What Changed: From Codex to ChatGPT Work
On May 12, 2026, OpenAI published an Academy article titled “Finance workflows with ChatGPT Work.” The piece explains how finance teams can use ChatGPT Work to turn everyday work inputs—close workbooks, revenue and expense dashboards, forecast updates, prior monthly business reviews, and owner notes—into review-ready deliverables. The key shift is that these workflows previously lived in the former Codex app; now they are available directly in ChatGPT Work at chatgpt.com or in the ChatGPT desktop app. The article includes a note that the accompanying webinar was recorded when the workflows were still in Codex, but users can now follow along in ChatGPT Work.
For product builders, this transition matters because it signals a move from a standalone coding assistant to a more integrated work environment. ChatGPT Work is positioned as a tool that requires no coding, making it accessible to finance professionals who are not developers. The emphasis is on starting from existing work products rather than forcing users to adopt a new data structure or workflow.
How It Works: Starting from Real Work Inputs
The core idea is simple: instead of starting from a blank document, finance teams feed ChatGPT Work the materials they already use. According to the OpenAI article, these inputs include:
- Close workbooks
- Revenue and expense dashboards
- Forecast updates
- Prior monthly business reviews
- Owner notes
ChatGPT Work then helps turn that context into concrete assets that the team can review, refine, and share. The goal is to spend less time assembling the first draft and more time shaping the story, validating the numbers, and preparing for the next decision.
This approach is notable for what it does not require: it does not ask users to change their habits or learn a new way of organizing data. The tool meets users where they are, which lowers the barrier to adoption. For product teams, this is a design principle worth noting: integrate with existing workflows rather than demanding new ones.
Practical Use Cases for Finance Teams
The OpenAI article highlights several top use cases for finance teams, though it does not provide detailed step-by-step instructions. The mentioned use cases include:
- Monthly business reviews
- Reporting
- Variance analysis
- Planning
These are common, repetitive tasks that often involve assembling data from multiple sources into a coherent narrative. ChatGPT Work aims to automate the assembly part, leaving the analytical and communicative work to humans. The article also points to a collection of use cases at learn.chatgpt.com/use-cases/collections/finance and a solutions page for finance teams at openai.com/business/solutions/finance/.
It is important to note that the article does not provide quantitative results, such as hours saved or accuracy improvements. It presents a direction rather than hard metrics. This is a useful reminder for product builders: not every AI feature needs a benchmark to be valuable; sometimes the value is in enabling a different workflow.
Limitations and Trade-offs
While the article is optimistic, it is also careful to note limitations. The most obvious is that the webinar was recorded in the former Codex app, and the workflows have since been migrated to ChatGPT Work. This means that some details in the webinar may not exactly match the current interface. Users are advised to follow along with ChatGPT Work directly.
Another limitation is the lack of specific examples or case studies with measurable outcomes. The article does not cite any companies that have successfully used ChatGPT Work for finance, nor does it provide before-and-after comparisons. This absence of evidence is not necessarily a red flag, but it means that readers should treat the claims as directional rather than proven.
For product builders, this suggests a trade-off: the tool is designed to be flexible and adaptable, but that flexibility comes at the cost of predictability. The output quality may vary depending on the quality and structure of the inputs. Teams will likely need to iterate and refine the prompts and inputs to get the best results.
What Product Builders Can Learn
One of the most striking aspects of the OpenAI article is the repeated use of the term “review-ready.” The output is not presented as a final answer but as an asset that the team can review, refine, and share. This is a crucial design philosophy: the AI is not meant to replace human judgment but to support it. The human remains in the loop, and the AI’s work is treated as a draft that requires validation.
For product builders, this raises a key question: is your output a draft that needs review, or a finished product that can be sent directly? The former is more likely to be accepted by professional teams because it preserves human control and allows errors to be caught during review. The latter may be faster but risks eroding trust if the AI makes mistakes.
Another lesson is the importance of starting from existing inputs. By allowing users to upload their own workbooks and dashboards, ChatGPT Work avoids the friction of requiring users to reformat their data. This is a powerful onboarding strategy: reduce the cost of trying the tool by making it work with what users already have.
Getting Started: A Concrete Next Step
If you are a product builder or a finance professional, the OpenAI article suggests a practical way to test ChatGPT Work: pick the most repetitive and time-consuming part of your monthly reporting process, gather the relevant workbooks and dashboards, and see what kind of draft ChatGPT Work produces. The goal is not perfection but to see if it can save time on assembly, allowing you to move faster into substantive discussion.
The article provides links to the use case collection and the finance solutions page, which are good starting points. However, it does not include specific prompts or code examples, so you will need to experiment on your own. The key is to treat the output as a starting point, not a final product, and to iterate based on your team’s standards.
In summary, ChatGPT Work represents a shift toward making AI assistants more accessible to non-technical teams. By focusing on review-ready outputs and starting from existing work inputs, it offers a template for how AI tools can be designed to augment human expertise rather than replace it. For product builders, the lesson is clear: design for review, not for automation.
Sources
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.