Diversity in Digital Transformation: Why It Matters

The Evolving Business Landscape: A Digital Imperative

Organizations across all sectors are fundamentally becoming technology and data-driven enterprises. A swift recognition and adoption of this reality is crucial for effectively addressing evolving customer demands and fostering growth.

Reimagining business models and leveraging digital technologies is now paramount to establishing novel processes, cultivating new organizational cultures, and unlocking fresh opportunities.

Debunking the Myth of Digital Transformation

A common misconception surrounding digital transformation is that it solely revolves around the implementation of technology. This is inaccurate.

Successful digital transformation is intrinsically linked to, and dependent upon, the embrace of diversity. Artificial intelligence (AI), while powerful, is a product of human intellect, benefiting from its capabilities while also inheriting its inherent constraints.

The Three Pillars of Diversity for Success

Consequently, prioritizing diversity within organizations and teams is essential. This consideration must extend beyond conventional understandings of the concept.

Key Dimensions of Diversity

Diversity, in this context, is best understood through three core pillars:

• Cognitive Diversity: Different ways of thinking and problem-solving.
• Experiential Diversity: Varied backgrounds and professional journeys.
• Demographic Diversity: Representation across various identity groups.

By focusing on these three areas, businesses can unlock the full potential of digital transformation and achieve sustainable success.

People

The human element is paramount in the development of artificial intelligence. It is people who are responsible for creating AI systems, and therefore, the composition of AI development teams must mirror the diversity found within the broader population.

This requirement extends beyond simply increasing female representation in AI and technology positions. It encompasses a comprehensive range of characteristics, including gender identity, racial background, ethnicity, skill sets, professional experience, geographic location, educational attainment, and individual viewpoints.

The rationale behind this is clear: diverse teams, when reviewing and interpreting data for decision-making, are less susceptible to biases stemming from their own limited experiences, advantages, and constraints. This reduces the risk of overlooking the experiences of others.

We possess a collective ability to leverage AI and machine learning for positive change and to shape the future. This potential is unlocked by assembling diverse teams that accurately reflect the world’s multifaceted perspectives and rich variety.

The Importance of Diverse Teams

A broad spectrum of skills, viewpoints, experiences, and geographic origins has been instrumental in driving our ongoing digital evolution. At Levi Strauss & Co., our expanding AI strategy team is not exclusively comprised of data scientists and machine learning engineers.

Recently, we actively sought employees from various global departments and intentionally provided training to individuals without prior coding or statistical knowledge. Personnel from retail operations, distribution facilities, and design teams were enrolled in our inaugural machine learning bootcamp.

This bootcamp was designed to enhance their established retail expertise with the addition of coding and statistical skills. We didn’t restrict applications based on prior qualifications; instead, we prioritized individuals who demonstrated curiosity, analytical thinking, and a tenacious approach to problem-solving.

Building a Diverse AI Workforce

The fusion of pre-existing retail expertise with newly acquired machine learning capabilities has empowered program graduates to offer fresh and valuable insights. This pioneering initiative within the retail sector has enabled us to cultivate a skilled and diverse pool of AI professionals.

The resulting team members bring a unique blend of business acumen and technical proficiency, contributing to more innovative and inclusive solutions.

The Importance of Data in AI and Machine Learning

The effectiveness of artificial intelligence and machine learning systems is fundamentally tied to the quality and quantity of data they receive. Often, data is narrowly conceived as structured information like numerical tables. However, the scope of data extends to any information that can be converted into a digital format.

For our company, this includes digitized images of the clothing – jeans and jackets – we’ve manufactured over the last 168 years. Customer interactions, recorded with appropriate consent, also constitute valuable data.

Furthermore, data is generated from sources like in-store customer movement heatmaps and consumer reviews. Essentially, any aspect of our operations that can be digitized is now considered data. A wider perspective on data collection is crucial for maximizing the potential of AI initiatives.

Addressing Data Scarcity in the Fashion Industry

Predictive models typically rely on historical data to forecast future outcomes. However, the apparel industry is relatively new to widespread digital adoption and AI implementation. Consequently, a frequent challenge is the limited availability of past data for reference.

The fashion world inherently focuses on anticipating trends and demand for novel products, items that, by definition, lack a sales history. This presents a unique problem for traditional predictive modeling techniques.

Leveraging Diverse Data Sources for Prediction

To overcome this, we utilize an expanded range of data sources. This includes images of new products alongside a comprehensive database of products from previous seasons. Computer vision algorithms are then applied to identify similarities between past and present fashion items.

This process enhances our ability to predict demand for new products, offering more precise estimations than relying solely on experience or intuition. Data- and AI-driven predictions are now supplementing established practices.

Utilizing Digital Assets and AI for Design

At Levi Strauss & Co., we are also employing digital images and 3D models to simulate the tactile qualities of clothing and even facilitate the creation of entirely new designs. Neural networks are trained to recognize subtle variations in jean styles.

These include details like tapered leg shapes, whisker patterns, and distressed finishes. The networks also analyze the physical properties of materials that influence drape, folds, and creases.

By integrating this information with market data, we can refine our product collections to align with evolving consumer preferences, ensuring inclusivity across diverse demographics. Moreover, AI assists in generating innovative apparel styles, while preserving the creative input of our designers.

The Importance of Diverse Approaches in AI Development

Beyond the necessity of diverse personnel and comprehensive data sets, a commitment to varied tools and techniques is crucial when developing and deploying algorithms. Certain AI systems employ classification methods that can inadvertently reinforce existing societal biases related to gender or race.

Consider, for instance, how classification algorithms often operate under the assumption of a binary gender system. Individuals are frequently categorized as either “male” or “female” based on physical characteristics and ingrained stereotypes. This approach effectively excludes and invalidates other gender identities.

Addressing this requires proactive measures from everyone involved in the field, across all organizations and sectors. We must actively work to mitigate bias and promote the development of techniques capable of representing the full spectrum of human experience. One strategy involves removing race as a variable from the data, aiming for a race-blind algorithm, while maintaining vigilant safeguards against the emergence of other biases.

Commitment to Diversity in AI is a core principle at our company, and we actively leverage open-source tools to achieve this goal. The collaborative nature of open-source development fosters greater diversity, as these resources are accessible globally and benefit from contributions from individuals with varied backgrounds and perspectives.

A practical example of this approach can be seen in our U.S. Red Tab loyalty program at Levi Strauss & Co. When users create their profiles, we intentionally avoid requesting gender information and refrain from allowing the AI to make assumptions.

Instead, we ask members to indicate their style preferences – choosing from options like Women, Men, Both, or Don’t Know. This allows our AI to curate personalized shopping experiences and deliver more relevant product recommendations.

Transforming Business Through Diversity

The integration of diverse people, data, tools, and techniques is driving significant innovation at Levi Strauss & Co., and throughout the fashion industry. We are transitioning from manual processes to automation, analog systems to digital platforms, and relying on intuition to leveraging predictive analytics.

This work builds upon our company’s long-standing commitment to social values, including equality, democracy, and inclusivity – principles we have championed for over 168 years. Embracing diversity in AI represents a contemporary opportunity to extend this legacy and actively shape the future of fashion.