Four procurement workflows, from scoped use case to production

Context

A global automotive group runs tendering and contracting at scale, across a large landscape of internal stakeholders who each own a different part of the process. As AI supplier through Jaden Data, I led a series of four projects embedded into those internal processes.

Business problem

Defining tender conditions, writing requests for proposals, evaluating proposals against predefined rules and negotiating contracts through iterative template exchanges between buyer and supplier consumed expert time and varied in quality from case to case.

My mandate

I designed the solution architecture and the AI processes, wrote the prompts for the workflows, and led the four projects end to end, from proof of concept through embedding into the internal processes to production use by the process owners.

Decisions

• Four projects, each a proof of concept first and then embedded into the process, each targeting one subsection (template generation, proposal evaluation, contract negotiation and related steps), each on a twelve-week cycle of six two-week sprints followed by go-live.
• A custom web interface tailored to the procurement workflows, integrated through tailored input masks and APIs, with vector databases over historical tenders and contracts for reference.
• Alignment cadence raised from bi-weekly to weekly with parallel sub-groups, so no single dependency could stall progress.

Delivery

Each project worked directly with its process owner group, who later used the application in production. The platform connected to the group’s existing AI services for compatibility and scale.

Outcome

An MVP deployed to production, with key steps of tendering and contracting automated and the process owners using the applications.

An MVP deployed to production, with key steps of tendering and contracting automated and the process owners using the applications.

Reuse

The integration patterns and the prompted workflows carried into later procurement work; the same client’s engineering teams were later onboarded to AI-assisted development with security standards and MCP practices.

Evidence

Project count, cycle length and cadence are project records; efficiency gains are the process owners’ qualitative assessment and are not quantified here.