Economics v0.1

Charles Dana · Monce AI · April 2026

The Equation

Cost(PO) = CVLM + CSnake + Cinfra

For the current Bedrock on-demand pricing (eu-west-3, Apr 2026):

ComponentCallTokens (typ.)$/PO
Stage 0 (Client ID)Haiku 4.5 × 11k in / 0.1k out$0.0013
Stage 1 (Doc Analyzer)Haiku 4.5 × 12k in / 0.3k out$0.0031
Stage 2 (Extractor)Sonnet 4.6 × 1–34k in / 2k out$0.042
Stage 3 (Rules)local$0
Stage 4 (Snake × 9.4k-label basis)local$0
Stage 5 (Validation)Haiku 4.5 × 13k in / 0.5k out$0.0044
Stage 6 (Router)local$0
Total LLM$0.051

A typical Ariba PO (Verizon, Satair) with 2–6 line items costs $0.05 in LLM fees. Multi-page POs with 20+ line items over 4–5 pages stay < $0.15.

Matching leverage

Stage 4 (Snake) is free (< 10 ms/line, in-process) but it resolves the heart of the problem: mapping heterogeneous manufacturer part IDs to Saft's 9,425-article master data. Without it, every PO is a manual lookup.

Manual (today)Monce pipeline
Avg time per PO8–15 min (data entry + SKU lookup)15 s automated + 30 s spot check
SKU match rate≈ 100% (operator does it)85–95% on seen families, audited
Error rate (wrong SKU entered)~2% (typos, wrong family)< 0.2% (Snake + fuzzy cross-check)
Share routed to auto-approve0%~70% target once stable

Break-even vs manual entry

Assume a fully-loaded cost of $45/h for a Valdosta planning clerk. Manual PO ingestion averages ~12 min → $9.00 per PO. The pipeline runs at $0.05.

Saving per auto-approved PO ≈ $8.95

At 30 POs/day for Valdosta (Verizon + Satair + direct customers), that's $268/day in recovered operator time, or ~$67k/year gross, before fixed infrastructure cost.

Fixed infrastructure

ResourceMonthly
EC2 t3.medium (eu-west-3)$30
EBS 20 GB gp3$2
Route53 + CloudWatch + data out$4
Total$36/month

Volume scaling

 10 POs/day  → $15/mo LLM + $36 infra = $51/mo
 30 POs/day  → $46/mo LLM + $36 infra = $82/mo
100 POs/day  → $153/mo LLM + $36 infra = $189/mo

t3.medium handles 3 concurrent workers — ~20 POs/minute sustained on a typical Ariba/Satair layout. The bottleneck is Bedrock throughput, not CPU. Snake scales linearly in |M|: the 9,425-label model trains in ~4 min on bucket=250, L=15, and predicts in < 10 ms/query.

Depth knob

The model_mode parameter on /extract trades cost for accuracy:

modestage 0/1/5stage 2$/PO
cheapHaikuHaiku$0.01
balanced (default)HaikuSonnet$0.05
accurateSonnetSonnet$0.12
VLM-dollars = accuracy-dollars

Cheapest first. Escalate only when Snake's top-1 confidence is below θauto.