MRV uses Speech-to-Text and expands accessibility for its audience
About MRV
MRV operates in the real estate and civil construction market as a platform focusing on housing solutions for the general population. The company’s goal is to reduce Brazil’s housing shortage. Currently, it has projects in over 160 cities in 22 Brazilian states.
Tell us your challenge. We're here to help.
Contact usBy deploying the tool for its virtual customer service, the company gained quality and speed in transcribing voice messages, thus democratizing the conversation with its target audience.
Results
- Service time reduced by a third due to increased productivity
- Voice-message transcription quality improved by 20%
- Increased accessibility has transformed customer experience
Ability to handle 9,000 voice-message interactions every month.
According to Brazil’s Geography and Statistics Institute (IBGE), civil construction has been on a tear recently, rising 2.1% from the final quarter of 2020. Grupo MRV&CO, one of the main players in this market, is based in Belo Horizonte and has 19,000 direct and indirect employees, and projects in over 160 Brazilian cities.
Grupo MRV&CO is positioned as a housing solutions platform including MRV and four other companies: Sensia, Luggo, AHS, and Urba. In recent years, the organization has been diversifying its product portfolio with the main goal of providing housing solutions tailored to every moment in its customers’ lives. The company’s experience allows it to plan the building process, keep its services constantly updated and stand out in the civil construction market, particularly for the cost-benefit it offers customers. MRV also provides purchasing facilities, takes part in housing programs and is partnered with real estate loan providers.
To continue having a positive impact, the brand invests in maintaining a good relationship with its strategic audiences, such as customers, prospects and service providers. In 2017, with an eye on trends and always looking to innovate, the company was already opting for tools to make customer contact easier, such as virtual assistant Maria Rosa - which was replaced by a new persona in 2020: Mia.
Using a messaging app, the virtual assistant talks to people in an automated, intelligent way and helps with negotiations, answers questions, discusses ongoing projects, or delivers quality customer service.
However, the company noticed that some of its customers chose to send voice messages instead of texts, either by preference or due to difficulties writing or typing. In a country such as Brazil, where the functional illiteracy rate hovers at around 29%, according to the latest data published by the country’s National Education Plan in 2018, creating a new way to interact with Mia via voice messages was not just about giving another option to consumers, but also a way to make Mia more accessible and build a bridge to inclusion.
Fully understanding those voice messages was essential for the project’s success, hence the need for an efficient service with low latency to reduce the time between questions and answers and maintain the quality of the service.
With that in mind, the company assessed the options available in the market and used a statistical method to choose the most suitable tool. Before starting the process, MRV set criteria regarding performance, latency, costs and ease of use. The company chose Google Cloud’s Speech-to-Text solution.
“We assessed the main players in the market and Google Cloud was way ahead because of Speech-to-Text, both in conversion time and transcription quality. In other words, we confirmed it would be a better, faster alternative.”
—Daniel Martinez, Cognitive Intelligence Core Coordinator, MRVThe road to deploying Speech-to-Text
The teams’ preparedness was essential for the project to run quickly, and the partnership with Google Cloud streamlined the process, with weekly schedules available to answer questions and discuss best practices. The deployment brought consistent results, reduced risks, and the possibility to have technical subsidies for occasional issues.
Today, the solution uses an orchestrator handling about 9,000 voice-message interactions every month and services those who wish to speak to Mia. This orchestrator must understand the request and forward it to one of the four expert bots or a human operator, depending on the complexity of the situation.
When the orchestrator identifies a message as a voice message, it sends it to the Speech-to-Text service to obtain the transcription, interpret the message, and forward it to the expert bots.
The assessment to incorporate Speech-to-Text was made in February 2021; by April, thanks to an agile structure, the tool was already in production, despite MRV’s complex bot architecture, with four bots using the same telephone number in the messaging app to provide the service.
“What we gained by making Mia accessible in a channel like this one means a lot to the company. One of our cornerstones is making customer-centric decisions. For us, an outstanding customer experience is far more important. It’s what we’re committed to.”
—Daniel Martinez, Cognitive Intelligence Core Coordinator, MRV“Imagine being able to read someone’s voice message explaining that they don’t know how to write or saying that ‘my cell phone is broken, I cannot write, only send voice messages.’ It’s very impressive,” Martinez says.
The coordinator explains that, to a certain extent, situations like these raise awareness in the teams, as the information comes from an audience that previously did not have a place to be heard.
Benefits have also been felt inside the company. From April to June 2021, service time was reduced by a third and the quality of audio transcriptions jumped 20%. Customer service employees can now get a full transcript, increasing efficiency in the team’s daily work.
There are still challenges at this stage. One possibility being assessed is providing a differentiated option so that MRV’s customers can choose Mia’s answers. This would demand technical solutions, including the decoding of graphic elements and images. Nevertheless, this plan could be implemented soon.
“We also want to expand channels using Google Home and Google Assistant. Regarding the use of AI and Vertex AI, we intend to explore the conversation analytics features to extract insights from our large number of interactions. There is a lot of potential ahead.”
—Daniel Martinez, Cognitive Intelligence Core Coordinator, MRVTell us your challenge. We're here to help.
Contact usAbout MRV
MRV operates in the real estate and civil construction market as a platform focusing on housing solutions for the general population. The company’s goal is to reduce Brazil’s housing shortage. Currently, it has projects in over 160 cities in 22 Brazilian states.