How a bot is helping us solve a very human problem.
“Any sufficiently advanced technology is indistinguishable from magic” —Arthur C Clarke
Five decades ago, when Weizenbaum from MIT introduced the world to ELIZA, or introduced ELIZA to the world, we experienced magic. Never before did anyone even think that something as lifeless as a computer screen can engage us in conversation. Yet, it did.
When we set about creating a bot for our HR organisation, our mandate was something along the same lines - a 24-hour support line that can talk to our people about their issues, questions, whatever they needed help with. If you can talk about it to someone in HR, you should be able to talk to this bot about it. Simple.
And thus, our very own HR bot was born.
Someone mentioned — but what’s so special about a bot. Nothing really. The bot is not what’s special. What the bot does, is.
Treebo HR team v/s The onslaught of routine HR queries
As it happens, we found that the HR team’s woes were entirely man-made!
- Most employees were not aware of where policies can be found and how to interpret them.
- Employees reached out personally to multiple people, multiple times in the HR team to have their queries addressed on basic people policies.
- Responses were frequently lost or delayed leading to poor employee experience.
Basically, because we were handling the routine HR queries manually, it was leading to a lot of complications and roadblocks.
Enabling technology for next generation HR query resolution
The bot would operate via Slack and will cover queries on the following:
- HR policies (e.g: how many days of leave can I take?)
- FAQs on HR processes (e.g: how do I initiate the PMS process? how can I share upward feedback?)
- Commonly understood practices (e.g: I needed an address proof from HR, what are the next steps?)
It’s important that people can use free form text rather than a syntactical one (that works on slash commands of Slack). Make it as human as possible and is syntax-free.
The team did a quick comparison between wit.ai & dialogflow (api.ai), and found that the latter one suited the solution more. (Having said that, for further enhancing the bot, we will be exploring luis.ai as an option too.)
The solution had two major components:
i) The Dialogflow Agent: Where we configured Entities and Intents for the initial dataset. We also integrated a “Fulfilment” feature, where in the Dialogflow agent also talked to our custom “Referral Service Endpoint” for status of referrals.
ii) The Bot itself: Which would convey all DMs to the Dialogflow Agent and get back the response from the latter. With the integration of Fulfilment feature, we made this bot contextual as well — i.e return something meaningful from a third party service based on a query. This brings in a large array of features which we can support in future with this bot — referral status being a small ask!
During the next few hours, we trained the bot on given dataset and we could achieve almost 60–70% accuracy if the relevant Entities were present in the question phrase.
The perks of having a HR bot in our ranks
A HR bot is a rarity and even more so, in a growing startup like Treebo. For us, it ensures all routine HR related queries are handled efficiently. This reduces the dependency on HR team, who had multiple people coming up to them to inquire about basic HR policies and processes. As a result, we have upped our “employee delight”quotient as well as registered a bump up in the productivity levels of HR team.
To top it all, this bot can be further trained to answer more and more queries about HR and people related questions. In fact, in time, its services can be extended to departments other than HR!
The rockstars behind this product are Satish Reddy, Rohan Saxena, and Raghavendra Reddy. Special thanks to Sonali, Thasim and Avnish for contributing to this post!
I hope this post was able to give you an insight into how technology is playing a vital role in the hospitality industry. Keep following this blog for more such posts and technology overview of Treebo.