A Platform's Worth of EAM Workflows in Seven Tools

The problem

Commercial EAM/CMMS platforms sell asset registry, condition monitoring, predictive maintenance, TCO reporting, compliance tracking, field dispatch, and capital planning as per-seat licenses. A 20-technician team on a mid-market Premium tier pays about $13,200 a year for that module list. The question this build set out to answer: how much engineering does that module list actually require, and what does it cost to run once built?

The build

Seven Python tools behind one language model: asset query, health analysis, failure prediction with composite risk scoring, TCO calculation, compliance tracking, field route optimization, and Monte Carlo capital planning with 1,000-iteration convergence. Orchestration is standard LangChain tool binding at temperature 0. The architecture is deliberately boring, and that is the point: the module list of a licensed platform fits in a reasoning layer, a tool layer, and an orchestration layer.

The AI-assisted demo build took roughly 40 to 80 engineer-hours, verifiable in the repository's commit history.

The measurements

• 59 of 59 unit and integration tests passing (37 tool tests, 22 capital-planning tests)
• End-to-end latency: 1.35 s single-tool, 8.70 s multi-tool (measured early 2026)
• Cost per query on the certified model tier: $0.0030 (AEQ Grid certification run, July 2026)
• The certified $1/MTok tier matched a $5/MTok frontier model 12 cells to 12 on the workload's non-trap query classes

Pricing note (added 2026-08-31): the certified-tier prices behind the $0.0030 figure were verified 2026-07-24 for the July 2026 certification run. On 2026-08-07 the vendor repriced the certified tier to $0.20 per million input tokens and $1.20 per million output tokens, roughly a 5x cut, so the run-date figure stands as an upper bound. The white paper's dated pricing postscript carries the full accounting.

The honest limits

The demo runs on a 50-asset synthetic portfolio, not a live customer deployment, and the route optimizer was measured against statistical simulation rather than a live road network. Total cost of ownership matters: with build and maintenance labor counted, the agent stack does not beat the per-seat license below roughly 130 seats. Small teams should keep their SaaS. The full accounting, including the assumptions an operator should replace with their own, is in the white paper.

Version history

Read the engineering

The complete definitions, measurement protocol, and methodology: Agentic Architecture for Enterprise Asset Management (technical reference, v2.1.1).