Over the past three years, multinational beauty company, Belcorp, has grappled with numerous challenges stemming from the pandemic, shifts in consumer behavior, disruptions in supply chains, the war in Ukraine, and inflation. To address the challenges, the company has leveraged a combination of computer vision, neural networks, NLP, and fuzzy logic.
“These circumstances have induced uncertainty across our entire business value chain,” says Venkat Gopalan, chief digital, data and technology officer, Belcorp. “As a result, we’ve found it imperative to foster greater agility and flexibility in our new product development process while maintaining high standards of efficiency, safety, and product quality.”
Belcorp operates under a direct sales model in 14 countries. Its brands include ésika, L’Bel, and Cyzone, and its products range from skincare and makeup to fragrances. As Belcorp considered the difficulties it faced, the R&D division noted it could significantly expedite time-to-market and increase productivity in its product development process if it could shorten the timeframes of the experimental and testing phases in the R&D labs.
“These stages significantly influence the iterative process of conceptualizing and rolling out a new product,” Gopalan says.
The R&D laboratories produced large volumes of unstructured data, which were stored in various formats, making it difficult to access and trace. That, in turn, led to a slew of manual processes to make descriptive analysis of the test results.
Belcorp’s answer was a new AI Innovation Labs platform, which has earned the company a CIO 100 Award in IT Excellence.
“The key objectives of this initiative can be summed up as first aiming to reduce our product development timeline by 20%,” Gopalan says. “Second, we’re striving to amplify the productivity of our lab sectors by 60%. Finally, our goal is to diminish consumer risk evaluation periods by 80% without compromising the safety of our products.”
Building the AI Innovation Lab Platform
Belcorp developed the platform in two primary stages. The initial stage involved establishing the data architecture, which provided the ability to handle the data more effectively and systematically.
“We transferred our lab data—including safety, sensory efficacy, toxicology tests, product formulas, ingredients composition, and skin, scalp, and body diagnosis and treatment images—to our AWS data lake,” Gopalan says. “This allowed us to derive insights more easily.”
The second stage focused on building algorithms and models to predict and simulate intricate biological conditions, accelerate discoveries, reduce risks, and optimize the cost-benefit ratio of technological developments using AI solutions. The team leaned on data scientists and bio scientists for expert support.
“These algorithms were built on top of an advanced analytics self-service platform, enhancing the agility of our data modeling, training, and predictive processes,” Gopalan explains.
Selling the project to executive leadership
Gopalan notes that the team considered building the platform using third-party SaaS, but ultimately decided on custom-built solutions due to the unique requirements of the R&D division, and the breadth and nature of the initiative. When the team presented the AI Innovation Lab initiative to the executive leadership team for approval, it showed them the five use cases with which it planned to start, along with associated potential value and costs.
“The business case studies highlighted how they would enable us to improve the safety, effectiveness, and performance of our formulas, and how that would translate into better time-to-market and operational savings,” Gopalan says. “To support this, we provided data-backed evidence and examples that demonstrated the positive impact of utilizing these technologies.”
Gopalan says that effectively communicating the potential benefits, demonstrating a clear ROI, and addressing any potential challenges were key to winning buy-in and support from the leadership team for the project.
Creating a cross-functional team
The team brought in experts from the R&D, technology, factory, and supply chain departments to provide a holistic view of the requirements for the project. The team spent about six months building and testing the platform architecture and data foundation, and then spent the next six months developing the various use cases.
“Deliveries were made in phases, and complexity increased with each phase,” Gopalan says. “It’s worth noting that each initiative carried its own unique complexity, such as varying data sizes, data variety, statistical and computational models, and data mining processing requirements. Therefore, setbacks or surprises were not uncommon, and we dealt with them as they arose. Working with non-typical data presents us with a reality where encountering challenges is part of our daily operations.”
Hurdles to success
As CIO, Gopalan says his biggest obstacles were the extensive and unstructured character of most of the data from R&D processes and external databases, the specific talent required for the project (including bio scientists, bio informatics professionals, technologists, and data scientists), and the cultural shift required to ensure the new platform’s acceptance.
To tackle the first challenge, Gopalan says the team concentrated its efforts on automating and cleaning the diverse data sources and formats to attain enough high-quality data to support robust analytics. They utilized data mining technologies to scrape and compile data for models from 23 international public benchmark databases, and compared that with data generated internally since 2016.
To address the second challenge, Belcorp hired new talent to bridge the knowledge gap among different teams and established a technology hub to recruit first-rate data scientists and data engineers to aid with the project’s design and implementation. Gopalan notes the data and technology team needed expertise and practical knowledge in a combination of areas, including:
laboratory processes to comprehend the data, biological processes, and business objectives of each use case
data architecture for efficient orchestration and connection of data and various platforms used in the end-to-end process
advanced analytics and AI to develop predictive solutions
software development to create customized plugins and Web apps to provide a visual interface for R&D analysts
talent training on data capabilities to ensure the end user could fully utilize the platform.
The last obstacle involved addressing the cultural change resulting from eliminating many of the laboratories’ manual processes.
“To overcome this, we trained the laboratory analysts on how to use the platform and piloted the initial use case to gather feedback,” Gopalan says. “Based on this, we made iterative changes to fine-tune the platform and its user experience. Furthermore, we succinctly conveyed the platform’s value and benefits to the end-users through a series of workshops and demos, thus ensuring the platform’s adoption.”
Now fully deployed, the AI Innovation Labs Platform has delivered 12 use cases to date that Gopalan says have yielded significant results. He points to cost savings from the reduction in laboratory tests, formulations, external software licenses, and the optimization of activities.
“The return on investment for the project stands at an exceptional 432%,” he adds.
Not only has the project delivered on expected results, Gopalan says it has also led to the digital transformation of R&D.
“Through the project’s implementation and exploration of data-driven insights, we have gained deeper insights into our product development process and customer needs,” he says. “This has opened doors to discovering new avenues for innovation and business growth, enabling us to identify and pursue additional opportunities that were previously untapped.”
Gopalan says developing the AI Innovation Labs Platform has given him five key insights into successful digital transformation involving AI and analytics:
Embrace the complexity of digital transformations. These transitions are intricate processes and mistakes are inevitable. “Rather than being deterred by these, take them as opportunities to learn and persist in your digital journey,” he says.
Follow a value-focused strategy. Focus your energy and resources on areas that have the potential to yield significant value: rapidly scale high-priority use cases, discontinue unsuccessful experiments, and use quarterly milestones for regular assessment.
Reimagine business processes. Only by reimagining and reinventing existing business processes can you truly tap the benefits of digital transformation.
Initiate an early impact narrative. A compelling success story, backed by endorsement from the executive team and prompted by a leading use case, is crucial to gain enthusiasm through the organization and among end users.
Recognize the importance of talent. Pinpointing the necessary skills and competencies, and aligning the right people in the right roles at the right time, is crucial to achieving success.
Artificial Intelligence, CIO, Data and Information Security, Data Center Management, Innovation, IT Leadership