There is a tremendous amount of buzz and excitement about chatbots and their potential application to the HR function. It's exciting to see the long-term potential of chatbots used for initial applicant screening, scheduling interviews and distributing employee engagement surveys. If you are going to use chatbots, it's important to understand how they work and how to set them up correctly.
How Do Chatbots Work?
Natural language processing (NLP) provides the human element to chatbots. There has been much progress in this field. Here are highlights of core NLP tasks:
- Tokenization. Breaks up a sentence into individual words and punctuation.
- Part-of-speech. Categorizes the words into typically 36 different types of tags (e.g., adverb).
- Chunking. Studies the words and puts them together into the appropriate phrases (e.g., noun phrase, verb phrase).
- Name entity recognition/classification. Identifies the word as a person, location, data, etc.
Once these tasks are completed, there are several online data libraries that allow the program to determine synonyms for the phrase used. For instance, an NLP library would know that "hi," "good morning," "good afternoon," "what's up" and "hello" all equate to a greeting.
Onboarding Your Chatbot
Now that we know NLP libraries enable a program to "read" text and understand its basic meaning, the next step is to train the system to interact in a dialogue. Every chatbot must be trained on what constitutes a question and what constitutes an answer. For example, if you are building an HR benefits chatbot, you could enter into the system:
Q: Tell me about our benefits plans.
A: We have three different plans: a CDHP, PPO and HMO.
Using NLP, the system would be able to understand different versions of the question and then connect to the database to find the right question and the right answer.
5 Tips to Enable Chatbot Effectiveness
Given how chatbots can be deployed in HR organizations, key stakeholders (managers, employees and applicants) should be immediately involved. This means that regardless of how small the test, it is important to make sure you are doing everything possible to make it successful. Here are five steps to consider to increase the probability of an effective chatbot implementation.
1. Provide extensive data and training. Chatbots gained popularity in customer service and help-desk environments because they had a tremendous amount of existing data showing permutations of questions and answers. However, in most HR organizations, such large volumes of data don't exist. A team should be available to help initially populate the database with common questions and answers.
2. Allocate ongoing support. There will always be scenarios when the chatbot gets stumped by a question. It may be due to how the question was worded, or because the chatbot doesn't know the answer. These systems all have mechanisms to capture null answers, and someone on your team will need to review these regularly to provide the right answers. This process will set up the chatbot for success so that the next time it receives that same question, the chatbot will know how to answer.
3. Define and map to your real-world process. Not only do chatbots answer questions, but many will take a user through a process—such as an initial applicant screening, open enrollment, scheduling interviews and distributing employee engagement surveys. However, there may be situations in which you won't want to enable a chatbot's functionality. For instance, scheduling phone interviews via a chatbot may be time-saving, but many firms will use a coordinator to schedule interviews because of the need, for instance, to book an interview room.
4. Think about geographic and language needs. Many of the chatbots operate only in English, so consider how this could impact your user base and your scale across the organization. Does this impact how you train or deploy the chatbot?
5. Create the chatbot's persona. In building the chatbot, you will need to provide a name, icon and tone. Do you want your chatbot to be formal, casual, funny? This is often an exercise where you can engage your branding team to develop a chatbot persona that best reflects the company's brand.
Start Playing Now
Chatbots have the potential to provide tremendous savings, make processes more efficient and increase customer satisfaction. However, your chatbot will not be able to answer all questions on its first day, and it will continually be learning the vocabulary nuances of your organization. Set expectations with senior leaders about the chatbot training process so your CEO isn't concerned with its initial lack of functionality. There is so much potential in the HR function for chatbots, and practitioners should be testing these tools to help improve a company's performance.
Franz Gilbert is vice president, Solution Provider Programs at Bersin™, Deloitte Consulting LLP.
As used in this document, "Deloitte" means Deloitte Consulting LLP, a subsidiary of Deloitte LLP. Please see www.deloitte.com/us/about for a detailed description of our legal structure. Certain services may not be available to attest clients under the rules and regulations of public accounting. Copyright © 2018 Deloitte Development LLC. All rights reserved