Customer Results

Real Workflows.
Measurable Outcomes.

Three deep-dives into how manufacturers replaced manual spreadsheet chaos with autonomous AI agents — and exactly what changed on the floor.

3 Verticals
MRO • Pharma • Production
< 4 wks
Avg. Time to First Value
Plug-In
Works on Your Existing Data
Natural Language
Query Interface
Industrial Manufacturing Multi-Plant MRO & Spare Parts

From Invisible Risk to
Proactive MRO Intelligence

A multi-plant manufacturer running critical production lines had no reliable way to know which spare parts would run out before the next PO arrived. The risk was invisible — until the line stopped.

4 wks
To full deployment
Zero
Unplanned line stops post-deploy

The Situation Before

  • Spare parts data spread across three separate file snapshots — inventory, last-price, and 90-day consumption — each maintained independently
  • No real-time flag for "at-risk" parts: a part was only identified as critical after a line stopped, not before
  • Price inconsistencies: inventory system showed $0 unit cost for many critical parts, making dollar valuations unreliable
  • Planners spent hours manually cross-referencing part codes across three files to answer: "what do we have and will it last?"

What ZeeHub Deployed

  • Unified data layer — auto-joins inventory, pricing, and consumption files on part code every shift, zero manual merging
  • At-risk detection engine — flags any critical part where days-of-stock-remaining is less than its lead time, before the stock-out occurs
  • Price fallback logic — when inventory shows $0 cost, automatically uses last-price file. Inventory valuations became trustworthy overnight
  • Natural language queries — floor managers ask plain-English questions and get instant, accurate answers

Live Queries the Team Now Runs Daily

"Which critical parts at our main plant will run out before the next delivery arrives?"
⚡ Found 4 at-risk critical parts — Bearing SKF-type (8 days remaining, 14-day lead time), Seal Kit A (3 days remaining, 12-day lead time)… Auto-reorder recommended for 3 of 4.
"What's our total critical parts inventory value across both plants this week?"
📊 Critical inventory: Plant 1: $512K, Plant 2: $335K. Total: $847K. 6 parts using price fallback (inventory showed $0). Non-critical: $241K.
"Show me parts with more than 180 days of stock — possible overstock?"
🔍 18 non-critical parts exceed 180-day runway. Top by value: Gasket Set ($18K, 340 days), Filter Kit ($12K, 290 days)… Recommend review for excess disposal.
"What percentage of critical parts are in stock vs. out of stock right now?"
✅ Critical in-stock: 94.2% | Out-of-stock: 5.8% (7 parts). Prior week: 91.3%. Trend improving. 3 of 7 have active POs already in transit.
94%+
Critical Parts In-Stock Rate
vs. reactive discovery only before
Zero
Unplanned Line Stoppages
due to parts shortage since go-live
3 files
Auto-Joined in Real Time
inventory • pricing • consumption
Seconds
To Answer Any Parts Query
was hours of manual work
Pharma Manufacturing Multi-CMO Supply Planning

Drug Substance Planning
That Answers “What If”

A pharma manufacturer coordinating drug substance and drug product batches across multiple CMO sites was running all supply planning in spreadsheets — with no way to model the downstream impact of demand changes or new batch decisions.

Days → Seconds
Scenario modeling time
4 CMOs
Unified DS/DP/FPP view

The Situation Before

  • Separate worksheets per CMO site with no unified pipeline view — each planner maintained their own version of truth
  • Any demand change required manually recalculating months-on-hand across every product × site combination — a multi-day effort
  • Batch addition decisions made without modeling downstream inventory impact — leading to overstock at some sites, shortfalls at others
  • Ending inventory projections were always weeks stale by the time leadership reviewed them

What ZeeHub Deployed

  • Unified supply dashboard across all CMO sites — live ending inventory, DS batches, DP batches, and months-on-hand per product and site
  • What-if scenario engine — "increase sales 15% starting Q1 2026" or "add 3 DS batches at Site A in 2027", modeled instantly
  • Natural language queries on live data — "What is ending inventory for our primary product at Site A in December 2028?" answered in seconds
  • Semantic caching for repeat planning questions — frequently asked queries return in milliseconds

