Chatbots and Conversational AI have gained immense popularity in the last few years to enable human-to-computer interaction. The technologies use natural language processing (NLP), providing an easy conduit for people to interact with computers via text or voice.
According to Google Trends, search volume around “chatbots” remains consistently high as businesses realize the value that conversational technology can provide to improve user experience. A study by Outgrow, an interactive content technology provider, predicts that 80% of businesses will integrate some form of chatbot system by 2021.
Chatbots deploy in a broad range of industries: e-commerce, banking, real estate, human resources, travel, and more. They serve various business functions but are most commonly used for sales, customer support, and information gathering.
Despite their advantages as a conversational interface, chatbots are not the be-all/end-all solution to human-to-computer process automation — far from it. They are, in fact, just the tip of a much larger, more intelligent iceberg.
If your company is standing on the tip of an iceberg and satisfied with your current chatbot use, read no further. However, if you have outgrown your current bot and long for something to support persistent, multi-person or system automations, read on.
What is a chatbot, and how does it work?
Chatbot, short for chatterbot, is a computer program that can interact conversationally with a person, either through voice or text.
There are two types of chatbots: FAQ or single task bots and Digital Assistants. Most bots deployed in enterprises are FAQ bots. They make it easier to find answers to frequently asked questions that may reside in disparate systems. You type in a query, and the bot sends the appropriate response (hopefully). However, as soon as one bot launches, enterprises realize they need another bot to answer additional questions from a different system. Before long, there are hundreds of bots to support all of the various queries and workflows inside of an individual business unit or service area. Once the maintenance overhead becomes visible, IT leaders often realize several bot platforms are deployed in the enterprise, and none of them speak the same dialect.
Digital assistants (DA), such as Alexa, Siri, Cortana, or Google Assistant, are the second type of bot. These are more sophisticated and use NLU, NLP, and ML to learn and evolve to increase personalization levels as they gather and process information.
As smart as these digital assistants seem; however, they often provide interactions that are far from anything resembling a conversation with a real-life human assistant. When asking a DA like Siri a complex query that requires context, the system often fails to interpret the intent, resulting in an incorrect response.
A 2019 digital personal assistant study by digital consultancy Perficient posed 4,999 questions to the most popular voice assistants and found that every personal assistant — from Google Assistant to Siri to Cortana — had dropped in accuracy from the previous year.
“This indicates that current technologies may be reaching their peak capabilities,” the study said. “The next big uptick will likely require a new generation of algorithms.”
What are the advantages and disadvantages of using chatbots?
Chatbots are capable of working in a variety of areas. However, customer service is where they excel. According to Grand View Research, 45% of end-users prefer them as the primary communication mode for questions regarding customer service.
Great Uses for Chatbots:
- Chatbots are excellent at handling simple FAQs, enabling faster responses;
- They are an inexpensive option compared to hiring human agents;
- They reduce call volumes and keep customers happy.
Customers see chatbots as a useful tool for:
- Getting a quick answer to questions;
- Resolving a complaint or problem;
- Getting a more detailed answer to a query;
- Identifying a human customer service agent.
Although chatbots today are becoming increasingly more intuitive, they still have limitations — and plenty of them.
Chatbot Issues and Shortcomings:
You will outgrow your chatbot
Chatbots solve simple problems, but companies wanting to extend their use across the enterprise will find them woefully lacking.
They can’t answer complex questions
Complex questions that need serious analysis are typically too difficult for chatbots. If a bot attempts to answer questions around several topics or a use case that is too broad, it will hardly provide a satisfactory user experience.
They are challenging to integrate into multiple systems
Chatbots that require integration into two or more systems to provide a single answer require significant software development effort. If you find bot projects are following traditional SDLC cycles, you will later realize your automation practices are unsustainable since they need so many development hours.
Chatbots are transient and stateless
Chatbots don’t remember previous conversations, so they cannot pick up where the conversation left off. If a workflow requires input from a person not immediately available, then the conversation stalls and must reinitiate when all required parties are present.
Chatbots lack the necessary security features an enterprise will require
In today’s age of data sensitivity and privacy, customers and enterprise security officers must trust the bots containing private data to comply with laws and mandates. Chatbots also lack auditing features required to meet compliance mandates. If there is ever an issue, you have to ask your IT development and operations departments to review terabytes of log data.
How Chatbots and Krista Differ
Now that we briefly covered what chatbots are, how they function, and the advantages and limitations of their use, let’s clarify how they differ from Krista.
Chatbots are just the tip of the conversational process technology iceberg. Put another way; they are like what a wheel is to a car. Wheels are great, but you need three more and an underlying chassis, engine, dashboards, and safety features.
