Illustrative engagement · representative of a 90-day Department Build · all names fictional
Department 03 · Logistics Ops

How regional CPG ops stops the war room.

What a forward-deployed engineering engagement looks like inside a $720M regional packaged-foods company. Synthetic client (Vermillion Foods, natural-ingredient brand with 6 plants and 7 DCs), real methodology, real artifacts. Eight agents shipped to production across SAP S/4HANA + Manhattan SCALE + o9. The team stopped chasing exceptions and started doing strategic sourcing.

Client archetype
$720M revenue · regional CPG · natural-ingredient packaged foods
Operations org
220 people · 6 plants · 7 DCs · 240 SKUs · 180 suppliers
Engagement window
90 days · 3w discovery · 12w build
Year-one ROI envelope
$2.0–3.5M · methodology below
Chapter 01

The premise

Vermillion Foods had migrated SAP from ECC to S/4HANA the prior year, deployed Manhattan SCALE for warehouse ops, and rolled out o9 Solutions for planning. The systems were modern. The work wasn’t. Coordinators spent their week chasing supplier exceptions through Slack and email. War rooms triggered for every Type-12 disruption. The COO wanted OTIF up to 97%, stockout rate down to 2%, and the team out of the war-room business and into strategic sourcing. Three weeks in, we had the picture below.

Systems detected & mapped
SAP S/4HANAconnected
ERP (post-2024 ECC migration)
OData + IDoc
Manhattan SCALEconnected
Warehouse Mgmt (7 DCs)
REST
o9 Solutionsconnected
Demand & Supply Planning
REST + cube export
Coupaconnected
Procurement & Supplier Mgmt
REST + Webhook
Project44connected
Transport visibility
REST + EDI 214
Snowflakeconnected
Data warehouse
JDBC
Power BIconnected
Reporting & dashboards
Direct query
NielsenIQ + IRIconnected
POS & market data
Weekly file delivery
8
systems mapped
31
distinct exception patterns documented
7
DCs · 5 owned · 2 3PL
180
suppliers · 240 SKUs · ~6,400 SKU-DC lanes
Where time was actually leaking

Heatmap from time-and-motion shadowing of 16 ops professionals across two weeks (coordinators, planners, buyers, warehouse leads, quality analysts). Exception chasing consumed nearly a third of the org’s capacity — most of it work that the systems should have absorbed automatically.

Exception chasing (Slack threads, supplier calls, war rooms)
31%
Manual demand forecast adjustments & overrides
22%
Supplier scorecard prep & QBR documentation
14%
PO amendment routing & price-escalation negotiation
12%
Inventory rebalancing decisions & inter-DC transfer planning
10%
Quality hold review & COA documentation
6%
Returns disposition & supplier chargeback admin
5%
Opportunity map · prioritized by ROI

Week-three deliverable: which agent to build first and why. P1 + P2 are the foundation — exception routing and demand sensing are how the war room shuts down. Everything else flows from those.

P1
Exception Routing + Inbound Visibility
$0.5–0.8M year-one·88% automatable·6w build
P2
Demand Sensing (POS + competitor signals)
$0.4–0.7M year-one·94% automatable·8w build
P3
Inventory Rebalancing (across 7 DCs)
$0.3–0.5M year-one·86% automatable·6w build
P4
Supplier Scorecard + PO Amendment
$0.3–0.4M year-one·82% automatable·6w build
P5
Quality Pattern Detection
$0.2–0.4M year-one·78% automatable·8w build
P6
Returns Disposition Automation
$0.2–0.3M year-one·88% automatable·4w build
P7
Cross-DC Transportation Optimization
$0.1–0.4M year-one·80% automatable·4w build
What was surprising

The systems weren’t the constraint. SAP, Manhattan, and o9 were all modern, all integrated. The constraint was the human-in-the-loop logic that connected them. Every Type-12 supplier delay required a coordinator to: check the DC’s safety stock manually, calculate transfer cost vs lost-revenue manually, find a backup supplier’s capacity manually, draft a customer ETA update manually. The agent doesn’t replace any of those systems — it sits between them and absorbs the work that humans had to do because nobody had built the connective tissue. Five exception types covered 71% of weekly war-room volume. Solve those five, the war room mostly disappears.

