A Tale of Two Chatbots: Search-Based vs. Guided Interactions

Since the dawn of the internet, we have had only two ways to locate information on websites: a navigation bar and a search bar. It seems astonishing that after 30+ years we would still only have those two main options. But a paradigm shift is on the horizon. Chatbots of various sorts are making waves in the online pool, sending ripples of rewards and rebuke through the private and public sectors alike.

Chatbots are emerging as the long-overdue third option for finding information digitally. But not all chatbot platforms are created equal. Thus, many companies are missing the mark with their chatbot solutions. Instead of creating a third option, they are simply repackaging the search option in a cleaner interface. Yes, the user gets to type their inquiry as if chatting, but the user interaction gets pretty stale after that.

In many ways, search-based chatbots are functioning partly on novelty and less on service-to-completion (sorry, Microsoft).

The reality is that users often don’t know what they are looking for in the first place when interacting with government. A search bot, just like a search bar, has a tough time linking the colloquial text of the citizen with the jargon that government agencies use. It’s like both sides are speaking a different language to each other and everyone is frustrated. Furthermore, search-based bots have little to no value after the user has located the information they are searching for. Users are still forced to read through confusing text while trying to parse out a process that is even more confusing. Where is the value in that?

Now, novelty aside, there is another option available in the chatbot space: guided conversation. Guided conversations favor the user selecting from a list of possible responses to drive the user to information and through complicated processes. Unlike search chatbots, which serve only one function, a guided chatbot can be used in a variety of ways, including RPA, brand and reputation management, automated services, information finder, and much more.

However, it should be noted that guided chatbots can suffer from the same issue as search bots. The same barrier exists in all government implementations of chatbots: How do we communicate effectively with the citizens that will utilize the service? The solution is two-pronged: Code-free chatbot builders and writers.

Code-free chatbot systems (such as Tars) take the programmers out of the picture, and we need programmers out of the picture when actually building a chatbot (sorry again, Microsoft). Code-free bots allow the designers of the system to focus on communication and business processes the chatbots are automating. It turns out that those trained to communicate effectively to various audiences and situations are actually the people you want designing your chatbots. Who knew?

Government’s biggest problem to date is communication.

The technological evolution of communication has steadily outpaced government’s implementation of digital resources. This can be attributed to the lengthy and expensive process of standing up such solutions at times. And let’s be honest, some vendors to government have famously provided little value for excessive costs. And chatbots are no exception (more on this later). So, let’s take a look at the numbers.

I have two award-winning website/chatbot systems for us to compare: Arkansas.gov’s search-based chatbot and Montana REAL ID’s guided chatbot. Which chatbot system offers the most value to the citizen? 

Arkansas.gov

According to the Associated Press, Arkansas.gov won the Mar Com “Gold Award for having exceedingly high standards as a government website,” with a chatbot feature that provides answers to citizen questions. Launched in 2017, the chatbot has been used nearly 4,000 times out of 800,000 site visitors.

Initial impression:

• I actually clicked the wrong widget (a feedback popup using a simple form) initially on page load because the chat widget took a while to load
• The chat interface is mobile-friendly, and persistent (it remembers my chat after I leave)
• It allows me to type out a custom message (I typed: “I need to file a business license. Can I do that online?”) and delivers a response with a link for follow-up action; however, I can also type “help” to view a list of possible topics
• The link dumped me on a Secretary of State page to continue my information seeking on my own (pics below)
• The system asks if the information was helpful by selecting a “Yes” or “No” button
• I selected “No” and the bot gave me an option to leave a message for a representative to answer in 24 hours or reword my question

The final verdict:

All-in-all, this is a glorified search feature. The issue with Arkansas’ search-based chatbot is the help ends before it really ever began. Yes, it gave me a link, but as someone who knows nothing of applying for a business license in Arkansas (is that even what they call it?), I’m still left confused at the end of my inquiry and forced to find the actual answers for myself. I’d like to know how many interactions with the bot have led to the user leaving a message for a representative and how many have been recorded as successful searches.

Mtrealid.gov

Disclaimer: I designed and built this website and chatbot. I will try to be as objective as possible in my assessment; however, feel free to leave a comment with your assessment of both website/chatbot features so we can compare.

Mtrealid.gov is considered the first “conversational” website in government and recently received a national award for “Excellence in Public Engagement” for its part in the overall Montana REAL ID information campaign. The user interface is an amalgamation of web elements and a guided chatbot interface that launches from a widget to specific information. The chatbot then offers additional information and links to other pages/resources and automated services such as the license fee calculator. The website and chatbot were launched Sept. 7, 2018, and the bot has recorded over 5,000 unique users with an interaction rate over 60 percent. Users have also completed specific services at nearly 30 percent (though not all interactions will require a service to be completed).

Initial impressions:

• All information presented in the bot is targeted and curated to supply specific information relevant to the user
• The interface is mobile-friendly, but not persistent (however since the web elements are almost all tied into the chatbot, a persistent interface isn’t needed)
• The user cannot type out their question; however, the current system is capable of offering a search-based feature, but the value is limited
• Service processes are delivered step-by-step and even filter to specific conditions that address different situations of the user; the automated fee calculator uses this feature to calculate license cost based on the specific factors of the user
• Upon completion of a service or an informational chat, the bot leverages the user for a rating and direct feedback; the user can leave their email for a response back if they wish

The final verdict:

Overall, the REAL ID chatbot delivers value from a global to a granular scale, though the subject matter is far more focused, the concept for the REAL ID bot can be scaled up to service a state or even federal website. The system could benefit from the addition of a search-based feature from custom user messages as long as the search led to a guided conversation to complete the service or information request successfully. The REAL ID chatbot also manages ratings and feedback portion more dynamically than Arkansas’ chatbot.

For me, bias or not, the guided chatbot provides a much more dynamic and engaging platform that assists customers in finding information and successfully completing processes. There are benefits to a search-based feature in a chatbot, but the communication gap between citizens and government renders the search inefficient and ineffective. However, this would be a great resource for government employees that understand the government lingo to find information for themselves.

Currently, there are only a few major vendors offering chatbot solutions. All of them to my knowledge produced search-based chatbots except one: govchatbots. This is a branch of Tars that is offering guided chatbots for government including training, support and development options.

What do you think of my comparison of search-based versus guided chatbots? Dropped a comment below or follow me on LinkedIn and comment when I post this article!

Levi Worts is part of the GovLoop Featured Contributor program, where we feature articles by government voices from all across the country (and world!). To see more Featured Contributor posts, click here.