Metcash Scales Coveo's Agentic AI Search Across Its B2B Marketplace: Conversion Up from 13% to Nearly 20%, Click Rank Down from 25 to 4 or Fewer

Metcash Rolls AI Search Across Its B2B Marketplace

Metcash scales AI across B2B e-commerce platform - retailbiz

Metcash is scaling the use of artificial intelligence (AI) across its B2B e-commerce operations to help independent businesses find products more quickly.

www.retailbiz.com.au

Australian retail trade publication RetailBiz reported in August 2026 that wholesale giant Metcash is scaling AI across its B2B e-commerce operations. Simon Williams, the company's General Manager of B2B Digital Engagement, gave the interview on the sidelines of the SAP NOW AI Tour in Sydney.

Metcash is one of Australia's largest wholesalers, supporting thousands of independent retailers under banners such as IGA, Foodland, and Mitre 10 across three divisions: supermarkets (including foodservice and convenience), liquor, and hardware. Sorted, the company's B2B marketplace, is a two-sided platform built on SAP Commerce Cloud that connects roughly 3,500 suppliers and processes around 48,000 transactions a week. With more than 75% of customers reportedly using search to discover new products, the quality of the search experience directly shapes transaction volume.

At the core of that search experience sits Coveo, the Canadian AI search and personalization vendor. Metcash officially announced its adoption of the Coveo AI-Relevance Platform on March 19, 2026, which makes this interview a progress report a few months into deployment.

From 13% to Nearly 20% Conversion: What Is Behind the Numbers

The clearest gains are showing up in the foodservice and convenience business, where the technology went live in March 2026. According to Williams, before the rollout only about 13 out of every 100 customers who came onto the platform ended up purchasing; that figure has now been lifted to close to 20. Expressed as a conversion rate, that is an improvement from roughly 13% to nearly 20%. Development began in February of the same year, meaning the shift happened in under half a year.

The change in click rank, the company's search quality metric counting how many products a customer scans before reaching the item they want, is even more dramatic. Under the old search, click rank sat between 20 and 25, so customers compared as many as 25 products before finding what they needed. Today the team treats anything past 4 as a sign of a major issue.

"

When you're a time-pressure business owner and your day is balanced between serving your customers, replenishing your stock, making orders, those seconds and those minutes really count.

Source: Simon Williams (Metcash GM of B2B Digital Engagement)

The improvement rests on semantic search that does not depend on keyword matching, combined with AI relevancy models that interpret intent. Because Sorted is used only by logged-in B2B customers, the platform knows each customer's business type and purchase history, and can rank results around what that business actually needs to operate. Compared with consumer e-commerce, where anonymous visitors dominate, the raw material for personalization is abundant from the start.

The effects go beyond faster search. Williams points to purchasing spreading into the long tail of the catalog. Customers traditionally bought only the high-volume lines on the welcome page, the chips, the Coke, the chocolate, and left. With better search they are now reaching much deeper into the catalog. Associated product recommendations, such as showing lids and containers to a customer searching for coffee cups, are contributing as well, lifting average order value and purchase frequency. Customer satisfaction scores keep rising month after month, Williams says.

That said, all of these figures are self-reported by Metcash, and no controlled test methodology or measurement conditions have been published. The conversion definition is also a B2B one, with a denominator of business customers visiting with intent to order. It differs in kind from consumer e-commerce conversion rates driven by casual browsing, so the levels cannot be compared directly.

The Coveo AI-Relevance and SAP Integration Play

According to the March announcement, Metcash integrated the Coveo AI-Relevance Platform into its SAP Commerce Cloud environment. Coveo has a partnership with SAP, and the pattern here is layering AI search and personalization on top of an SAP commerce foundation. Replacing the discovery experience without ripping out the underlying commerce stack has become a realistic option for large enterprises wary of core system overhauls.

Coveo itself has been laying agentic AI groundwork throughout 2026. In February it announced a Hosted MCP Server that lets LLMs such as ChatGPT Enterprise and Claude connect directly to its unified index, and in March it launched Conversational Product Discovery, which embeds conversational AI directly in the search experience. The latter is built on an agentic orchestration architecture that coordinates multiple AI capabilities to interpret shopper intent. In June the company was named a Leader in Gartner's Magic Quadrant for Search and Product Discovery, and its fiscal 2026 results reported that commerce drove roughly 60% of new business.

Metcash's rollout sequence is also worth studying. The technology went into the hardware business before Christmas 2025; development for foodservice and convenience began in February 2026 with deployment the following month; and the liquor business is scheduled for late 2026. Scaling division by division, verifying results at each step, illustrates a risk-contained way to embed AI into a platform that thousands of businesses rely on for daily operations.

Product Data Readiness Decides the Return on AI

What Williams emphasized in the latter half of the interview was not the AI itself but product data quality. He asks suppliers and vendors to provide accurate product information through standards platforms such as GS1. Product imagery from smaller suppliers was called out specifically as a challenge, with Williams stating that simply having an image can improve per-product conversion by up to 40%. That 40% figure is a company claim based on internal experience and has not been externally verified, but the underlying point, that basic product information work determines what search AI can deliver, applies to every e-commerce operator weighing AI search.

The same structure carries over to operators in Japan and elsewhere. For ChatGPT shopping features and agent-driven purchasing, structured product data and imagery are the preconditions for being "read" by AI. The sequence of fixing product master accuracy, image coverage, and standards-based data exchange before investing in advanced AI features holds in B2B and B2C alike. Search AI is an amplifier of product data, and if the source data is thin there is nothing to amplify. Metcash's case backs up that plain fact with numbers.

The contract value, implementation cost, and specific terms between Metcash and Coveo remain undisclosed.

Summary

Metcash's AI rollout will complete its coverage of all three divisions with the liquor deployment planned for late 2026. The next thing to watch is whether the conversion and click rank gains seen in foodservice and convenience reproduce in a different category. On Coveo's side, conversational search and MCP-based agent connectivity are waiting in the wings; once those land, Sorted moves closer to a procurement channel where AI shoulders the ordering work rather than a marketplace customers search themselves. The practical returns of agentic commerce are quietly accumulating in B2B procurement for time-pressed business owners, a step ahead of the flashier consumer showcases.