AI When entering the creative industry, it is often misconceived as "machines taking over the design process." However, the collaboration during New York Fashion Week on September 18th took a different approach: engineers worked together with designers Jane Wade and Sergio Hudson to create two specialized tools that addressed the recurring issues of model fittings and stage setup adjustments. The focus was not on creating a garment that looked novel, but on reducing the need for rework in the processes of sample garment production, fitting trials, lighting coordination, and venue preparation.
A large amount of an independent designer's time is not spent on drawing. Coordinating with factories, communicating with suppliers, having models try on outfits, matching accessories, and managing venue budgets can consume entire workdays. Google Envisioning Studio and Google Labs therefore did not start by creating a general chatbot for designers. Instead, they involved engineering staff in the actual processes to observe which decisions were made repeatedly and which errors only became apparent during the physical product development stage. After that, they turned these steps into actionable tools.
Jane Wade uses Styling Suite to combine clothing, shoes, accessories, hairstyles, and makeup on digital models. As Google states, traditional on-site casting and fitting processes usually take up to three full working days for a team. Digital rehearsals cannot replace the examination of fabric drape, size adjustments, or real-person walks on the runway, but they do allow the team to see whether there is an imbalance in body proportions before cutting additional samples, identify any missing elements in a particular outfit, and check whether accessories are drawing too much attention away from the main visual focus. What is saved is not "inspiration time," but rather the material and back-and-forth communication required to identify obvious issues.
Both tools start from specific constraints, rather than forcing a general model into the process.
Sergio Hudson faces a different kind of cost. Every time the lighting, props, and movement paths in the fashion show are changed, the production team may need to redo the 3D rendering, which increases the budget with each modification. Runway Visualization suggests that they first adjust the venue layout, lighting, and props in a virtual environment, and then compare the models' walking routes. The value of these tools lies not in producing a beautiful render, but in allowing creative choices to be discussed within the constraints of budget and space.
These two sets of tools reflect a step often overlooked in the implementation of AI for enterprises: to first narrow down the problem. The styling tools are not responsible for contracts, and the runway tools are not responsible for the supply chain; they are designed to create workspaces around high-frequency, visual, and easily revisable decisions. A narrower scope actually makes it easier to define what is good or bad, and it also allows for the designer's final judgment to be better preserved. Google states that the related results have already been presented on the runways of New York Fashion Week, but this is still a collaboration between two designers and two workflows, and it cannot be generalized to indicate that the entire fashion industry has completed the AI transformation.
Google also allows users to describe the tools or workflows they desire in natural language, and to build them themselves within Flow. The lack of a code-based entry point lowers the barrier to experimentation, but it does not automatically solve the issue of data organization. Brands need to prepare clothing images, sizes, materials, available props, venue restrictions, and budget boundaries; if the asset names entered are confusing or the colors are distorted, even a smooth generation result could lead the team in the wrong direction. The design industry is particularly concerned with color, texture, and pattern, and there must be a verification step between screen previews and the actual products.
Whether a tool is truly useful depends on its ability to integrate into the existing collaboration chain. Designers, stylists, pattern makers, and production companies require different information. If the generated results can only be viewed on a closed page, teams still have to take screenshots, rename files, and perform manual synchronization. Mature products should retain versions, annotations, reasons for adoption, and sources of materials, so that a single selection can be carried forward throughout the production process. Otherwise, the claim of "fast generation" will only shift the bottleneck to approval and file organization.
Moving from a pilot program to a regular practice, the real issues are copyright, cost, and who holds the ultimate decision-making power.
For fashion companies using generative tools, the first question to answer is whether the materials can be uploaded. Unreleased series, model photos, brand logos, and supplier quotes may all be subject to confidentiality or rights restrictions. Teams need to clearly define training and retention policies, access rights, and methods for cleaning up after the project is completed. If digital models are based on real people's appearances, portrait authorization is also an issue; when referencing historical clothing and artistic styles, it is important to distinguish between internal inspiration boards, commercially usable materials, and protected designs.
Next is cost accounting. Saving three days for on-site trial installations sounds impressive, but digital tools also require organizing materials, debugging processes, training employees, and reviewing results. The most reliable way to assess the impact is not by counting how many images were generated, but by whether the number of sample garments decreased, whether there was less rendering rework, whether changes to the set design were detected earlier, and whether the final production costs were within budget. If the team merely has an additional set of generation tools but still repeats all the steps according to the original process, AI may actually increase rather than reduce the workload.
Finally, there is the issue of creative responsibility. Styling Suite can list combinations, but it cannot determine what a brand wants to convey in a particular season; Runway Visualization can simulate lighting, but it cannot bear the reputational costs of a fashion show failure on behalf of the designers. Keeping people in their respective roles is not just a comforting statement; it must be reflected in their authority: models are responsible for comparison and rehearsals, designers decide what to adopt, and the production team confirms whether it can be put into practice.
The most valuable aspect of this collaboration is that it doesn’t present AI as an “automatic designer.” Instead, it functions more like a digital prototype room and a rapid scenario sandbox, allowing for earlier experimentation with what would otherwise be expensive and time-consuming trial and error processes. In the creative industry, the tools that are truly willing to be used over the long term are not necessarily those that claim to excel at creativity the most, but rather those that enable teams to avoid making unnecessary prototypes, reduce the need for costly lighting adjustments, and free up limited time for actual creation, without taking away the control from the creators.
