Analytics powered by data and artificial intelligence (AI) promises organizations the ability to speed up decision-making, unlock insight upon which they can base positive action, and even drive changes to their business models. The experience of Peruvian insurance company RIMAC Seguros y Reaseguros is that all those benefits are achievable – but not without putting in a significant amount of groundwork.
RIMAC’s pioneering work in data and AI dates from the arrival of CEO Fernando Rios at the business in 2017. Following a stint in banking (an industry that has worked hard to capitalize on new data-driven technologies), Rios joined RIMAC as Vice-President of Insurance and Marketing. He felt that a data-driven approach could pay off for RIMAC as well, enabling a personalized and customized approach to insurance that would shake up the industry.
Rios’s first challenge was getting RIMAC’s existing data team up to speed. He hired Miguel Paredes, a machine learning (ML) and data science specialist, to head up a small team tasked with proving that data analytics could deliver better outcomes to specific business problems. Paredes soon discovered that RIMAC was using data in only very limited, traditional contexts, such as reporting. The business was undertaking some business intelligence work but, at the time, the group only had the capacity to serve one-third of the organization. Moreover, it lacked the bandwidth to develop mature data governance and management structures.
Paredes decided to put the work on advanced analytics and AI on hold, focusing instead on getting RIMAC’s data in order. This proved challenging, with the team forced to reach out to several business units for access to their data, and then extract this data themselves before analyzing it for quality.
In year one, RIMAC’s areas of focus included data governance and democratization, as well as the launch of a new data platform. Collecting and aggregating data from the insurer’s core IT systems required new processes and architecture, and the company struggled to secure skills in areas such as cloud engineering. But Paredes recognized that, without this foundational work, it would be impossible to build and deploy AI models.
Even at this early stage, the project was driven by the desire to realize tangible business benefits. Paredes’ team continually re-ran their models as part of the development process, feeding in new information relating to the needs of the different business units. Early successes included a five-percentage-point reduction in customer churn in vehicle insurance, worth around $750,000 a year in policy renewals, and the development of a customer classification metric that improved customer acquisition by 10%.
These achievements helped win support for Paredes’ work among the business’s leadership, but making progress was still tough. One issue was that the team was simply responding to requests from the business, rather than being free to explore the underlying model. The data science team then worked in isolation to develop an algorithm, rather than collaborating with their business partners, meaning that the models took too long to develop and often proved unscalable.
By mid-2019, RIMAC was sufficiently confident in the project to take it to the next level. Paredes recruited Rosario Alcedo, an analytical transformation specialist, to help him build a center of excellence (CoE) for advanced analytics and AI. Their objective was to develop RIMAC’s analytical capabilities in order to achieve the full potential value of solutions.
The pair were also determined to work more closely with the rest of the business. The CoE would serve the insurer’s business units – life, health and enterprise insurance – but also co-ordinate with other functions to deliver value. Alcedo insisted that business partners should feel they owned the solution, to the point where they could confidently explain the algorithm in use.
Running with this new approach, Alcedo and three key project leads began spending a considerable portion of their time working on understanding business problems and strategic opportunities under consideration.
The CoE also began to add atypical and unstructured data capabilities to their work (audio for speech analytics and images for computer vision, for example). The team subjected this data to more sophisticated algorithms and deep learning techniques. Again, several successful use cases proved that the project was moving forward. RIMAC was able to use computer vision to assess vehicle insurance claims, hugely reducing processing time, with an accuracy rate of 95%. Elsewhere, analysis of customers’ past behaviors helped RIMAC identify those likely to pay premiums late, or even to skip payments; that helped it boost collection rates by 15 percentage points.
Another win came from speech analytics, which enabled RIMAC to analyze the reactions of customers calling its call center, and to act accordingly to enhance customer experience. As a result, positive sentiment doubled in a year.
Naturally, these successes began to attract attention, with business units approaching the team with a growing number of new requests. This prompted the CoE to implement a prioritization system. Each initiative was assessed according to three criteria: feasibility based on data availability; the execution-readiness of the business unit; and potential impact.
Fast forward two years and RIMAC was ready to go even further. In mid-2021, Rios oversaw a reorganization, with the CoE reporting to RIMAC’s CMO, and the data engineering team moving under the umbrella of IT (reflecting the interconnected nature of the two functions’ work in areas such as control mechanisms, applications and cybersecurity). With the CoE divided into three groups – data science, ML ops, and an AI project lead team ‒ the ultimate goal was to scale up the project.
