McQueen Autocorp is a leading US-based used car selling company with a buyer network spread across the entire United States. McQueen Autocorp offers a nationwide modern car selling service, facilitates sellers with an instant offer for their car(s) using price prediction algorithms and real-time bids from an extensive network of licensed buyers.
Disclaimer - The name McQueen Autocorp is a placeholder as there is an NDA signed between both parties.
The client operates a cloud call center that handles about 5000 - 8000 calls a day, providing quotes for used cars. About 150 agents take these calls throughout the day from various locations across the world.
During certain peak days, like the financial year-end in March or peak hours during the day, such as Noon EST, when the traffic from the PST time zone also comes in, there would usually be an influx of calls agents may not have the capacity to handle.
As a result, many calls are kept in long queues (sometimes even 10 mins) until a representative is assigned. Due to these long queues, most callers would hang up the phone or go to a competitor, resulting in a loss of potential revenue for the client.
The client required an automated way to handle this influx of calls without having to scale their cloud call center team, as this was usually a spike for a few hours and not consistent throughout the day.
Given the human limitations of scaling man-hours at the time of need and the additional costs, the team at Maruti Techlabs proposed an AI-based chatbot solution.
With this, an incoming call that was kept on queue for more than 2 minutes would begin play- ing an IVR, asking the caller whether they would like to get the quote for their car via an SMS chatbot.
If the caller selected YES, the call would end, and an SMS would be sent to them outlining the steps to start the process.
To adhere to the strict regulations in the US, which prevents companies from bulk sending SMSes, the user was required to explicitly grant permission to start the quoting process by having to send START.
Once the permission is granted by replying with START, the chatbot would begin asking a series of questions required for the system to generate a quote.
‘What is the make of your car?’
‘How many miles does the car have?’
During this whole process, due to the open nature of input boxes and human nature, not every visitor would input the required information in the required format. Some visitors would ask counter questions or respond to questions with answers that didn't match the format of the question or something completely different.
The chatbot was trained to recognize the following scenarios -
Dynamic Responses : If you look closely at the image, the visitor has responded with its make and model as Jeep Grand Cherokee. Obviously, Jeep Grand Cherokee is an SUV, and hence the chatbot has used the term SUV while requesting the zip code.
Model Name Correction : In this scenario, the visitor has entered the model with incorrect spelling. Sometimes, a manufacturer may have two models with similar names, causing an inaccurate price quote.
To prevent this, we leveraged Natural Language Processing (NLP) so that the bot is capable of detecting misspelled words and suggesting the right word to confirm the next step.
Smart Suggestion : If you do not know the model of your car, you’re not alone; there are thousands of models out there, and sometimes we may not remember the model of the vehicle that we’d like to sell.
If the model name is not identified or provided by the visitor, the AI chatbot smartly suggests a list of possible model names. The visitor can now choose the model and move ahead in the flow to generate their car quote.
Synonyms/Slang Detection : With most users being from the US, local slang was a massive issue for call center representatives who weren't from the US or not familiar with the culture. Based on this, Maruti included slang detection to understand the words most commonly used and map them with their original names.
In this case, Chevy is Chevrolet.
Global Context : As the users were asked for their car’s make and model, the bot would identify if the user provided just make and not model, or both collectively. This helped in improvising on the next set of questions by not asking them repetitively.
As the user entered ‘An Oldsmobile Aurora’, the chatbot detected the Make=Oldsmobile and Model=Aurora and moved to the next question, i.e., zip code.
We follow Agile, Lean, & DevOps best practices to create a superior prototype that brings your users’ ideas to fruition through collaboration & rapid execution. Our top priority is quick reaction time & accessibility.
We really want to be your extended team, so apart from the regular meetings, you can be sure that each of our team members is one phone call, email, or message away.