James Reed
Founder & CTO, WYRM
Builds agentic systems for procurement, tenders, OSINT, contracts, and the cross-domain intelligence layer underneath them.
About
James Reed is the founder and technical lead of WYRM, the agentic engineering and operating-intelligence platform. He architected the Jörmungandr decision orchestrator that powers every WYRM product line — the flagship engineering products WYRM MEP and WYRM Data, the Procure, Ledger, Cyber and Legal add-on modules, and the standalone WYRM Healthcare line — along with the public REST + MCP API surface and the live-feed pipeline behind sanctions, CBAM, FX, tender, OSINT, and contract-review capabilities.
Before WYRM, James spent a decade as a Python systems engineer working on real-time data pipelines, geospatial analysis, and cross-domain OSINT tooling. His writing focuses on the parts of agentic AI that matter for production buyers — confidence ensembles, dissent surfacing, audit trails, and the boundary conditions where multi-agent systems break.
He posts working code more often than he posts opinions, and prefers showing the agent trace to claiming the agent works.
Recent posts
- MEP
Why we built WYRM MEP
The drain in a building-services practice is in the deliverables, not the design. WYRM MEP is an agentic studio that sits on top of the tools you already own, does the production legwork, and gives your engineers their week back — checked by two QA gates, signed by a human.
2026-06-09 · 8 min read - MEP
Why WYRM MEP never lets a language model size a cable
A general-purpose model that is usually right becomes a liability the moment 'usually' turns into a mis-sized cable. WYRM splits the work: agents read, plan and draft; algorithmic engines do the maths; two QA gates check the join. Bounded, visible failure instead of unbounded, silent error.
2026-06-10 · 9 min read - MEP
The change cascade: MEP coordination is a graph problem
The expensive failure in MEP is rarely the first design — it's the change that ripples through six documents and gets missed in one. WYRM MEP keeps the whole job connected so a single change cascades correctly across every discipline, batched into a digest, never a stream of pings.
2026-06-11 · 8 min read - MEP
From brief to deliverables: the WYRM MEP workspace
Drop the ERs on the Project tab and WYRM extracts the design constraints, locks them against their source paragraphs, and runs every downstream tab — sizing, EPC, deliverables — from that single spec. Spec, schedule and drawing stay reconciled by construction, not by hand.
2026-06-12 · 8 min read - MEP
"Can I swap this for that?" — the MEP QA Advisor
Most MEP coordination errors start with an innocent substitution made under time pressure. The £20 MEP Advisor tier is built around that single question — swap this fan coil for that 21 kW unit — and answers it against the standards, with the downstream knock-ons surfaced before you commit.
2026-06-13 · 6 min read - Company
Why We Built WYRM Procure
Procurement teams lose hours every week to tab-switching between sanctions portals, commodity feeds, FX rates, and supplier directories. We built WYRM Procure — an agentic AI platform — because we believe the answer should come to the buyer, not the other way around.
2026-03-18 · 6 min read - Technology
Agentic AI for Procurement, Explained
Agentic AI procurement is not a dashboard. It is a set of specialist agents operating in parallel across sanctions, FX, commodities, shipping, carbon, and supplier data, returning one ranked procurement answer with full evidence trail.
2026-03-20 · 8 min read - Engineering
Building the Agentic Procurement Pipeline
Seven specialised agentic workers run simultaneously every time a buyer searches a product. We fuse their outputs — sanctions risk, CO₂ liability, FX exposure, commodity forecast, shipping cost, supplier capacity — through a predictive scoring model and return the ranked answer. Here's the architecture.
2026-03-25 · 12 min read - Technology
Agentic AI vs RAG vs Copilots: What Procurement Actually Needs
RAG answers questions. Copilots summarise screens. Procurement needs something that reasons across live markets, sanctions, and freight — then shows its working.
2026-04-22 · 10 min read - Compliance
Agentic AI for CBAM Category Managers: A Field Guide
CBAM liability calculated in spreadsheets after the PO is signed is a compliance risk. Agentic AI moves the calculation to the moment of decision.
2026-04-24 · 12 min read - Public Sector
Agentic AI for UK Government Procurement: Act 2023 Alignment
The Procurement Act 2023 asks buyers to justify supplier decisions on record. Agentic AI makes the justification a by-product of the decision itself.
2026-04-26 · 10 min read - Ledger
Agentic AI for bid + tender intelligence
Tender platforms aggregate notices. They do not tell you which to bid on, and they certainly do not draft the response. WYRM Ledger does both — capability-match against win history, then a structured draft aligned with the awarding authority's evaluation criteria.
2026-05-08 · 8 min read - Cyber
CVE triage at scale — why EPSS + KEV beat CVSS alone
CVSS is theoretical severity. EPSS is real-world exploit probability. KEV is observed exploitation. Use all three — weighted correctly against the operator's footprint — and the weekly action list drops from hundreds to single digits.
2026-05-09 · 9 min read - Legal
Why clause playbooks beat one-off contract review
Contract review without a documented playbook is opinion. With one, it is policy enforcement. WYRM Legal turns your negotiated baselines into a comparison engine — every incoming contract gets the same treatment, every time.
2026-05-10 · 8 min read - Data
Entity resolution is the cross-module substrate
A sanctioned supplier, a CVE-exposed processor, and a counterparty in a DPA are often the same legal entity wearing three names. Entity resolution is what lets a finding in one WYRM module trigger the right action in the others.
2026-05-11 · 9 min read