One crucial initiative at this stage was the launch of an “AI factory” to standardize processes, scale up model deployment, and share knowledge and best practice. The planning of this unit was a deliberately inclusive exercise that sought views from business users, data engineers and the IT department, which had been charged with implementing the technology. Their objective was to move away from piecemeal AI initiatives to a more systematic approach that involved working with partners on business problems and value cases.
Key elements of the AI factory include a “feature store,” which is a repository of clean, standardized data and a system to allow data scientists to access features of the model. It also includes an AI platform, a set of tools, guidelines and templates that save data scientists’ valuable time, and an MLOps platform that transforms the algorithms they develop into consumable solutions. A model governance framework is also important, setting out requirements for performance monitoring.
Over time, RIMAC has also begun to develop a federated model that supports the scaling of its AI activities. As business units began to compete more strongly for attention from the CoE, they also began to push for distributed capabilities. RIMAC began to develop a model for this in its health insurance business, where a dedicated team of 10 worked on organizing and structuring claims data. One positive outcome was the launch of an incentive program to encourage customers to adopt healthier lifestyles.
RIMAC’s progress on AI may sound straightforward, but the reality was anything but. At each stage of the project, all those involved faced significant challenges. First, they uncovered a number of gaps in the data, both in terms of availability and quality, with only limited data sharing across business units. While this might, at a glance, seem inconsequential, there could be significant commercial repercussions. For instance, RIMAC’s health insurance business knew whether a customer had a child, but the life insurance business, ideally placed to sell important protection policies to such a customer, did not receive this data.
Another problem was organizational dynamics. Having split its capabilities between advanced analytics (within the marketing function) and data engineering (IT), RIMAC had to manage the tensions that arose between the two teams (a not uncommon occurrence, with the latter responsible for the complex task of delivering clean, usable data, and the former under pressure to deliver quickly for the business).
Indeed, increased demand from the business has become a significant challenge. The CoE initially focused on quick wins, particularly in vehicle and health insurance, but life insurance leaders saw the bigger picture and wanted a share of the benefits. As a natural consequence of this, the demands on the team have grown very rapidly.
Other obstacles include a lack of alignment of the objectives set for the CoE and for individual business units. This has sometimes led to issues around prioritization and co-operation. Similarly, defining and accounting for the value created by the team’s AI models has proved challenging.
One crucial initiative at this stage was the launch of an 'AI factory' to standardize processes, scale up model deployment, and share knowledge and best practice.
Another learning is that business leaders inevitably take a short-term view of data, focusing on what is relevant for the next decision. Subtly nudging mindsets towards a longer-term perspective was, therefore, a key part of managing both expectations and approaches to the new facility.
Having worked so hard to build up RIMAC’s capabilities, Rios is looking forward to harnessing them for transformative benefit. He expects the business’s newfound digital competencies to play a crucial part in moving the insurer into the wellbeing marketplace. Here, RIMAC has an opportunity to build a positive ongoing relationship with customers, rather than a one-off transaction that gets forgotten about until and unless it’s needed.
That shift began in 2022, with the development of Estar Bien, RIMAC’s new wellbeing platform. The platform combines the insurer’s internal data with big data sourced externally in order to enable real-time customization of the value proposition for customers.
The intention is to move RIMAC into a space where it operates as a trusted adviser that can improve customers’ lives. If it can achieve this status with its customer base, RIMAC can truly say that its commitment to data has paid off.
Loh Chin Hua, CEO of Keppel, tells Jean-François Manzoni how bold restructuring, strategic discipline, and a long-term vision transformed an industrial conglomerate into a global asset manager...
Companies like Nestlé, Toyota, and Intel drive innovation by systematically nurturing curiosity, creating diverse teams, balancing autonomy with visibility, and embracing external partnerships....
With solar power flooding grids, the duck curve creates both challenges and opportunities for businesses. Learn how smart companies can harness this shift to cut costs and boost efficiency....
The pandemic sparked a shift to remote work that’s proving hard to reverse. According to Prof. Manju Ahuja, hybrid models are the new norm—driven by employee expectations, technological advances, and a changing...
Explore first person business intelligence from top minds curated for a global executive audience