commercetools has taken a pragmatic approach to generative AI (GenAI), emphasizing practical applications over flashy features. The company enables GenAI by providing open access to extensive data, enabling applications like code generation and debugging — presenting itself as a facilitator rather than a direct provider in the GenAI landscape.
In the new world of GenAI, commercetools has steadfastly embraced a philosophy where pragmatism trumps the allure of shiny new features. It’s crucial to elucidate from the outset that we, at commercetools, are not the purveyors of the Language Learning Models (LLMs). Instead, our unwavering commitment to an open default approach means that these LLMs, developed and offered by third parties, come pre-trained on our expansive and publicly available data, spanning documentation, code and more, dating back to 2011.
Our engagement with GenAI is not about direct provision but about creating an environment where it can be applied meaningfully and practically by leveraging the wealth of data we’ve made openly available. Applications like code generation and debugging, simplifying integrations, and parsing documentation become possible, not by our offering of LLMs but by enabling them to be well-versed with commercetools through our extensive, open data. This approach not only accelerates time-to-value but also strategically exposes data (always opt-in) to enhance and facilitate interactions and functionalities.
Although we do not offer LLMs, the pragmatic applications of GenAI, as enabled by our strategy, percolate into various operational facets of commercetools. It provides a framework where interactions with projects are not just possible but are enriched and enhanced, offering capabilities like code generation and debugging which, while not production-ready, certainly position themselves as valuable assets in developmental and demonstration contexts. Auto-GPT plugins and real-time sync capabilities are examples where LLMs, trained on our data, create a landscape of possibilities in interactive engagement and user prompt responses.
However, it’s imperative to underscore the boundaries of our engagement with GenAI. While we empower GenAI applications by providing a fertile ground of data and information, the generated code, especially that which is utilized for demonstrations, may not be production-ready and may necessitate further refinement and validation. Moreover, while we facilitate data exposure to LLMs, this is always an opt-in feature, safeguarding our steadfast commitment to data privacy and client control.
Concluding, commercetools stands as a facilitator, not a provider, in the GenAI landscape. Our strategy, embedded in pragmatism and an open default approach, ensures that while we do not offer LLMs, we create an environment where they can be applied in a manner that is both meaningful and practically beneficial to our clients and partners. It's a fine dance of enabling innovative and efficient GenAI applications while upholding unyielding standards of data security, control and pragmatic operational applicability.
To learn more about how commercetools is experimenting with GenAI, read our blog post Prompt engineering for a commercetools implementation using generative AI.