How HeadChannel Built an AI Tender Finder for Cohaesus Group

Cohaesus Group is a collective of specialist agencies, founded in 2008. As the group grew, managing tenders across its agencies became complicated.

The Challenge

Tenders were spread across sources such as Contracts Finder, TED Europe and Stotles, with nowhere to bring them together. Agencies duplicated effort, missed deadlines and lost opportunities.

Qualifying each tender took careful review. With so many services across the group, it was hard to tell which agency should respond, and the answer depended on domain knowledge and past work that wasn't easy to find. Coordination ran through informal Slack messages, and the group sometimes chased tenders it had little chance of winning while missing better ones.

The Approach

HeadChannel built a proof of concept on n8n, a low-code automation platform, so it could plug into the tools Cohaesus already used.

• One feed. Tenderiser checks Contracts Finder, TED Europe and Stotles on a schedule, extracts each new tender's details and documents, and keeps them in one place.
• AI analysis. OpenAI and Anthropic models read each tender and pick out the service area, sector, budget range, timeline and requirements.
• Scoring for each agency. Each agency uploads a knowledge base of its services, focus areas, past projects and preferred bid types. Tenderiser scores every tender against it for domain fit, complexity, estimated effort and strategic value.
• Alerts and a shared board. When a tender passes an agency's threshold, Slack notifies the people concerned and the tender goes onto a shared Trello board.

Google Workspace handles the documents. One developer and one product owner delivered it in three weeks, with input from each agency.

The Outcome

Manual qualification time fell by more than 80%. Triage that took hours of discussion across agencies now takes minutes: relevant tenders surface straight away and the rest are filtered out without manual review. Each agency sees only the tenders that suit it, and the shared board has cut duplication.

Stakeholders praised the interface and the accuracy of the scores. Plans for the next phase include user feedback to refine scoring, AI-drafted responses based on past bids, compliance checks against company documents such as ISO certificates, Google Gemini as a further model, and PGVector storage.