Table of Contents
Chatbots are largely ineffective for today’s needs
Types of Chatbots on the Market
Framework for Selecting a Chatbot
Chatbots are dramatically innovating the world of business communications across a variety of industries. As customers demand more speedy, efficient means for problem resolution, chatbots play a vital role in improving business processes.
As an industry, conversational AI is projected to reach $1.3 billion by 2025. This is with as many as 90 percent of all queries expected to be answered by chatbots in 2024. The Juniper Research study suggests that this could ultimately save as much as $0.70 per customer interaction.
Whether an organization is dealing with thousands of students at an institution or millions of community members, that cost savings ultimately makes chatbots a no-brainer for improving organizational efficiency. Yet the technologies that power chatbots have not kept up with the growing needs of most institutions.
The vast majority of chatbots on the market are too clunky and are more of a nuisance than they are helpful in the customer experience. In this guide, we’ll look at why many chatbots on the market fall short, and what buyers need to seek instead.
Chatbots are largely ineffective for today’s needs
If you’ve used a chatbot in the past, chances are that some of those experiences have deterred (rather than enhanced) your customer experience. While chatbots have come a long way toward improvement in recent years, there are several reasons why customers are frustrated by the current process.
Chatbots that get stuck
Nobody wants to talk to an AI chatbot only to learn that the bot doesn’t understand them, prompting users to restate their questions, becoming unresponsive, or routing the conversation to a human. A bot that doesn’t work is more frustrating than one that isn’t available. It’s misleading and only creates a more frustrating experience.
One of the biggest challenges is that when many bots get stumped, they often revert back to the start of the workflow, requiring the user to repeat their entries. This is particularly frustrating when conversations are passed to a live human and user entries are not stored. In this case, the user must start all over again with the customer service representative.
Most chatbots don’t remember a conversation when a customer asks a new question as little as a day later. In fact, it might not even remember the conversation five minutes later if the bot closes a conversation with the user. When the problem is more complex or requires some level of follow-up, this can quickly annoy the user if they need to backtrack and explain the same problem multiple times.
Sometimes chatbots can be a bit too eager to chat. It pops up on every page and keeps asking how they can help. Of course, this is neither helpful nor warranted. Another problem is that the bot simply gets in the way of the information you’re trying to navigate, rather than going away when it’s no longer needed.
Gaps in user intent
Many chatbots have a difficult time interpreting the meaning behind a question. For example, the question “How much does this cost?” is ambiguous for many chatbots. The user may be inquiring about total cost, or they may want to hone their out-of-pocket expenses. A bot wouldn’t necessarily make that distinction unless the user was more specific, and many chatbots lack the nuance and contextual understanding to overcome this.
Types of Chatbots on the Market
There are a number of chatbots on the market aimed at solving business problems with varying complexity. Some chatbots work very well for simple tasks like ordering food or e-commerce. Others are built for more complex tasks and require more customization. This section analyzes the overall market of the chatbot landscape, in addition to the pros and cons of each type available.
Pros: Point-and-click bots are fairly straightforward and easy to set up. They are created using decision-tree hierarchies where the information is presented to the user in the form of buttons. A point-and-click bot is most useful when you’re only looking to help the user accomplish a finite number of very simple tasks, such as speaking to a customer service rep or finding a specific program or product. They’re also simple to build and can be done with a little-to-no technical background.
Cons: The biggest disadvantage to a point-and-click bot is that users can’t converse with the bot itself. They’re stuck with selecting from a menu of options that might not otherwise apply to their circumstances. Another problem is that point-and-click bots ultimately require a lot of human intervention, so unless you have sufficient staff in place, they might not be efficient enough for your needs.
Pros: If you know the types of questions your customers are most likely to ask, a rules-based chatbot (also known as linguistic) might be the answer for you. Most rule-based bots are designed using a series of ‘if-then’ statements to match various questions and answers together. It’s a popular option because the implementation is simpler, making it a lower-cost alternative. This is an excellent choice for institutions seeking to test the waters with a bot that can answer frequently asked questions. Rules-based chatbots work well when user requests can be anticipated and streamlined.
Cons: Rules-based chatbots have a difficult time answering questions that are not arranged in strict accordance with their programming. Likewise, these chatbots require a high degree of specificity and fine-tuning in order to function at a high level. The interactions can at times feel robotic because they have a limited understanding of the user’s input and require manual intervention to improve.
03. Keyword recognition-based
Pros: Keyword-driven chatbots are trained to understand the gist of what the user is saying by learning specific keywords meant to trigger a certain response. They are similar to rules-based bots but have an additional degree of sophistication because they have some natural language understanding. They give users the ability to ask more advanced questions, providing a more natural, conversational feel.