Chapter 02

Tribal knowledge becomes executable

Six plants, 7 DCs, 180 suppliers, 240 SKUs. Half of how Vermillion’s ops team actually runs production was in nobody’s SOP — it lived in the heads of the senior coordinators who had been there through every supplier transition, every dock-constraint workaround, every weather-driven inventory build.

Documented rules · extracted from existing playbook

Vermillion had a 47-page operations playbook in Confluence. The agent extracted 12 executable rules from it. Six of the most-cited shown below.

Extracted rules12 rules · 6 shown
01Supplier delay >48 hours auto-escalates to Procurement Lead with alternative sourcing options97%
02Disruption events impacting >$500k revenue trigger COO escalation within 2 hours98%
03PO amendments exceeding 15% over base PO require director-level approval per MSA-2201896%
04Supplier on-time delivery below 85% for 3 consecutive months triggers performance review94%
05Warehouse allocation recalculates when inbound shipment ETA changes by >12 hours95%
06Demand signal weighting: POS data 40%, retailer orders 30%, seasonal index 20%, promotional calendar 10%93%
Unwritten rules · captured from interviews

What lived in people’s heads. Captured through structured interviews with senior coordinators, planners, the procurement bench, the QA team, and the regional warehouse leads. Each rule has a named source so it can be revisited as the org changes.

Pacific Components always ships 2-3 days late on orders over $200k. Build buffer into timeline; don't rely on quoted lead times.

Source · Senior Supply Chain CoordinatorSupplier-specific

DC-7 (Atlanta) has loading dock constraints that require staggered inbound scheduling on Mondays — too many trucks book the 8am slot otherwise.

Source · Warehouse Ops LeadDC-specific

Q4 holiday season: double safety stock on top 50 SKUs starting October 1, not the policy date of November 1. Two years of stockouts taught us this.

Source · Demand Planning AnalystSeasonal pattern

When o9 forecast disagrees with sales team input by >20%, always use the higher number for top 10 accounts. Sales has direct customer signal we don’t.

Source · Director of OperationsKnowledge concentration

DC-3 (Tampa) always carries 12% extra of hurricane-zone SKUs in summer months — not in any official document, but every coordinator knows.

Source · Senior CoordinatorRegional pattern

Mark at procurement handles all Coastal Co-Pack exceptions because of timezone and relationship context. Don’t auto-route those.

Source · Procurement LeadKnowledge concentration
Agent configuration · per-workflow confidence thresholds

Every agent ships with an explicit confidence threshold. Below it, the agent escalates to a named human; never silently fails. Quality-related decisions ship with stricter thresholds because of recall risk.

Demand Sensingthreshold 90%
94% auto-resolved6% escalated to human
Exception Routingthreshold 85%
81% auto-resolved19% escalated to human
Inventory Rebalancingthreshold 88%
86% auto-resolved14% escalated to human
PO Amendmentthreshold 95%
73% auto-resolved27% escalated to human
Quality Pattern Detectionthreshold 90%
78% auto-resolved22% escalated to human
Returns Dispositionthreshold 90%
88% auto-resolved12% escalated to human
Why thresholds, not certainties

In CPG, a wrong agent decision can mean a stockout that loses a Walmart slot, a quality issue that becomes a recall, or a supplier dispute that breaks a relationship. The whole stack is built around the agent declaring uncertainty rather than guessing. Below threshold, work routes to a named human with the full reasoning trace attached. Quality and recall-risk patterns trigger the highest thresholds; routing decisions can run leaner.

Chapter 03

Day-in-the-life

Three perspectives on the same Monday morning, three weeks after deployment. The Coordinator saw a 4-item queue (down from 30+). The Director Ops saw a 220-person team operating at 300-person effective capacity. The COO saw the weekly ops cycle at Day 3 of 5, on track.

