Product update · October 16, 2026
From AI assistant to agentic catalog operations
Tabgrid is bringing AI closer to the structured product workflow—and making longer tasks observable, bounded, and reviewable.
The idea: useful autonomy, with explicit boundaries
An agent should do more than answer a prompt, but it should not become an unbounded operator. Tabgrid task plans let a team define an ordered list of supported product tasks, select the exact products, choose an approval mode, and follow progress as the agent works through one step at a time.
A plan, not an open-ended command
A plan can contain up to 20 sequential tasks and target up to 200 products. Current task-plan actions cover quality checks, AI edit proposals, AI enrichment proposals, and linking selected products to an existing category. The plan is durable: teams can inspect its status, pause or resume it, retry an eligible failed step, or cancel the remaining work.
Teams can choose approval mode, bounded autonomous execution, or a hybrid workflow. Autonomous still requires a deliberate authorization of the exact saved plan and product scope. Paid AI changes have a stronger rule in every mode: approve proposal generation, review the exact proposal, then separately approve applying it. The task agent does not get arbitrary database, browser, publishing, or delete tools.
This is an agentic approach grounded in a structured PIM: the model proposes work through a closed action catalog; the product enforces scope and workflow; people retain review over generated changes.
AI actions in the product workflow
AI enrichment, title and description edits, translations, image generation, and category actions are becoming part of the same product workflow. Where enabled, teams can start with a single product or select products for bulk work. Generated copy is previewed before it is applied; image generation previews prompts, image counts, and costs before it is authorized.
The objective is not to replace product expertise. It is to reduce repetitive work while keeping product claims, terminology, imagery, and channel requirements reviewable by the team.
One organized library for product media
A folder-based media library makes the configured object-storage structure navigable from the product workspace. Teams can browse files such as images, videos, and PDFs, then select existing assets for one product or a bulk product action rather than repeatedly uploading the same media.
Reporting that connects channel delivery with sales
The Reporting module can export date-ranged feed-delivery activity into an Excel workbook, with channel and feed summaries, delivery detail, and a product-sales sheet when sales data exists. Reports are stored in the configured private object-storage bucket. They describe events recorded by the system; they do not reconstruct historical feed runs from before event capture began.
Teams can also schedule product-sales imports from connected WooCommerce, Shopify, Magento, or Amazon stores, or map a custom HTTPS JSON API. Schedules run from hourly to every seven days and can be run on demand. Source coverage and credentials determine which records are available.
A clearer view of AI usage costs
Billing now separates monthly spend for the in-app assistant, the task agent, and other AI usage while preserving the existing total. The view groups recorded usage; it does not duplicate metering or charge an event a second time.