Skip to content
Giallo

AI for fashion: agents that know your brand's sizes, fabrics and sales campaigns.

Where does AI actually help a fashion brand?

Using AI for fashion and apparel means handing agents the repetitive work around each garment: size advice based on real measurements, product pages with fiber content taken from the spec sheet, analysis of fit-related returns, and wholesale orders. Giallo Studio builds them with Claude, GPT and Gemini; style, collection and pricing remain the brand's decisions.

An AI agent links a white mannequin and a fabric swatch. A yellow ribbon floats like silk and wraps around the mannequin

We write for independent brands, boutiques with an online store and companies that sell both to consumers and to retailers. We follow a garment's journey, from the spec sheet to the customer asking which size to get, and show where an agent we build can take over the work, with which model and under what oversight.

Where time gets lost today.

These are the points we hear about most often. Each one is a possible first agent.

“What size should I get?” is the question of every day

Customers ask in chat and in DMs whether a style runs true to size, large or small. Whoever answers goes from memory or grabs a tape measure in the warehouse, and the same question comes back with every new arrival.

An agent answers from the garment's measurements and the fit notes written by the brand, and suggests a size with a clear explanation of why.

Product pages written in a rush before every drop

Right before a launch you need dozens of product pages: fiber content, care instructions, fit, details. People copy from the spec sheet, get a percentage wrong, forget a language.

The agent drafts the page from the tech pack and photos, transcribes the fiber content from the label, checks that each variant's color matches its name and flags what is missing before publication.

Size returns that nobody analyzes

Return reasons are collected in a drop-down menu and sit in an export. That a pair of trousers keeps coming back for being too short only surfaces at the end of the season.

An agent reads return reasons and comments garment by garment and flags early when a style tends to fit differently from the size guide.

Store orders in Excel, emails and messages

During the sales campaign, buyers send orders in different formats: spreadsheets, photos of the order form, transcribed voice messages. Someone retypes them into the management software, checking minimums and variants.

The agent reads orders in whatever format they arrive, turns them into a clean draft for the management software and flags nonexistent variants, unmet minimums and deliveries to double-check.

The AI agents we design, one by one.

The size and fit agent at work: it reads the garment's real measurements, unrolls the tape measure across the sizes and recommends an M, explaining where it fits snugger; for someone between two sizes it mentions the free exchange.

Size and fit agent

Recommends a size based on the garment's real measurements and the brand's notes, explaining its reasoning.

When a customer opens the chat from a product page or messages the brand.

  1. Reads the garment's measurement chart, the stated fit and the size worn in the photos, if the brand provides it.
  2. Asks the customer which size she usually wears in another of the brand's styles, or for a few key measurements.
  3. Suggests a size and says where the garment might feel tighter or longer.
  4. If data is missing or the customer is between two sizes, it says so openly and suggests the free size exchange, where offered.

Connects to: Website chat · Instagram DMs · Size charts · Fit notes

Claude Sonnet
It can explain a size recommendation in the brand's tone and does not force an answer when the data falls short, which is what prevents a return.
Gemini Flash
It handles the spikes in questions on launch days and during sales, on the simplest requests, at a low cost per conversation.

Who approves. The brand approves size charts and fit notes; the agent never promises refunds or exchanges outside the published terms.

Product page agent

Prepares product pages in every sales language, with fiber content and care taken from the official documents.

When the spec sheet and photos of a new garment arrive, before every drop or season change.

  1. Reads the photo of the sewn-in label and reports fiber content and care symbols exactly as printed, without inferring them from the fabric.
  2. Compares each variant's swatch with the color name on your tag, for example “tobacco” or “sage,” and flags swapped variants.
  3. Writes the description in the brand's tone from the tech pack and translates it into the store's languages.
  4. Puts everything in a review table for the product manager, with the label photo next to the transcribed fiber content.

Connects to: Label photos · Swatches and still lifes · Tech pack · Ecommerce platform

Gemini
It reads images and printed text in the same request: it transcribes the label and compares the swatch color with the variant's color name.
Claude Sonnet
It keeps the brand's voice in translations too and does not add materials or finishes that the tech pack does not list.

Who approves. The product manager approves fiber content, care instructions and copy before the page goes live.

Returns and fit agent

Reads return reasons and comments and flags garments that fit differently from what the size guide promises.

Every week during the season, on the returns logged over the previous days.

  1. Collects returns with garment code, size, reason and the customer's free-text comment.
  2. Groups comments by theme: too tight, too long, fabric not as expected.
  3. Compares each style's trend with the others in the same line.
  4. Proposes concrete fixes: a fit note to add, a measurement to check, a photo to reshoot.

