Apples to apples comparison of enterprise technology is difficult, and it is no exception for AI solutions in Customer Service.
We have compiled a framework of the most important questions you should be asking about Agatha (and other providers) so you can make apples to apples comparisons:
To answer this question, we need to first clarify the range of AI solutions offered for customer service. We see solutions broadly falling into customer service automation and customer service augmentation.
Customer service automation (also known as ticket deflection)
Customer service augmentation
Agatha Answers is a tool for customer service augmentation and increases agent productivity. Agatha lets your agents tap into all of your corporate knowledge within their helpdesk--no more opening five screens to find an answer. Agatha recommends answers to agents, which is especially helpful when the question is complex or the agent is new. And Agatha surfaces best responses from previously-answered questions so your agents learn from each other.
Agatha is powered by Natural Language Understanding (NLU), which is a subcategory of Natural Language Processing (NLP), which is a subcategory of Artificial Intelligence (AI).
Regardless of whether you’re in the market for a chatbot or an answer recommendation engine (see above), a best in class AI tool for answering questions should be able to interpret the underlying intent of a question, without requiring triggering keywords. NLU is the technology that is able to interpret the “intent” of a question without keyword matching.
An example is how Google search has evolved over the years from keyword search to intent based. Years before, if someone in San Francisco wanted to know the weather, they had to search something similar to “weather San Francisco”
Now, they should be able to search “how hot is it outside” and Google will understand the intent of the question even though you never specified “weather” as a keyword.
Data is the fuel that powers AI tools, and we have noticed big differences in the type of data various AI solutions for customer service can leverage. The three main types of knowledge data are:
Centralized knowledge: Help articles, cloud docs, and macro templates that companies have manually created as “sources of truth” are by far the most common type of data powering AI solutions for customer service. While it is a great best practice to have centrally organized knowledge, without proper processes to refresh it, it risks becoming outdated after a new product launch or organization change.
Organic knowledge: Past customer service tickets and email/slack conversations contain collective knowledge that is organically generated by your organization every day. The knowledge is more real-time and relevant, but are more cumbersome to sift through.
Specially-created knowledge: Some tools require you to specially create knowledge in order to power their AI tools. For example, the solution might require you to transfer all your enterprise knowledge onto their platform or it might require you to create question-answer templates (this is common in chatbots). When you are required to create templates, it starts becoming a macro instead of a true AI solution.
A best in class AI tool should be able to tap into the entirety of your corporate knowledge, both centralized and organic, without requiring specially-created knowledge. You should choose a tool based on the type of knowledge your organization can support. We have noticed that amongst our customers, smaller companies without the resources for pre-created knowledge can benefit from Agatha for surfacing recommendations based on the company’s past tickets/conversations. In contrast, larger companies want both answer recommendations powered by centralized knowledge (to surface relevant macros and help articles) and organic knowledge (to see best-in-class ticket responses).
Agatha recommends answers to a wide range of questions from common to unique. Agatha's prediction for macros/canned responses is over 85% accuracy. Agatha's prediction for ticket tags and categorization is also over 85% accuracy. In general, we see Customer Service agents using Agatha on over 40% of incoming cases.
In evaluating any enterprise solution, you have to compare the upfront cost of implementation and the cost of ongoing maintenance.
Time to implementation:
Knowledge preparation (days): Agatha can index your entire corporate knowledge in a few days.
Train model (days): Agatha can be trained in days and ready to start as early as next week.
Employee learning curve (none): Agatha is integrated into your helpdesk tool to minimize the learning curve for your employees.
Because Agatha is directly connected to your knowledge sources, Agatha automatically updates as your information grows.