Merge Builds Product and Business Workflows Around OpenAI Tools

Merge is using OpenAI technology across software development and business operations, according to a case study published by OpenAI. The company says its teams use GPT-6 Astra, ChatGPT and Codex to review product code, support marketing, prepare sales updates and organize executive dinners.

The account offers a snapshot of how one software company is applying multiple OpenAI products beyond a single chatbot workflow. It also shows the limits of the available evidence: OpenAI’s published material is the primary source, while the accompanying Google News item does not provide additional reporting or independent verification.

What Merge is using OpenAI for

OpenAI News identifies three tools in Merge’s internal workflow: GPT-6 Astra, ChatGPT and Codex. The company reportedly uses them across both technical and commercial tasks rather than restricting AI to engineering.

The most direct product-development use is code review. OpenAI says Merge uses GPT-6 Astra and Codex to review product code. That suggests an AI-assisted engineering process in which models help examine changes before they are merged, although the available source does not specify whether the tools approve code, flag potential defects, generate tests or simply provide suggestions to developers.

Merge also uses OpenAI tools to support marketing, according to the case study. The source does not identify the campaigns, content formats or review process involved. That makes it difficult to assess how much of the work is automated versus accelerated by human teams.

On the business side, OpenAI says Merge uses AI to prepare sales updates and plan executive dinners. These examples place the technology in routine coordination and communication work, where the value may come from organizing information and producing first drafts rather than making final decisions.

A multi-tool deployment, not a single use case

The reported pattern is notable because it connects engineering, marketing, sales and executive support in one company. For product teams, Codex is associated in the source with code review. For broader workplace tasks, ChatGPT and GPT-6 Astra are presented as part of Merge’s operating toolkit.

That division should not be treated as a formal architecture. OpenAI’s summary does not explain which model handles which task in every workflow, how employees move work between the products or whether Merge has built internal integrations around them. It also does not disclose the size of the teams involved, the volume of work processed or the amount of time saved.

Still, the examples point to an important deployment choice. Merge appears to be evaluating AI as a shared capability across departments rather than as an isolated engineering experiment. That can reduce the number of separate tools employees need to learn, but it also raises questions about permissions, data handling and consistent review standards across very different types of work.

What the evidence does and does not show

The specific product and adoption claims come from OpenAI’s official Merge case study. As a vendor-published account, it should be read as reported customer usage rather than independent validation. The source confirms that OpenAI describes these workflows, but it does not provide audited productivity results, error rates, customer outcomes or a comparison with Merge’s previous processes.

The mention of GPT-6 Astra is also significant but difficult to evaluate from the supplied material. OpenAI provides no technical specifications, pricing information, context window details or benchmark results in the available summary. Without those details, the report cannot establish whether the model is materially better for Merge’s work than other available systems.

Likewise, the source does not say whether AI-generated code has introduced defects, whether human reviewers remain mandatory or how Merge protects proprietary product information. Those omissions matter because code review and sales preparation involve different risks. A mistaken code suggestion can affect reliability, while an inaccurate sales update can distort internal planning or customer communication.

Implications for builders and enterprise buyers

For builders, Merge’s reported approach highlights a practical starting point: map AI tools to recurring workflows instead of deploying a model without a defined job. Code review, marketing assistance and sales reporting have different inputs, outputs and approval requirements. Treating them separately can make it easier to measure quality and assign responsibility.

The example also reinforces the need for human checkpoints. The source describes AI helping with product code and business materials, but it does not claim that Merge has removed people from these processes. In most enterprise settings, the important implementation question is not whether a model can produce an answer, but who verifies it before the answer affects software, customers or executives.

For enterprise buyers, the case study offers evidence of breadth rather than proof of return on investment. A company considering ChatGPT, Codex or another model would still need to test performance on its own repositories, marketing standards and sales data. It would also need clear controls for access, retention, confidential information and auditability.

The multi-department pattern may nevertheless be commercially relevant to OpenAI. If customers use the company’s products across technical and administrative work, OpenAI can compete for a larger share of workplace software budgets. But that opportunity depends on showing reliable results in each workflow, not merely demonstrating that a model can be used in many places.

What to watch next

The next useful signal would be more detail from Merge or OpenAI on measurable outcomes. That could include changes in code-review time, the number of issues identified, marketing production volume or the time required to prepare sales updates.

Readers should also watch for information about deployment controls. Details on repository access, approval requirements, model monitoring and handling of sensitive sales or executive information would clarify whether these uses are experimental or embedded in daily operations.

Finally, additional customer accounts could show whether Merge’s approach is unusual or part of a broader shift toward shared AI platforms. Independent reporting, rather than vendor case studies alone, would be especially valuable for assessing reliability and actual business impact.

Creati.ai perspective

Merge’s example is less a proof of dramatic automation than a useful illustration of AI becoming a layer across ordinary company workflows. The reported uses are varied, but the evidence does not establish that any of them has produced quantified gains.

The practical lesson for AI builders and buyers is to evaluate each workflow on its own terms. Code review needs technical accuracy and strong controls; marketing and sales work need factual consistency and approval discipline. OpenAI’s account shows where the tools are being used, but the harder questions—quality, governance and measurable value—remain open.