Krista may be the most advanced chatbot you’ve ever seen. That’s because it’s not a chatbot at all. It is an intelligent automation platform that uses the chatbot interface to communicate with people and systems in a human-like, conversational manner. Krista can speak to computer systems and listen for events to then notify humans when their expertise is needed.
Don’t get us wrong. We’re not throwing the chatbot baby out with the bathwater. Krista embraces and extends the chatbot concept. However, in our world, a chatbot is a “channel,” or “medium”, just like a telephone or computer browser. Krista works with a variety of mediums to help people interact with systems. In addition to chatbots, Krista communicates via Slack, email, a web-based client, mobile apps, text messaging, and other mediums your people are accustomed to.
It is not the interface at issue, however, but what Krista’s underlying technology enables it to accomplish.
Our goal from the outset was to create technology that understands people, not make people understand technology. We say if you can text, you can use Krista. Literally, if your employees can text, they can interact with numerous systems to support customers, consume enterprise services, deploy IT changes, or update important KPIs.
Chatbots Fit for the Enterprise
Enterprise technology is complex and full of constraints. Technology and systems often span multiple decades, most of which weren’t designed to work together. Technology advancements like service-oriented architectures, APIs, cloud, and machine learning make it easier for us to interact with these systems, but enterprises still need help integrating people and technology. With that, I’d like to provide a brief guide on what to look for in a chatbot fit for large enterprises.
Provide a simple authoring environment
Enterprises need to provide business users with an intuitive self-service authoring environment to build automation and conversational artificial intelligence. Building business workflows shouldn’t require extensive software development skills.
Enable integrations into multiple backend systems
Enterprise workflows will require information from various systems, not just one. Therefore, make sure your chatbot can integrate directly via a standardized integration model to supporting systems for processes that span more than one system of record.
Enforce role-based access
When integrating multiple systems in a workflow, your automation must adhere to enterprise security standards and provide data based on your appropriate level of access. Some chatbots have access to PII, and you must be able to enforce data security and compliance.
Use a stateful chatbot
A chatbot fit for the enterprise understands state and can enable persistent messages. It knows what you talked about last time. State is significant for messages that involve more than one person.
Enable multi-person conversations
Chatbots should be capable of multi-person conversations. Conversations involving multiple people are crucial for approvals, milestones, and checkpoints for processes requiring a human to verify the process or override a rule.
Enterprise chatbots should support rich-based text and dashboards
Chatbots are great at looking up information but lack visualization. If you want to enable self-service for your business users, they will want to visualize data based on their specific queries. Use a chatbot that supports data visualizations.
Enterprise chatbots should be able to start conversations with systems and finish with humans
If you genuinely want to integrate your people, process, and technology, use an enterprise chatbot that can listen for event triggers or schedule future events in the context of the conversation your people are used to.
What Krista Can Do That Chatbots Can’t
Like chatbots, Krista enables anyone to build and create workflows around business process constraints following the way humans already communicate. Unlike chatbots, Krista offers a standardized extensible authoring environment and an enterprise integration model. This means you don’t need to have technical development skills to integrate one or more systems into the same bot interface. If you want to interact with one person and one system, chatbots are great. If you’re going to interact with more than one system or more than one person, use Krista.
Krista can do things that chatbots aren’t equipped or designed to:
Benefits to Using Krista
Increase Staff Effectiveness and Efficiency
Employees converse with Krista, similar to texting, instead of logging in disparate supporting IT systems. (e.g., CRM, HR, Expense Mgmt., Accounting, Supply Chain, DevOps, and many others.)
No More Expensive, Ongoing Staff Training
Eliminate training and retraining costs from turnover or when you acquire or update maintenance-intensive IT applications. If your employees understand how to text, they already know how to converse with Krista.
Digital Transformation
Krista achieves C-level goal attainment and provides a 100% mobile experience for customers, partners, and employees. Krista involves stakeholders in creating their own human-to-system-to-human conversations, therefore enabling sustainable digital transformation.
Eliminate IT Backlog
Eliminate wait times since IT no longer needs to build custom apps or dashboards across various platforms. Business owners utilize Krista conversations to update and access business KPIs.
Extend Your Chatbot with Intelligent Automation
On their own, chatbots are sufficient to answer simple questions, solve easy problems, and perform basic transactions. But their ability to scale to resolve more complex issues, integrate with backend systems, or handle multi-user conversations leaves enterprises wanting more.
However, chatbots coupled with the power of intelligent automation is a win-win combination. Implementing your chatbot with a flexible automation platform will expand its core abilities and extend its utility and lower overall cost of ownership. To extend your chatbot, please contact our team to set up an evaluation or reach out to me directly.
Read our comprehensive LLM testing guide
See how several LLMs fared in our test and steps to run your own tests for each of your use cases.
