Today, we want to share a story about how we accelerated and simplified hiring by implementing an HR chatbot. It provides information about the company, job vacancies, collects resumes, and offers test tasks. All of this is done automatically with a seamless transition to a human operator if needed.
Our Pain Points
Previously, we spent too much time sorting applicants' resumes, communicating with them, copy-pasting, and sending test tasks to each one, as well as answering the same questions repeatedly. The step-by-step hiring process looked like this:
These eight steps with a hundred applicants might not seem overly complicated, but they consume many hours of manual work. We would gladly dedicate this time to our core activities, which unfortunately can't be paused during the hiring process. So, we've long wanted to automate the procedure.
Our team has been around since 2017. For nearly six years, we've been hiring people, and last year, we established our own chatbot development team. The team has already helped many of our clients automate their business processes, and finally, it was time for us to eliminate our manual work.
As a result, a new member joined our team — the HR-bot. We based it on our existing chatbot platform: a system for building automated dialogue systems on RASA, Botfront for the visual interface, and the Chatwoot service for operators, so a live person could intercept the dialogue.
This is what the HR-bot looks like for us:
And this is how it looks for the user:
The HR-bot can not only collect resumes from applicants but also answer questions, provide information about job vacancies, and the company as a whole. To do this, we compiled FAQ scenarios: we divided the pool of potential questions by topics, wrote answers to them, and trained the bot to recognize what the user wants to know.
At any stage of the conversation, one can access a menu to initiate another scenario. For example, to inquire about job vacancies, submit a resume, or a completed test task. Most importantly, if the bot doesn't understand what's being asked of it, it offers to bring in a human. The applicant can also request to speak with a human at any time, and the bot will transfer the conversation to the team.
During the communication process with the candidate, both the bot and the operator can take turns participating. The operator can hand the dialogue back to the bot, and the bot can again transfer it to the operator, depending on the scenario or at the request of the applicant.
With the introduction of the HR-bot, the hiring process has become much more comprehensible, faster, and painless. It now looks something like this:
1. We describe and post a job vacancy, asking applicants to respond using the HR-bot.
2. The candidate completes the scenario, sends their resume and test assignment, which then appears in the Chatwoot dialogue.
3. Depending on the position, you can designate a responsible team member who will communicate with the candidate in Chatwoot and review the test assignment.
The reviewer in the Chatwoot admin panel can respond to the candidate in Telegram and create notes accessible only within the company.
4. After screening the resumes and checking the test assignments, each participant in the process sees the candidates in a convenient list with contacts. They can be filtered by statuses: 'Awaiting first interview', 'Awaiting second interview', 'Considering', 'Rejected'.
In this way, we have virtually no chance of missing someone, failing to provide feedback, or not scheduling an interview. Moreover, the Chatwoot lists help us to re-select from candidates we liked but who didn't get an offer. If a new position opens up, we can reach out to them again.
After the introduction of the HR-bot, we were finally able to focus on other important tasks. Before, processing 100 candidates took us about 25 hours of pure time. 15 minutes for each candidate: send a test, answer questions, accept the test, send feedback. Now, the chat-bot does all this work.
If you have questions, please ask them in the comments. And if you want one for yourself, we'd be happy to help build it on our platform.