Inside Eko’s Arkansas “capture factory”: turning every product into AI training data

The operation

It has been reported that Eko runs a sprawling “capture factory” in Arkansas where rows of workers photograph, measure and tag consumer goods — everything from shampoo bottles to sneakers — to build digital product catalogs. The aim, reportedly, is to create standardized, AI-ready records that retail-focused models can chew on: high-resolution images, SKUs, barcodes, dimensions and descriptive metadata all organized at scale. It’s boring work, meticulous work, and crucial work if you believe the future of shopping will be driven by models that need clean, consistent product data.

Why it matters

Why should you care? Because data is the new supply chain. Retailers, marketplaces and AI vendors have been starved for high-quality, labeled product datasets; Eko’s bet is that hand-built catalogs will make their models better at matching, recommending, generating listings — and maybe even automating returns and merchandising. It has been reported that the company positions these catalogs as foundational infrastructure for commercial retail AI. The upshot: what looks like mundane cataloging quietly powers slick consumer-facing features.

The human angle

There’s a small, sharp irony here. AI gets lauded as an automated leap forward, and yet so much of the work behind it is decidedly analog: people in Arkansas lifting items, adjusting lighting, typing descriptions. It has been reported that workers perform repetitive, detail-oriented tasks to ensure each product is captured correctly. Who gets the credit when the models go live? Who bears the tedium now so models can be “efficient” later? That question lands hard when you imagine the scale required to catalog “everything we buy.”

Bigger picture

This fits a larger trend: firms are investing in proprietary datasets as competitive moats. But building exhaustive, up-to-date product catalogs is expensive and never really finished. Allegedly, that unfinishedness is exactly why buyers are willing to pay — fresh, accurate data is rare. If Eko succeeds, the company could become a quiet backbone of retail AI; if it fails, it will be a reminder that some problems aren’t solved by models alone. Either way, the capture factory is where the future of shopping gets grounded in human hands.

Sources: wsj.com