01Business problem
The group wants better sourcing conditions for its own-brand range through scale: products that are sold across all its brands are increasingly sourced at European level, while local sourcing teams keep buying the products that are specific to one brand. At either level, a tender with one supplier in it leaves little to negotiate with. Yet only a third of the group's own-brand products had an approved alternative supplier.
02Intervention
One team had built a first version of a research agent in one of the local markets. In a joint team with another consultancy, we took the idea to European scale in a deliberately narrow proof of concept, scoped with the sourcing team: a research agent on external sources, a report generator and a draft outreach email. The sourcing managers set the bar before the first search: what a shortlist must contain (name, country, product categories, packaging format, certifications, minimum order quantity, company size, website), which geography counts and the one test that mattered most, whether the agent finds at least one credible supplier they did not know yet.
03Role of data & AI
A supervisor agent splits the sourcing brief into parallel research tasks. Researcher agents search the web in the local language and in English, starting from trade-fair directories and public supplier lists, and check each candidate for certifications, contacts, capacity and location. Every claim is verified before it reaches the report. The shortlist scores each supplier on the criteria a sourcing manager needs.
04Adoption
Sourcing managers and supplier category managers were involved from the first prompt and score every output, in a loop of co-designing prompts, defining the bar, running the searches, scoring and refining. The agent drafts the outreach emails; a person reviews and sends them. The proof of concept is now being scoped into a minimum viable product, built into the sourcing workflow from the moment a supplier needs replacing to the signed contract, with usage tracked per category from day one.
05Measurable impact
The proof of concept passed its success criteria: nine in ten proposed suppliers matched the brief, and every shortlist carried the required fields and closed with next steps. The business case is a coverage number. In one local market, tenders with an approved alternative supplier delivered 2.4% additional savings on average. Raising coverage from 33% to 40% of own-brand products, the first target, brings that saving to every tender that gains an alternative. The European saving is not measured yet; the minimum viable product will track it.
06Time to value
Two months from scope alignment to a validated proof of concept.