What-If Scenarios the Planning Team Now Runs in Seconds

Scenario A — Demand Surge
"If sales increase 20% starting Q2, when does our primary product inventory drop below 3 months on hand at each site?"
⚡ MOH breaches 3.0 at Site A in Aug and Site B in Oct under this scenario. Recommend adding 2 DS batches at Site A in Q1 to buffer. Want me to model that addition?
Scenario B — Batch Addition
"Add 4 DP batches at our secondary CMO in 2027. Show ending inventory impact vs. baseline."
📊 Adding 4 DP batches at Site B increases ending inventory ~18% vs. baseline by Dec 2027. MOH improves from 4.2 → 5.1 months. Overstock risk low. Side-by-side chart ready.
Seconds
Scenario Modeling Time
was 2–3 days per scenario
4 CMOs
Unified in One View
DS • DP • FPP pipeline
Live
Batch & Inventory Tracking
across all products and sites
Real-Time
MOH & Ending Inventory
was always weeks stale
Process Manufacturing OEE / Uptime Maintenance Planning

Machine Uptime Tracking
That Runs Itself

A process manufacturer tracking uptime across dozens of production lines was manually reconciling maintenance execution logs against master plans every shift — calculating daily production impact by hand from maintenance codes.

Auto
Uptime calc from maintenance codes
Per-Line
OEE visibility every shift

The Situation Before

  • Two separate files — a Master Plan (scheduled maintenance) and an Execution Log (actual events) — maintained by different teams with no automated reconciliation
  • Maintenance codes each carried different hour deductions. Engineers calculated production impact manually per machine, per shift
  • When execution deviated from the master plan, actual uptime was only known the following day after a supervisor reviewed both files
  • No machine-level OEE tracking — production impact reported as a plant average, masking which lines had chronic downtime patterns

What ZeeHub Deployed

  • Execution-first reconciliation — AI merges execution log and master plan automatically, with actual execution always overriding scheduled events
  • Code-driven uptime engine — maintenance catalog maps each code (weekly service, tool swap, monthly shutdown, major stoppage) to its hour deduction; uptime % calculated per machine automatically
  • Machine-level OEE dashboard — shift supervisors see real-time uptime per line, not averages. Chronic downtime machines surface immediately
  • NL queries on maintenance data — "Which lines had more than 2 hours of unplanned stoppages this week?" answered instantly

How the Uptime Engine Works (Fully Automated)

Step 1
Ingest Both Sources
Execution log + Master plan loaded each shift automatically
Step 2
Reconcile
Execution overrides master plan. Standard shifts = 0 deduction. Heavy maintenance = hours deducted per code.
Step 3
Calculate Uptime %
Uptime = (24 − maintenance hours) ÷ 24. Per machine, per shift, per day.
Step 4
Surface on Dashboard
Machine-level OEE live for every shift supervisor. Alerts fire when lines drop below threshold.
Standard Shift
Production / No Impact
0 hrs deducted
Weekly Service
Scheduled Maintenance
−1.5 hrs
Monthly Shutdown
Planned Overhaul
−3.17 hrs
Major Stoppage
Bi-Monthly / Quarterly
−19.66 hrs
Real-Time
Uptime per Machine
was known only next-day
Auto
Plan vs. Execution Reconcile
zero manual cross-referencing
Line-Level
OEE Visibility
not just plant averages
Catalog
Driven Rules
update hours, not code

More Case Studies in Progress

Deployments completing now — detailed write-ups publishing Q2 2026

Coming Q2 2026

Automotive Tier 1 — JIT Orchestration

Coordinating just-in-time parts delivery with Tier 1/2 suppliers based on live production pace. Deployed across multiple plants.

Coming Q2 2026

CPG — Multi-Vendor Spend Intelligence

Unifying fragmented packaging vendor data, surfacing invisible spend, and automating reorder across 10+ suppliers.

Coming Q2 2026

Food & Bev — FSMA Compliance Autopilot

Automating FSMA 204 traceability, supplier cert management, and audit trail generation across multiple SKU lines.

Want a Case Study from Your Plant?

We run a low-risk, fixed-fee pilot on one workflow. You see results in 4–8 weeks — then we document the outcome together.

Start Your Pilot
ZeeHub AI ZeeHub AI
© 2026 ZeeHub AI • Backed by Berkeley SkyDeck, Alchemist, Google, Microsoft