Rachel Torres, Senior Supply Chain Coordinator · DC-4 opened her queue Monday morning and saw four items requiring human judgment. Down from 30+ before the agents shipped. Everything else was handled overnight across 7 DCs.

What agents handled overnight
34
Exceptions triaged
127
POs validated
89
Shipments tracked
7
Supplier alerts processed
Before deployment, this took ~7 hours/day for the coordinator team. After deployment: 30 minutes spent reviewing the queue below.
Queue snapshot · 4 items required human judgment
Sorted by revenue impact · agent recommendations attached
EXC-2847high
Pacific Components 3-day delay · DC-4 stockout risk
340 units short. Reroute from DC-7 modeled (cost $4.2k, prevents $159k revenue+penalty exposure). ETA 18h.
91%
POA-1247high
Apex Raw Materials · 18% price escalation request
Above auto-tolerance. Counter at 13% recommended per NielsenIQ commodity index. 2 backup suppliers at +11% premium identified.
88%
QPD-1018medium
GlobalPack Industries · seal-integrity defect trend
0.3pp/week increase 6 weeks. Equipment-change root cause 87% confidence. Pre-emptive inspection active. 4-week earlier vs manual.
94%
DFC-0847medium
SE region demand anomaly · 3 SKU groups
+30% above forecast. Likely competitor recall transfer. Recommend +22% replenishment for 14 days. Pending Demand Director approval.
84%
Weekly time saved
Hours/week previously spent on exception chasing · now reclaimed for sourcing strategy
+22h
Mon
+28h
Tue
+19h
Wed
+31h
Thu
+38h
Fri
Chapter 04

Under the hood — how a supplier disruption moves

A supplier disruption arrives. The Main Operations Agent spawns five specialized sub-agents in parallel — each assembling different context, checking different thresholds, surfacing different decisions. The orchestrator synthesises a recommendation. The team gets the full plan; they decide whether to approve.

Incoming disruption · Type 12 supplier delay
Pacific Components Ltd
EXC-2847 · Apr 29 2026 · Detected 08:14 ET via Project44
340 units short
Shipment TRK-44821 — 3 days late
ETA: May 2
Affected SKUs: HPC-201, HPC-204, HPC-209
12 orders
DC-4 (Chicago) — stockout risk 87%
Critical
Main Operations Agent
Orchestrates 5 sub-agents in parallel
spawns sub-agents →
Context Assembly Agent1.4s98%

Assembled context: 340 units short, 12 customer orders affected, DC-4 safety stock at 15%, DC-7 has 420 surplus units, backup supplier Meridian has capacity.

Exception Classification Agent0.9s96%

Classified as Type 12: Supplier Delay — Partial Shortfall. Severity: High. Matches repeatable pattern (1 of 31 automatable types).

Rule 7: Stockout risk >80% triggers automatic safety stock replenishment from nearest DC
Impact Assessment Agent1.4s94%

Revenue at risk: $127k across 12 orders. 3 orders have delivery commitments within 48 hours. 2 key accounts affected. Alternative sourcing available.

Rule 2: Disruption impacting >$500k triggers war-room escalation within 2 hours
!Resolution Planning Agent2.1s91%

Recommended: Reroute 280 units from DC-7 safety stock + expedite 60 from Meridian Parts. Cost: $4,200. Alternative: wait 3 days, risk $127k revenue. Recommend: approve reroute.

Rule 10: Expedited shipping requires VP approval if cost exceeds $25k per incident
Execution Agent

Draft actions assembled: DC-7 → DC-4 transfer order, Meridian Parts expedite PO, Pacific Components delay notification, customer ETA updates. Awaiting approval.

The feedback loop

When a human corrects the agent’s recommendation — say, choosing a partial-reroute split-source over a full reroute — the correction is captured. The next similar disruption pattern gets the corrected playbook automatically.

