Multiasistencia: Reading vital documentation with machine learning tools

About Multiasistencia

Based in Spain, Multiasistencia handles insurance claims on behalf of major insurance companies, liaising between policy holders, insurers, and repair firms to resolve claims rapidly and efficiently.

Industries: Financial Services
Location: Spain

About About Pervasive Technologies

Pervasive Technologies builds predictive solutions with AI and machine learning, specializing in the intelligent treatment of images and video streaming.

Multiasistencia built two machine-learning-based digital products that automate the reading of vehicle claims forms and invoices from housing insurance claims and convert them to structured data.

Google Cloud Results

  • Opens, reads, processes, and closes 60% of claims automatically in testing
  • Detects fraud by comparing costs with historical information using machine learning
  • Reads handwritten forms and accurately finds information from highly varied document formats
  • Creates new claims in less than a second, compared to five seconds manually

Cuts claim handling costs by 35% and aids fraud control

Anyone who has made a claim on their insurance policy knows how important it is to get a fast response from their insurer. Insurers understand that, but they also need to check every claim for fraud and accuracy. Multiasistencia works to make that process more efficient for everyone involved. Founded in 1983, the company manages insurance claims end-to-end for insurers in its native Spain, as well as Portugal, France, Italy, and Brazil.

"For 35 years, our value proposition has been to provide claims management services that both reduce the cost of claims for insurers, and increase the quality of service for policyholders," says José Antonio Molleda Fernández, the company's General Director of Strategy and Corporate Resources. "We act as a white label brand, so when policyholders calls their insurance company, they speak to one of our staff, and we then orchestrate the whole process until the claims have been paid out."

"By embracing automation, predictive models, machine learning, and big data, we're evolving into a comprehensive technological partner for insurers. We want to transform the way they work. That's our vision. We're not waiting for someone else to disrupt the industry. We want to be a leader."

José Antonio Molleda Fernández, General Director of Strategy and Corporate Resources, Multiasistencia

With the emergence of disruptive new technologies, Multiasistencia made a strategic decision to focus on implementing digital transformation throughout the industry. The team began by optimizing the claims process, identifying ways technology could be used to make tangible improvements for insurers.

"By embracing automation, predictive models, machine learning, and big data, we're evolving into a comprehensive technological partner for insurers," says Jose Antonio. "We want to transform the way they work. That's our vision. We're not waiting for someone else to disrupt the industry. We want to be a leader."

Finding new ways to read data from documents

In many industries, manual data entry remains a crucial step in making information ready for analysis and processing. For Multiasistencia, it's an aspect of every claim handled for its clients. And although it takes time, accuracy is vital for processing sensitive and important information on behalf of policyholders and insurance companies.

"Our insurance claims handlers used to receive documents and invoices describing damages, the necessary repairs to fix them, and their costs," says José Antonio. "They picked out the relevant information, analyzed it, and made a judgement about whether the costs were in keeping with market standards. It was a long, labor-intensive, completely manual process."

By applying machine learning to that process, Multiasistencia looked to collect data from documents more quickly, more accurately, and at a lower cost. And by making more data available, it aimed to simplify the identification of fraudulent or unreasonable claims.

Multiasistencia already used G Suite for communications and, impressed by the machine learning tools on Google Cloud Platform (GCP), approached the Google Cloud team to discuss its objectives. "Google Cloud introduced us to partners that would help us with our project," says José Antonio. "Due to its deep knowledge and competences in the use of AI to read unstructured and handwritten documents, we began working with AI-specialized partner Pervasive Technologies. We saw an opportunity, and because it was on GCP, we knew we could try it out, thanks to the flexible, pay-as-you-go pricing offered by Google Cloud."

The teams set to work on two distinct, but related challenges: vehicle accident claims, which are submitted in handwriting on standardized forms; and housing insurance claims, which are typed, but come in many different formats.

"Together with Google Cloud, we've helped reduce claims handling costs by 35 percent and shorten the duration of claims, helping insurance companies develop more competitive and attractive products for their customers."