Cons: The biggest area where keyword recognition-based bots fall short is that they struggle with questions that are similar, yet require some nuance. As a result, you need staff that can help tweak the bot’s lexicon so it won’t get stumped. Some chatbots use a combination of keywords and buttons to bridge the gap from whatever the bot can’t comprehend to a menu of options.
Pros: Think of this kind of chatbot as a completely blank slate. It will answer as many questions as you want and has some element of machine learning where the bot constantly improves. The advantage is that a custom-designed bot gives administrators full control of the content, and can converse at a relatively high level. It also allows for the ability to add additional content so the bot’s brain can be as comprehensive as you wish.
Cons: Building such a bot is tedious, requiring subject matter experts to contribute large quantities of information to the bot’s knowledge. It also requires constant maintenance to continue improving the brain’s capacity, which makes it a costly and timely endeavor. Many custom-built bots are also limited to point-and-click functionality, which means they lack natural language understanding.
05. Machine learning
Pros: Machine learning bots are by far the most advanced chatbots yet. They take preexisting training data from your own content and turn it into questions and answers on their own. That training data is also layered with existing knowledge that the bot pulls from other bots in order to strengthen its own knowledge. Machine learning chatbots provide a superb user experience and enable the users to ask the most natural and advanced questions.
Cons: Given the high level of sophistication involved, machine learning bots require a more substantial investment to get started. Many machine learning bots on the market come with templates designated for a specific industry, so they can’t be customized to a company’s individual needs. This can create a clunky user experience when the organization wants to provide a more personalized customer service.
Framework for Selecting a Chatbot
Selecting a chatbot in 2022 can feel daunting with so many different types on the market. When it comes to finding the right fit, your decision committee could look at a variety of factors including price, effectiveness, and even reputation. To this point, many chatbots have become commoditized, making it difficult to discern best-of-breed chatbots from the rest of the pack.
As chatbot technology advances, it’s imperative to adjust expectations. Organizations must adopt a comprehensive framework that better integrates chatbots into their overall business operations.
The following criteria represent some key areas businesses should analyze when evaluating their chatbot investment.
Time to Launch
Seasonality impacts businesses across the country from higher education to local governments and everything in between. What that means is that every business decision is impacted based on the timing of when a new solution can be implemented. Yet, buying cycles rarely align perfectly with the implementation time.
For example, let’s say a primary care provider wants to implement a chatbot right in time for flu season but makes the purchase in July. That doesn’t provide a lot of time to get a bot right-sized for your business. On top of implementing the bot itself, it needs time to train itself on answering questions relevant to your business. This often comes in the form of templates or questionnaires that you must complete, and when your time is limited, that isn’t always an option.
The process for implementation can be different depending on your selected technology. If the vendor is using templates, there is a relatively short time to launch since the bot’s brain is already matched to your specific industry.
However, when there’s customization involved, vendors tend to ask hundreds or even thousands of questions in order to nail down how your bot should respond in various situations. Not only is this extremely time-consuming for your team, but it takes an immense amount of time to program.
In this primary care example, the bot might not be built until well after flu season, which means the bot will either be implemented at a sub-optimal time or have a long-delayed launch until it can get tested in an appropriate environment. It’s no wonder that given these challenges, many businesses opt to continue with the status quo, rather than advocate for change.
Fortunately, new chatbot technologies are emerging that dramatically reduce the time to launch, even with a high level of customization and sophistication. These bots take all of your existing content and use it to generate relevant questions that lead to the answers your content provides. This creates a more seamless process and allows your bot to be built in less than an hour.
Customize to Business Needs
The overall efficacy of a chatbot ultimately depends on its ability to work within your business needs. Any chatbot within your consideration stage must be custom-built to understand topics unique to each department, rather than taking a one-size-fits-all approach via templated content. This means it should meet specific objectives and allow all departments to engage in a conversation through a seamless, unified experience. As businesses look to differentiate themselves with a chatbot, it’s crucial to get past canned libraries and answers that don’t enable a unique, competitive edge.
Another example would be the ability to communicate with customers on any platform. Most chatbot consumers who are inexperienced with the space think about bot capabilities as the pop-up that won’t go away on your browser.
The truth is that this perception couldn’t be further from the truth. Chatbots must be accessible to customers regardless of platform, including text, email, social media platforms, mobile applications and even at-home devices like Amazon Alexa.
