Stampli faced a fixed launch deadline while its design team was allocated to other projects. To meet the timeline, the company turned to OpenAI’s Codex and ChatGPT Work, using the models to generate design assets, copy, and code snippets that would normally require weeks of manual effort. By integrating these generative tools into its workflow, Stampli compressed what would have been a multi‑week production phase into a matter of days.
Codex is a decoder‑only transformer fine‑tuned on publicly available source code, offering a 4 k‑token context window and accessed via OpenAI’s commercial API under a standard usage license. ChatGPT Work runs on the GPT‑4 Turbo architecture, which provides a 128 k‑token context window and is likewise offered through the API with comparable licensing terms. The large context windows allowed Stampli to feed extensive design briefs and specifications into the models, receiving coherent outputs that could be directly incorporated into the launch materials, thereby enabling parallel workstreams despite limited designer availability.
Why this matters
The Stampli example shows how API‑based large language models can mitigate human resource constraints in product cycles, emphasizing that model context length and licensing terms are practical factors for enterprise adoption; while the case demonstrates measurable time savings, it also highlights the need to evaluate ongoing API costs, data privacy, and model reliability when relying on generative AI for core deliverables.
Share this article
Found this insightful? Share it with your community on Reddit, X, or copy the link.