Client work · Managed AI agents

Years of member-facing engineering for Coop Sweden

Coop Sweden · Retail — 3 million+ member cooperative · Multi-year relationship with our sister brand ebi.ai

The client

Coop Sweden is one of Sweden’s biggest grocery retailers — a cooperative owned by its 3 million+ members, whose membership accounts, rewards and payments sit at the centre of the business.

Mach1: the membership master system

Beneath the member experience sits Mach1, Coop’s membership master system — the platform handling member accounts and the financial machinery around them. EBI’s work spanned the integration architecture connecting Mach1 to the surrounding estate, requirement mapping across the account and event flows, and a multi-year support arrangement keeping it healthy. Banking-adjacent member data, at population scale — engineered and supported quietly.

Cooper: the assistant three million members can ask

Through our sister brand ebi.ai, Coop deployed Cooper — an AI assistant with its own persona, roadmap and app delivery plan, answering member queries that would otherwise queue for the contact centre. Publicly reported results put the assistant at 91% of queries answered across millions of shoppers. Bought not as software for Coop to operate, but as an outcome delivered and improved continuously — Service as Software before the term existed.

The covid-era footfall system

When the pandemic hit, the same engineering relationship produced something nobody had specified in advance: computer vision tracking how many people were inside each store, surfaced through the assistant so members could ask the best time to shop safely. Trend data accumulated automatically, letting the grocer predict busy periods store by store.

Why it matters if you’re reading this

One client, three different shapes of work — a core data platform, a managed AI service, and a rapid-response innovation — across a relationship measured in years. That combination (data foundations + agents + speed) is exactly the model EBI now offers as standard.

Millions of customers, one back end? We've run that errand before.