José Antonio Molleda Fernández, General Director of Strategy and Corporate Resources, Multiasistencia

"Housing claims and vehicle accident claims demand very different algorithms," says Salvador Marimon, Director at Pervasive Technologies. "For housing claims, there are thousands of different formats, each of which puts the information that we need to find and extract in a different place. So, we don't have a set structure that we can use to train a machine learning model. For auto accidents, on the other hand, there's one standard European Accident Statement form and our algorithms have to read the handwriting used to fill it in."

The teams built a training architecture with App Engine and Cloud Functions as a processing layer. Combining Cloud Machine Learning Engine with TensorFlow, Cloud Vision API, and Cloud TPUs, the solution hosts and applies more than 30 machine learning models in order to read documents, storing images in Cloud Storage and predictions in Cloud Firestore. In tests, the resulting solution processes 60 percent to 65 percent fully automatically, flagging only 35 percent to 40 percent of claims for human attention.

Using Firebase, the team deploys this architecture as Software as a Service products for two main types of claims: housing and vehicle claims. Because they're accessed through APIs, the solutions are easily integrated into client infrastructure. They minimize the workload of claims handlers by flagging unreadable or unusual claims, while improving accuracy by cross referencing information with CRM databases. And with predictive models trained on data from decades of managing insurance claims, the systems are formidable tools for fraud detection.

"We tested Cloud Vision API on handwriting recognition over 12 months and the evolution has been spectacular," says Salvador. "We know that Google uses machine learning algorithms in its core businesses. That gives us confidence that Google Cloud products will make use of the most advanced AI technologies available."

"You don't need 100 percent automation to make major improvements. If algorithms can read key parts of a form, we can reference a database to fill in the rest. Open text fields still need to be manually checked, but automation makes the workflow so much more efficient."

José Antonio Molleda Fernández, General Director of Strategy and Corporate Resources, Multiasistencia

Delivering the incremental changes that truly transform a process

Multiasistencia is already working with one of the three leading insurance companies in France for its housing insurance processes, and with one of the top three insurance companies in Portugal on its vehicle insurance processes. "Together with Google Cloud, we've helped reduce claims handling costs by 35 percent and shorten the duration of claims, helping insurance companies develop more competitive and attractive products for their customers," says José Antonio. "Every insurance company has its own procedures, but they all save in the same ways: faster claims handling, more accurate fraud control, and more efficient claims processing thanks to early identification of problems." Over thousands of claims, incremental efficiencies soon stack up, such as in opening a new claim: a manual task that used to take five seconds, now takes less than one second.

"You don't need 100 percent automation to make major improvements," José Antonio adds. "If algorithms can read key parts of a form, we can reference a database to fill in the rest. Open text fields still need to be manually checked, but automation makes the workflow so much more efficient."

As Multiasistencia readies for deployment in the Spanish and Italian insurance markets, it is investigating new ways to use GCP tools to improve efficiency, from using Cloud Vision API on photographs to automatically assess costs, to opening claims automatically from phone calls with Cloud Natural Language.

"Working with Google Cloud has helped us make things so much easier for our staff, and so much more reliable for our customers," says José Antonio. "And this is just the beginning. It's great to know we can expect more benefits as our partnership with Google Cloud evolves."

About Multiasistencia

Based in Spain, Multiasistencia handles insurance claims on behalf of major insurance companies, liaising between policy holders, insurers, and repair firms to resolve claims rapidly and efficiently.

Industries: Financial Services
Location: Spain

About About Pervasive Technologies

Pervasive Technologies builds predictive solutions with AI and machine learning, specializing in the intelligent treatment of images and video streaming.

Google Cloud Platform logo

12 Months FREE TRIAL

Try Kubernetes Engine, BigQuery, and other Cloud Platform products with $300 in free credit and 12 months.

TRY IT FREE
Google Cloud Platform logo

12 Months FREE TRIAL

Try Kubernetes Engine, BigQuery, and other Cloud Platform products with $300 in free credit and 12 months.

TRY IT FREE