Building operational AI agents
Dema appears to be expanding its AI‑agent platform from a primarily analytical tool toward a more operational workflow layer for larger commerce organisations. The moves suggest a focus on tighter integration, governance and execution capabilities across enterprise environments.
Public announcements detail new features such as automated weekly trading reports, expanded ad‑library scraping, asset transfer flows, version history with rollback, ownership controls for apps and integration of product recommendation tools.
-
Dema announced that Represent uses an agent to automate weekly trading reports, including recommended actions and publishing to Teams at 9:30, replacing a manual process that gathered data from Shopify, Excel, PowerPoint, ERP and other sources.
-
Dema announced that agents and apps can now read TikTok Ads Library, Meta Ad Library and Google Shopping results, in addition to existing web scraping, enabling daily checks of competitor ads, creative, pricing and product visibility.
-
Dema added a transfer flow for agents that considers dependent assets such as skills and apps, showing linked components and allowing users to select what to move.
-
Dema added version history and rollback for agents, apps, skills and instructions, making agent‑built workflow assets more controllable and safe for operational environments.
-
Dema described that agents now select app details when building or renaming an app, and users can change these details only when they own the app or the underlying agent.
-
Dema added Depict to the agents' integration catalog, allowing agents to work with product recommendations, site search and collection sorting in e‑commerce stores.