Hey everyone,
I just finished building something I’m pretty excited about and wanted to share it with the community.
The problem: I work with trades businesses (electricians, landscapers, welders, general contractors) and they all have the same pain point — the procurement process is brutal. They’re missing opportunities buried across government portals because the owner is on the tools all day. When they do find an RFP, they spend 12-20 hours per bid manually reading documents, calculating pricing, checking compliance requirements, and writing proposals from scratch. Most of that time is spent on work that can be automated.
What I built: An 8-agent AI pipeline in n8n that automates the entire contract procurement lifecycle
Some things I learned building this:
The document quality difference between Haiku and Sonnet is significant for the client-facing documents. Haiku is fine for structured data extraction and scoring, but Sonnet produces genuinely consultant-grade output for the pricing guides, compliance checklists, and proposal frameworks. Worth the extra cost for Agents 5-7.
The two-entry-point design on Agent 2 turned out to be really important. Most procurement systems only catch public portal postings, but a huge portion of valuable contracts come through private channels — direct email invitations, vendor lists, GC subcontract requests, legacy renewals. The webhook intake path handles all of those.
Running costs are low — approximately $0.55-$1.10 per opportunity through the full pipeline, so under $15/month per client.
Every workflow is fully documented — README sticky notes, configuration sticky notes, and node-level notes on every single node explaining what it does and how to customize it.
I’ve packaged this as a template that other consultants and n8n builders can import, customize for their jurisdiction, and deploy for their own clients. If you’re interested, or want more information, it’s available here: n8n Contract Procurement Workflow — 8 Agent System for Contract Discovery to Final Proposal
Happy to answer any questions about the build, the architecture decisions, or how specific agents work. This was a fun project and I learned a ton about multi-agent workflow design in n8n.