Connects to: Returns export · Inventory management software · Size guide

Claude Opus
It has to reason over many comments written in different ways and find the common thread: a weekly task where understanding well matters more than answering fast.

Who approves. The product manager decides on changes to the size guide and product pages; the agent proposes, it does not publish.

Wholesale order agent

Turns store orders into clean drafts for the management software and flags errors before confirmation.

During the sales campaign, every time a buyer sends an order or a change.

  1. Reads orders as spreadsheets, PDFs or photos of the paper form.
  2. Matches styles, colors and sizes to the codes in the management software.
  3. Checks order minimums, existing variants and the delivery windows set by the brand.
  4. Prepares the draft order confirmation for the sales office.

Connects to: Showroom email · Order sheets · Fashion management software

GPT
It extracts order lines from mixed formats with structured output, which the connector validates before writing the draft.
Codex
It writes the connector to the management software, through APIs or exports if the software offers them; it does not work on buyer data.

Who approves. No confirmation goes out without the sales office, which decides on discounts, exceptions to minimums and delivery dates.

Which AI model we use, and for which task.

We don't tie you to a single provider: we work with Claude by Anthropic, GPT and Codex by OpenAI, and Gemini by Google, and we choose the version during testing, on your own cases.

Claude Sonnet

  • Size and fit agent. It can explain a size recommendation in the brand's tone and does not force an answer when the data falls short, which is what prevents a return.
  • Product page agent. It keeps the brand's voice in translations too and does not add materials or finishes that the tech pack does not list.

Gemini Flash

  • Size and fit agent. It handles the spikes in questions on launch days and during sales, on the simplest requests, at a low cost per conversation.

Gemini

  • Product page agent. It reads images and printed text in the same request: it transcribes the label and compares the swatch color with the variant's color name.

Claude Opus

  • Returns and fit agent. It has to reason over many comments written in different ways and find the common thread: a weekly task where understanding well matters more than answering fast.

GPT

  • Wholesale order agent. It extracts order lines from mixed formats with structured output, which the connector validates before writing the draft.

Codex

  • Wholesale order agent. It writes the connector to the management software, through APIs or exports if the software offers them; it does not work on buyer data.

How we choose between models: the best AI model for agents.

How we work with you.

  1. Inside a season

    We start from the brand's calendar: launches, sales, the sales campaign. We pick where an agent eases the busiest moment.

  2. Size charts and the brand's vocabulary

    We gather size charts, fit notes and the way the brand describes its garments, and turn them into sources and rules for the agent.

  3. Connecting the store, warehouse and management software

    We check what the e-commerce platform and fashion management software expose, and build connectors with the minimum permissions for each agent.

  4. A trial on a capsule

    We test on a small collection or a continuing line, compare models and extend to the other lines once the product pages come out clean.

Tell us about your next drop

Tell us how you sell, online, in boutiques or to retailers, and where questions and product pages pile up: we'll propose the first agent to build for your brand.

Prefer the full form? Go to the contact page.

No spam or sales pitches. Response within 1 business day.

We use this data to contact you and, if we work together, to manage the project. We store it in Giallo's internal management system, on servers in the European Union, where only our team can see it; by email we only receive a notice, without your data. If you do not become a client, we delete it 3 months after our last contact. Details and your rights: privacy policy.

Do you prefer to write or call us?

Frequently asked questions, straight answers.

There isn't just one: very different jobs coexist in a fashion brand. We give size advice and customer replies to Claude Sonnet, labels and swatches to Gemini, retailer orders to GPT. Before deciding, we test the models on your own garments, including the ones with the trickiest fits.

Yes, if it has the garment's real measurements and fit notes. Without that data it guesses, which is why the first job we do is put the brand's size charts in order.

Yes, through Meta's messaging APIs for business profiles. The agent replies in DMs with the same rules as the website chat and hands collaborations, press and complaints to a person.

That is not what our agents do. We work on text, data and conversations; the images remain those from your photo shoot.

Yes, as drafts. The agent reads orders in different formats and prepares them for the management software, but the confirmation is always sent by the sales office.

It can help organize the data, not decide what is needed. Once the requirements for textiles are defined, the product-sheet agent will be able to collect the data required.

AI agents for fashion brands

Which garment do we start with?

Tell us how you sell, online, in boutiques or to retailers, and where questions and product pages pile up: we'll propose the first agent to build for your brand.