Building Custom AI Tools to Solve Real Operational Bottlenecks
As a premium fashion distributor managing products across multiple brands and suppliers, every new season meant a fresh wave of manual work: standardising incoming supplier data, reviewing imagery, renaming assets, and preparing product content for their website. Not the most exciting work for the team and processes that definitely weren't built to scale with them.
Rather than trying to automate everything at once, we took a targeted approach. We mapped the workflows that were consuming the most time, generating the most inconsistency, or sitting at the intersection of both, and we built AI tools to solve exactly those problems. The goal was never to replace human judgement, but to remove the repetitive, low-value work that was slowing the team down and introduce a meaningful review step to ensure quality didn’t diminish.
The result is a set of four custom AI functionalities embedded directly into their Sanity CMS, designed to fit the way the Moxtex team already works, not to disrupt it.
1. AI-Powered Fabric Swatch Selection for Faster, More Consistent Product Imagery
The Challenge: Fabric imagery was critical to product trust. But manually selecting a representative swatch for every item was time-consuming and inconsistent.
The Solution: We built an AI-powered system that:
- Analyses uploaded fabric images
- Selects the most representative sample
- Flags items for manual QC before going live
- Displays the swatch directly within the product description
The Result: What would require careful manual review for every product now happens instantly, consistently improving both internal workflows and the customer experience.
2. AI Automated Line Sheet Standardisation Across Multiple Fashion Suppliers
The Challenge: Each brand or supplier creates line sheets in their own format and sends these to Moxtex, resulting in a lengthy and tedious standardisation process before the data can be used. This type of process is always prone to human error given the volume of data being handled.
The Solution: We built an AI workflow that:
- Ingests line sheets from any brand or supplier format
- Extracts all relevant data
- Transforms everything into a single structured format inside Sanity
- Flags for manual QC before upload
The Result: The process now runs in the background, quietly saving time and dramatically reducing errors, integrating seamlessly into the team's existing Sanity workflows.
3. Automated Image Naming and Asset Management for Large-Scale Product Catalogues
The Challenge: Large seasonal uploads. Multiple images per product. Manual renaming. For content-heavy teams, this is never a joyful task, but if not done properly, you’re going to run into issues further down the line.
The Solution: We built an AI tool that:
- Analyses product type and orientation
- Applies a standardised naming convention automatically
The Result: No more inconsistent asset names. No more manual renaming. Just clean, predictable file management. Across a full season of products, this saves hours, if not days of manual work.
4. Automated SEO Meta Titles and Descriptions for Every Product Synced from Gung to Sanity
The Challenge: Every time a new product is synced from Gung (their PIM system) into Sanity, it needs a meta title and meta description before it can go live. These small but critical pieces of SEO copy would have been written manually, product by product, meaning a huge backlog with each new product drop and potential for inconsistent quality depending who writes the metadata. At peak season, with hundreds of products syncing through, this backlog would have a direct impact on time-to-live.
The Solution: We built an AI tool that triggers automatically whenever a product is synced from Gung into Sanity. It:
- Reads the incoming product data: name, category, brand, and key attributes
- Generates a well-structured, SEO-optimised meta title following a consistent template
- Writes a compelling meta description that balances search visibility with brand tone
- Presents both for editorial review inside the Sanity UI before publishing
The Result: Every new product arrives in Sanity with SEO-ready metadata already drafted. The editorial team no longer starts from a blank page, they review and refine rather than write from scratch. At scale, across hundreds of seasonal products, this means faster time-to-live, consistently better search visibility, and significantly less effort from the team.
Why Targeted AI Integrations Deliver Real Results, Without Overhauling Your Stack
These four tools are a good illustration of what well-targeted AI actually looks like in practice. None of them try to take over entirely, each one includes a human review step, because the goal is to augment the team's capability, not remove their oversight. What they do is eliminate the low-value, repetitive work that would drain the Moxtex team time and block product launches. Replacing it with something faster, more consistent, and easier to manage.
The common thread across all four is specificity. Rather than reaching for a generic AI solution and hoping it fits, we identified the exact friction points in Moxtex's workflows and built tools designed around how they actually operate. That's what separates AI that gets used from AI that gets shelved.
If your team is spending significant time on tasks that are repetitive, high-volume, or dependent on consistency, there's a strong chance a targeted AI integration could make a meaningful difference. The Moxtex project is proof that you don't need to overhaul your entire stack to feel the impact, you just need to start in the right place.