Think about the diverse range of customers you attract. When a customer has a complaint or needs an answer to a question, many opt to direct message the brand on Twitter, rather than call a customer service hotline. At the same time, that brand must account for other social media channels such as Snapchat and TikTok. This is because teenagers don’t communicate via Twitter while customers aged 45+ are more likely to opt for email.
Integrating these channels together can be a significant challenge. However, many enterprise solutions offer the ability for live agents to provide chat support over web and text. And, if no live agents are available, the system must route the user’s question to a designated inbox so that it can be answered directly.
Speaking of customer engagement, chatbots must provide a level of personalization that incorporates the ability to detect a user’s emotions and offer a conversational feel. They should also respond appropriately when a customer mentions a named entity (a department or name of something specific to that organization). For example, if a user says they’d like information on a specific person’s meeting availability, the bot should know what those hours are and relay them back to the user. If a bot fails to perform this task, it shouldn’t be considered a viable option in your chatbot search, as the inevitable result is another inbound inquiry to your institution.
Finally, there is an aspect of branding here that is worthy of consideration. According to Forbes, sharing a consistent brand presentation across platforms increases revenue up to 23 percent. Put another way, if you are truly looking to integrate a chatbot into your current workflow, and its presentation is disjointed from the rest of your branding, what impression does that leave with your users? It shows them that a chatbot isn’t really part of your company and just another tool that you’re using to improve efficiency.
But when you make the bot a part of your brand by changing the logo, the colors, the avatar, and even the bot’s personality, it starts to feel more like a part of your company and begins to represent your brand. While branding isn’t the most critical aspect of a chatbot, it shouldn’t be overlooked, especially if your brand identity is closely protected.
Meets Accessibility Requirements
Companies are facing a moral and legal imperative to make their customer experience more inclusive. As more organizations seek ways to widen their customer base, more marketers place priority on making their websites more accessible. However, this often requires an overhaul that most companies aren’t prepared for - especially when implementing a chatbot. The most effective way to get ahead of the curve is to simply source a chatbot solution that meets accessibility requirements. Some state-of-the-art solutions meet, or even exceed accessibility requirements out of the box. If you’re unaware of the latest accessibility requirements, there are a few questions you should ask your vendor of choice when investing in a chatbot.
First, get familiar with Web Content Accessibility Guidelines. These guidelines are useful for anyone dealing with development and have a mandate to create better accessibility. WCAG changes often, so when you’re investing in a chatbot, it must be nimble in allowing you to continue improving website accessibility. For example, the chatbot you’re vetting should allow a screen reader to read the text from a chatbot for the hearing impaired. In addition, the chatbot should support high-contrast ratios, enable keyboard navigation, and enable users to increase text size without loss of functionality.
There should also be some straightforward instructions on where users can find accessibility features, settings and support within the application. Meeting accessibility requirements requires plenty of forethought when creating a product of any kind, let alone, a chatbot. But if you’re a business that prides itself on being customer-centric, your audience will hold your feet to the fire if you don’t walk the talk. Finding a chatbot that is WCAG-friendly will go a long way toward building and maintaining trust with your audience.
In fact, some chatbot technologies have the ability to automatically detect accessibility gaps on the webpages on which they’re installed. These solutions leverage connections with WCAG databases that refresh in real-time, ensuring that institutions receive real-time updates related to any on-page compliance issues.
When investing in a chatbot, you want technology that crawls your content organically. It keeps your bot’s brain up-to-date with the most recent content so your business never needs to worry about the bot’s knowledge becoming stale. The solution must at least monitor the institution’s website for updated information and alert administrators to discrepancies between web content and chatbot knowledge, or ideally adjust chatbot knowledge accordingly.
Any AI chatbot must get more intelligent with time by receiving input from administrative users who wish to add chatbot knowledge.
This is done by retraining the AI algorithm on regular intervals to improve the chatbot’s ability to understand natural language questions. You’ll also want your administrators to have the tools they need to edit or add responses as needed without assistance from a vendor.
This can be especially relevant if your business is in a constantly changing industry where information rapidly changes. While other bots often require paying for customization, having one that crawls your site automatically prevents the added cost of owning a bot that many often provide.
Ivy.ai developed new technology to meet many of today’s requirements, and much more. By leveraging self-building bot technology, Ivy makes it possible to build pre-trained, conversational chatbots that know your brand's content in minutes and stays up-to-date with your evolving web content.
The days of building templated chatbots that require maintenance are over. Ivy delivers a custom bot that learns all of your unique content in a matter of minutes. With preloaded training data, your bot will be ready to go on day one. Ivy automatically generates questions and answers in mere moments, building bot knowledge directly from your website, knowledge base and other documents.