Graas buys Trustana: product data is what agents sell

A retail AI company just spent its Series B on a product-data platform. Singapore-based Graas raised $17 million and acquired Trustana, folding product-data enrichment into the knowledge graph that powers its retail AI agents. That's the whole thesis in one transaction: in agentic commerce, the constraint isn't the model — it's whether your catalog is good enough for an agent to sell from.

What actually happened

Per TNGlobal, the August 13 Series B was led by LemmaTree — an investment firm founded by Singapore state investor Temasek — with Integra Partners, Tin Men Capital, The Xander Group, IncredWealth and Orzon participating. Terms of the Trustana acquisition were not disclosed.

Graas, founded in 2022, says its platform processes over $1 billion in live commerce transactions, with customers including Unilever, Puma and Schneider Electric across Australia, Southeast Asia and the Gulf. Trustana, founded in 2020, automates product-information enrichment and structuring for retailers and distributors; its client list adds David Jones, Chemist Warehouse and Toys"R"Us.

The stated rationale is specific. Trustana's enrichment feeds Graas' Commerce Knowledge Graph, the database underneath its agents. CEO Prem Bhatia framed the combination of product, customer and inventory data as what lets AI agents drive sales across e-commerce and general trade, per e27.

Why product data quality matters for your business

You don't need a knowledge graph. You need the thing a knowledge graph is compensating for.

Agents fail on data, not on reasoning. An agent asked "does this fit a 2019 model" cannot infer compatibility from a marketing paragraph. It needs an attribute. We've watched this play out across Shopify's structured-catalog conversion numbers and Akeneo's agentic product-data push: the merchants winning in AI channels are the ones whose attributes are complete, typed and consistent.

Enrichment is now a funded category, which means it's a real cost. When a Temasek-backed firm pays for a product-data company as its first acquisition, that's a market signal about where the work is. If you're planning an AI commerce project, budget the data cleanup as a line item. It is usually larger than the integration.

Product plus customer plus inventory is the useful shape. The reason Bhatia names all three is that an agent recommending an out-of-stock item, or the wrong variant for a returning customer, destroys trust faster than a bad answer. If those three systems don't talk to each other in your stack, the agent layer will just surface the disconnect at higher volume.

Own the enriched output. Whatever tool does the enrichment, the structured attributes it produces belong in a system you control and can export. Enrichment work is expensive to redo. Don't let it live only inside a vendor's graph.

Sources: TNGlobal, e27.