Exception Routing · last week’s correction
Agent planned
Auto-Reroute
Reroute all 340 units from DC-7
Human corrected
Split-Source
Reroute 280 from DC-7 + 60 expedited from Meridian
Exception routing rule updated. Future Type 12 exceptions with partial availability will evaluate split-sourcing before full reroute. Reduces DC-7 safety-stock impact and creates supplier optionality.
Before correction
85%
After correction
94%
Change detection · auto-absorbed during Q1

Three real ops changes from Q1 — supplier lead-time shifts, demand pattern adjustments, new supplier onboarding. Each detected automatically and absorbed without manual reconfiguration of routing rules.

Jan 15, 2026
Supplier lead time change: Pacific Components

Average lead time increased from 14 to 18 days. Safety stock calculations and PO timing rules auto-adjusted across affected SKUs.

2 rules auto-updated
Jan 28, 2026
Demand forecast adjustment: Home care category

POS data showed 22% above forecast for 3 consecutive weeks. Demand sensing weights recalibrated. Pre-build triggered at DC-3 and DC-5.

1 rule auto-updated
Feb 1, 2026
New supplier onboarded: Meridian Parts

Backup supplier added to exception routing options. Lead time, quality standards, and capacity constraints integrated into resolution planning.

3 rules auto-updated
Chapter 05

What changed

Month 1 vs Month 4 at Vermillion Foods. The agents got smarter, disruption response got faster, the team shifted from war-room exception chasing to strategic sourcing.

Before & after · eight measurable outcomes
Metric
Before
After
Δ
Disruption response time
2-4 days
<6 hours
−95%
Stockout rate · top 50 SKUs
6.8%
2.4%
−65%
SKU forecast accuracy (4w-out)
71%
86%
+15pp
PO amendment rework rate
23%
4%
−83%
Inventory carrying costs
baseline
−19%
$9.3M release
Returns disposition cycle
5-7 days
Same day
−95%
OTIF (4-week rolling)
91.2%
97.2%
+6pp
Supplier QBR cadence
Quarterly manual
Continuous
auto-generated
The accuracy curve

Agents get smarter every week. Human corrections and SOP changes are absorbed automatically. Overall accuracy lifted from 85% in week 1 to 96.8% by week 16.

100%90%80%70%Week 1Week 4Week 8Week 12Week 16Feedback loop activated · week 696.8%
85% week 189% week 6 (feedback loop)94% week 1296.8% week 16
Full audit trail · every action, timestamped, traceable

Every agent action with timestamp, reasoning, confidence, and human approvals. Searchable. Filterable. Used for retailer compliance audits, supplier QBR documentation, and quarterly board reporting.

08:14:22
Exception RoutingEscalated to human
Pacific Components 3-day delay flagged. DC-4 stockout risk 87%. Reroute decision routed to Director Ops.
91%
08:15:48
Demand SensingEscalated to human
SE region 30%+ deviation across 3 SKU groups. Likely competitor recall transfer. +22% replenishment recommended.
84%
08:16:31
Supplier ScorecardAuto-processed
Q1 scorecards generated for 180 suppliers. 8 flagged off-contract. QBR meetings auto-scheduled.
99%
08:18:14
Inventory RebalancingAuto-processed
DC-4 stockout risk: 4 inter-DC transfers initiated. 280 units. Cost $4.2k prevents $159k revenue+penalty.
96%
08:19:48
PO AmendmentEscalated to human
Apex Raw Materials 18% price escalation. Above auto-tolerance. Counter at 13% recommended; 2 backup suppliers identified.
88%
08:20:55
Quality PatternEscalated to human
GlobalPack defect trend 0.3pp/week × 6 weeks. Equipment-change root cause. Pre-emptive inspection active.
94%
08:22:09
Inbound VisibilityAuto-processed
412 inbound shipments tracked. 14 with >12h delays. 3 critical-path routed. 8 absorbed by buffer. 6 informational only.
95%
08:23:41
Returns DispositionAuto-processed
64 returns dispositioned: 41 restock, 14 supplier-recover, 7 scrap, 2 refurb. 7 routed to Returns Specialist.
89%
They own all of this

Vermillion owns the agents, the data, the rules, the methodology. We did the work; they keep everything.

