Catalog management used to mean spreadsheets: a master file, a dozen formulas, and a prayer before every import. In 2026 the workflow looks different. AI handles the drafting, rules handle the math, and humans handle the judgment. This guide describes the modern catalog workflow end to end — and where each piece fits.
The old workflow and its costs
The spreadsheet era of catalog management has four recurring costs:
- Time. Adding 50 products means 50 rows of manual typing across 15+ columns, plus variant rows multiplied on top.
- Errors. A shifted comma, a pasted-over SKU, a price typed with a currency symbol — each one fails an import or corrupts a product silently.
- Fragility. The master spreadsheet depends on formulas nobody remembers writing. One broken reference and the next export is wrong.
- Bottlenecks. Catalog changes queue behind whoever owns the spreadsheet and knows the formulas.
None of these costs are obvious day to day. They surface at scale — a seasonal menu, a vendor switch, a platform migration.
The modern workflow
1. Generate with AI, structure with code
Describe the catalog in plain language — products, prices, sizes, add-ons — and let AI draft the product data. The critical design decision is what the AI does not do: it drafts names, descriptions and plausible values, while deterministic code handles structure (column names, option-group encoding, quoting) and exact math (price deltas, percentage changes). AI for language, code for numbers.
2. Convert instead of rebuild
Platform moves used to be rebuild projects. Now they are conversions. A Shopify export converts to the Hyperzod structure with variants collapsed into option groups and prices recalculated as deltas. A DeliverLogic menu export converts with its option groups and add-on prices intact. The source file stays untouched; the converted result is reviewed before use.
3. Edit by describing
Bulk changes become sentences: "make all products 15% cheaper", "set inventory to zero for the discontinued category", "rename the Beverages category everywhere." The tool finds matching rows and applies the change, showing the result in a preview. This replaces the find-and-replace-plus-formula dance that caused most catalog errors in the first place.
4. Review before export
The preview grid is the quality gate. Inferred values are highlighted, validation runs continuously — required fields, unique SKUs, plain numbers — and nothing is exported until the error count is zero. The review step is what makes AI-assisted generation safe: the model drafts fast, and the human approves deliberately.
5. Export to the platform's exact format
The export writes the target platform's column structure — for Hyperzod, the PRODUCT.* columns with option groups — sanitized against formula injection and encoded as UTF-8. The file that reaches the import is the file you reviewed.
What still needs a human
Three judgments AI should not make alone:
- Truth. Whether a fact is real — materials, certifications, origin claims.
- Pricing strategy. What a discount should be; the tool computes it exactly, but you choose it.
- Taxonomy decisions. Which category a product belongs to, and when the category tree itself needs restructuring.
Getting started with the modern workflow
Start small: describe a five-product catalog in CSV Pilot, review the preview, and export. Then convert a real export — Shopify or DeliverLogic — and compare the result to a hand-built version. Most teams adopt the workflow in an afternoon, and the spreadsheet marathon becomes the exception rather than the routine.
The tools changed. The discipline — review before you publish — is the same as it ever was.