoc tools, like spreadsheets and application scripts. All of these practices defeat the purpose of having a chatbot in the first place.
Why are we stuck in this mess?
A key reason for not progressing past POCs is the fact that most conversational AI platforms today are based on the machine learning/deep learning rabbit hole, in which algorithms only learn from examples. This process is also known as inductive learning.
To build the AI bot, subject-matter experts (SMEs) spend weeks collaborating with engineers to create intents, which train the bot to understand verbal requests or patterns of behavior and convert them into actions, also known as intents.
The integrity of these intents will determine the success or failure of the bot. If the bot can’t recognize the intent behind the query, it’s useless.
But creating intents is often a manual and time-consuming process and managing their deployment and evolution without compromising the overall accuracy of the system and quality of service becomes a tall task. Such a system is, by design, not scalable.
The Key to Scalability: Knowledge
CogniCor takes a radically different approach that is built around an area that is often overlooked in contact center software—knowledge.
Businesses generate thousands of digital assets daily, ranging from user manuals and website FAQs to policy documents and videos. These assets contain routine procedural knowledge, process and product information, policy details, etc.
Combined with the data available in a CRM and data generated by other linked systems, it is easy to see that every enterprise sits on top of a knowledge stockpile that is overlooked and underused. That’s where knowledge graphs come into play.
Knowledge graphs represent vast amounts of interrelated concepts and information in a graphical format. They encode concepts, topics, and features around the topics that are relevant to an organization’s operations.
For example, in financial services, a product or service is often connected to a complementary product or service. When the client wants to take a mortgage, they may need insurance as well. Or when the customer takes a mandatory distribution from an IRA, they may consider alternative investment accounts. Similarly, you might have the appropriate content to help a customer resolve an issue about a particular account’s login, however that information might be the same for all your other accounts and you may want to reuse that in your information architecture without duplicating it.
Knowledge graphs model the real world in an elegant, scalable way. Applying machine learning technology on top of knowledge graphs results in intelligence that is enriched by the information represented in the knowledge graph.
Doing more, with a leaner team
Instead of weeks, what if intents could be created in minutes? Instead of a large team with hundreds of engineers, testers, content writers, and business owners, what if a lean team of a handful of knowledge engineers can manage an entire conversational AI deployment?
With a knowledge-graph-driven platform that uses deductive AI capabilities, this is entirely possible.
With knowledge maps, conversational AI platforms can directly tap into the expertise embedded in documents and extract intents, attach content to relevant areas in the knowledge graph and use pre-trained domain libraries to accelerate deployment time.
With a no-code interface that manages the end-to-end process, including version control to archival and analytics, a lean team of mostly knowledge engineers can manage the deployments.
Business information becomes completely accessible
Instead of depending on SMEs, the AI can learn directly from the business’ pre-existing assets. In short, the AI becomes the SME, reducing the time and resources required to build the bot.
Business knowledge resides in various formats, ranging from PDFs to text documents, to JPEGs, to videos. When contact center agents need to find specific information within a business, it can be like searching for the proverbial needle in a haystack.
Once companies can harness and manage their institutional knowledge in a knowledge graph, both staff and customers can get answers to their questions instantly. And, when used by live contact center agents, they can respond to the customer’s needs faster than ever before. And this faster service leads to a long list of other benefits: shorter handle times, lower per-interaction costs, higher agent productivity, increased first contact resolution, and let’s not forget higher CSAT
Standardized and personalized information across all channels and touchpoints
Traditional knowledge management platforms are a bit like managed chaos. They have dozens, if not hundreds of different authors—each contributing different content. Some of this information is outdated. Some of it is contradictory. Some of it is duplicated.
To add even more confusion, each author has a different writing style. Instead of being stored in one place, content is distributed everywhere—across brochures, websites, and videos. When a customer or contact center agent finds information, it is very likely that someone asking the same question will hear an entirely different answer.
But if business knowledge is managed by the digital assistant, customers and employees can get consistent responses across all touchpoints—whether it’s from a digital assistant on the company intranet, Facebook messenger, or Alexa.
Coupled with advanced call center technologies such as cloud contact center that offers skill-based agent routing, this lends itself to a far more efficient and intelligent contact center that goes beyond today’s static IVR call trees.
A Step Towards Cognitive Enterprise and Digital Companions
With conversational AI systems grounded in knowledge graphs, companies now have the staggering ability to reuse existing enterprise knowledge. They can also create new knowledge in the digital assistant, based on demand. Regardless of which method they use, digital assistants can eventually become the brain trust of the enterprise, acting as a single point of truth.
From the user’s perspective, going beyond scripted chatbots, knowledge-driven conversational AI has the potential to evolve into true digital companions of their users, not only in assisting them with information and tasks but also in thinking and collaborating with them.
Reimagining conversational AI as a knowledge management platform is thus the stepping stone towards becoming a truly cognitive enterprise.
CogniCor’s Digital Companion integrates with CXone to understand your customer’s questions and find information faster for your support staff. Deploy in self-care portals, in front of Live Chat, and on Agent dashboards to reduce calls, agent handle times, lower per-interaction costs, and increase customer satisfaction.