The client owns the agents

Every workflow, every rule, every model. Deployed on their infrastructure, inside their VPC, within their security perimeter.

Data never leaves their environment

Processing happens in their environment. No supplier or operations data sent to external servers. Full compliance with their security and retailer-data policies.

Walk away anytime

Zero vendor lock-in. They keep everything if the engagement ends. The IP is in the methodology, not the output.

Like hiring an architect

They own the building. We designed and built it. The blueprints, the structure, the systems. All theirs.

Working capital impact

$2.0–3.5M year one.

Range, not point estimate. CPG CFOs read these numbers carefully and they need to survive board scrutiny. Below is how the value gets created — each line tied to a specific agent and a specific measurable outcome.

$0.6–1.0M
Working capital release

Inventory carrying costs down 19%. Optimized safety stock levels across 7 DCs based on better demand sensing. $9.3M working capital released; conservative carry-cost computation at 7-11% yields the range.

Driven by · Agents 01, 05
$0.5–0.8M
Stockout avoidance

Stockout rate on top-50 SKUs from 6.8% to 2.4%. Revenue not lost. Plus: zero Walmart OTIF penalty exposure this quarter (vs $48k Q1 prior year). Conservative — only counts top-50 SKUs.

Driven by · Agents 01, 02, 05, 06
$0.6–0.9M
Headcount redeployment

14 coordinator FTE-equivalents previously running 35-50% on exception chasing have been redeployed to supplier development, strategic sourcing, and cross-DC optimization. No layoffs — capacity moved up the value chain.

Driven by · Agents 02, 04, 06, 08
$0.3–0.8M
Disruption-cost avoidance

Disruption response time 2-4 days → <6 hours. War room reduced from 90 min/12 people to 25 min/5 people. Proactive Type-12 handling saves expedited-freight costs and prevents retailer penalty exposure.

Driven by · Agents 02, 03, 07
Methodology footnotes

All baselines are pre-engagement (the prior fiscal quarter at Vermillion). Working capital release computed at 9% blended carry cost (warehouse + opportunity cost). Stockout avoidance uses contribution-margin-weighted lost-sales model on top-50 SKUs only (conservative; long-tail SKU losses excluded). Headcount redeployment uses fully-loaded comp ($85k median for senior coordinators benchmarked against 2025 ASCM data). Range exists because cohort sizes are still small (n=2 quarters of post-engagement data). We never bill more than the lower bound of created value.

What we don’t claim

What agents don’t do well.

Especially in supply chain — where one quality issue can become a product recall and one missed retailer commitment can lose a slot — knowing the limits is the only way to deploy something that survives reality.

  • Replacing supply-chain judgment.Agents don’t replace the Director of Operations. The decisions where supplier relationship history, regulatory shifts, or strategic supplier portfolio decisions matter still belong to humans.
  • Inventing strategy.Agents are excellent at applying known-good rules. They don’t set sourcing strategy, choose which supplier base to consolidate, or decide which retailers to prioritize for OTIF compliance. Those are COO-level decisions; the agent surfaces evidence to support them.
  • Adapting to truly novel disruptions without help.A pandemic, a port strike, a geopolitical shift, a new regulation — the agent flags and escalates the first time. The second time, after a human handles it, the rule is captured. Initial novel-event handling is always slower than steady-state.
  • Quality decisions without human signoff.Recall risk decisions, quality holds, regulatory submissions — these always require human attestation. The agent prepares the evidence pack; QA Director or COO makes the final call.
  • Operating without instrumentation.Every agent ships with eval gates and audit-trail instrumentation. Drift is real; if the org isn’t willing to maintain the operating budget for evaluation, the agents drift.
Continuation

What ships next at Vermillion.

Quarter two: scope expansion to S&OP automation, retailer order-quality scoring, and co-packer capacity optimization. Quarter three: BOT (build-operate-transfer) optionality. The methodology is portable; the agents are theirs.