OpenAI published a case study detailing how Stampli, an AI-powered procure-to-pay platform, uses ChatGPT Work and Codex to accelerate product marketing from prototype to launch. The Deep Finance product launch moved from initial demo to public go-to-market in about six weeks, with the marketing team estimating a reduction from 243 modeled active role-hours to roughly 77 — a 3.16x speedup — while maintaining full human review on all customer-facing materials.
Stampli's product marketing team built a GPT-powered system that pulls context from Jira tickets, GitHub repositories, meeting notes, and other internal sources to keep help center articles, presentations, and one-pagers current. Director of Product Marketing Melad Zahedi said the automation has multiplied the output of a small team by 10x, producing hundreds of pieces of content weekly compared to just a couple before. The same infrastructure supported the Deep Finance launch across a seven-part blog series, launch emails, a webinar and deck, social and paid creative, a PR Newswire release, the product web page, and sales enablement materials. Codex also handled roughly 90% of the polished hero animation work before a contractor finished the opening scene and final format.
What's new
- Deep Finance launch timeline: Prototype to public GTM launch and first shipped product in ~6 weeks
- Production hours: Estimated 243 modeled active role-hours without Codex vs. ~77 with Codex (3.16x faster)
- Animation workload: Codex handled ~90% of polished hero animation before contractor completion
- Daily content scale: Hundreds of pieces weekly vs. a couple previously, per Zahedi
- Data sources integrated: Jira, GitHub, meeting notes, HubSpot, Gong, and other systems via API
Beyond launch events, the system serves as a "second brain" for employees preparing for meetings. In one executive session, a team member used Codex to retrieve and analyze metrics stored across HubSpot and other systems during the call — a task Zahedi said would have taken the FP&A team half a day, completed in roughly 20 seconds of keystrokes. The time saved has shifted product marketing from context reconstruction toward advising VPs and C-suite leaders on corporate and product strategy, with ChatGPT Work also serving as a thought partner for brainstorming, stakeholder personas, and pressure-testing recommendations.
Why it matters
The case study illustrates a practical deployment pattern for agentic AI in a mid-market B2B company: connecting fragmented internal data sources (Jira, GitHub, CRM, call transcripts) to an LLM-powered workspace that both maintains living documentation and executes production tasks. Stampli's approach keeps humans in the loop for final approval while offloading synthesis, drafting, and data retrieval. For organizations evaluating ChatGPT Work, the Stampli example shows a measurable before/after comparison on a real product launch, not a controlled benchmark.
Our take
The 3.16x speedup figure is specific to a modeled workflow for one launch, not a universal productivity multiplier. What distinguishes Stampli's deployment is the infrastructure investment — building automated pipelines from source systems into the AI workspace — rather than ad-hoc prompting. That upfront engineering is the replicable part; the hour-count savings will vary by team maturity and data hygiene.
