The 'Dark Factory': Building AI at Scale
Klaviyo, a prominent e-commerce marketing platform, has achieved significant scale, reaching $1.5 billion in valuation and growing to 2,300 employees. A key to this growth, according to co-founder and co-CEO Andrew Bialecki, is a highly disciplined approach to building AI products, which he detailed at SaaStr AI. Bialecki emphasized that his presentation wasn't about vision decks, but about the actual build system. This system, which he terms the 'Dark Factory,' focuses on the practical, often unglamorous, work of engineering AI at a large, public company.
The core insight that propelled Klaviyo was moving beyond simple email metrics like open rates. Instead, Klaviyo connected directly to the e-commerce cart, understanding transactional data. This foundational principle of deep integration and data utility underpins their current AI efforts. The 'Dark Factory' concept implies an operational focus, where the machinery of AI development runs continuously and efficiently, often behind the scenes, much like a manufacturing plant. It’s about the plumbing, the infrastructure, and the rigorous processes that enable the creation of sophisticated AI features without constant fanfare.
Bialecki explained that for a company of Klaviyo’s size, building AI isn't about a few brilliant researchers in a lab. It’s about systematizing the entire process, from data ingestion and model training to deployment and monitoring. This requires a robust engineering culture and tooling that can handle the complexity and volume of data generated by millions of e-commerce merchants. The 'Dark Factory' is where this complex machinery operates, ensuring that AI capabilities are reliably integrated into the core product and deliver tangible value to customers.
Composer: The Orchestration Layer
Central to Klaviyo’s AI build system is a proprietary internal tool Bialecki calls 'Composer.' This is not a product for customers, but an internal development environment designed to abstract away the complexities of AI model development and deployment. Composer acts as an orchestration layer, allowing engineers to manage the lifecycle of AI models more effectively. This includes everything from data preparation and feature engineering to model training, evaluation, and serving.
Think of Composer less like a single AI model and more like a sophisticated factory floor manager. It doesn't perform the core AI tasks itself, but it directs the resources—data scientists, engineers, compute power, and code—to accomplish them efficiently. This abstraction is critical for scaling AI development. Without such a system, each new AI feature or model iteration would require significant manual effort to set up, train, and deploy, creating bottlenecks and slowing down innovation. Composer streamlines these workflows, enabling faster experimentation and iteration.
The tool allows for defining complex AI pipelines as code, making them versionable, testable, and reproducible. This mirrors best practices in traditional software development but applied to the unique challenges of machine learning, such as managing large datasets, tracking experiments, and handling model drift. By providing a consistent framework, Composer ensures that AI development is not a black art but a repeatable engineering discipline, accessible to a broader set of engineers within the company, not just a select few AI specialists.
The L3 Mandate: Skill Elevation for All
Perhaps the most striking aspect of Klaviyo’s AI build strategy is the company-wide mandate for every employee to reach 'L3' proficiency by June. L3, in Klaviyo’s internal leveling system, signifies a level of skill and autonomy that allows an individual to independently tackle complex tasks and contribute meaningfully to AI initiatives. This is not about making everyone an AI researcher, but about ensuring a baseline understanding and capability across the organization.
This aggressive upskilling goal is a direct response to the pervasive nature of AI in Klaviyo’s product strategy. Bialecki believes that to truly leverage AI, the entire company needs to be AI-literate. This means that engineers understand how to integrate AI models, product managers understand how to identify AI opportunities, and even customer-facing roles have a foundational grasp of what Klaviyo’s AI capabilities are and how they benefit customers. It's a cultural shift, embedding AI thinking into the company’s DNA.
The mandate reflects a strategic decision: rather than relying solely on hiring specialized AI talent, Klaviyo is investing heavily in developing its existing workforce. This approach not only scales AI development more rapidly but also fosters a deeper sense of ownership and understanding among employees. The 'Dark Factory' requires skilled operators, and L3 represents the minimum standard for those operators. This ensures that the complex machinery built with Composer can be effectively utilized and maintained by a broad base of capable individuals, turning a company-wide AI vision into a practical reality.
Implications for E-commerce AI
Klaviyo’s approach highlights a critical shift in how B2B SaaS companies are building and deploying AI. The focus is moving from speculative, high-level AI visions to the grounded reality of engineering robust, scalable systems. The 'Dark Factory' concept, coupled with internal development tools like Composer and aggressive employee upskilling, suggests a mature strategy for embedding AI deeply into product offerings. For e-commerce businesses, this translates to more intelligent, data-driven marketing tools that can be deployed reliably and evolve quickly. Competitors will need to match this operational rigor, not just the AI features themselves, to remain competitive.