Cleveland Works¶
This appendix is part of the AI Innovation Sprint engagement with Cleveland-Cliffs Inc. It covers the site-specific roadmap, project sequencing, investment rationale, and champion accountability for Cleveland Works. It is designed to be read as a standalone document or alongside the main report.
| Term | Definition |
|---|---|
| H1 | Horizon 1: Bridge the Gap (Months 0-6) — quick wins, data integration, immediate pain relief |
| H2 | Horizon 2: Build the Foundation (Months 7-12) — production-grade systems, cross-process linkage |
| H3 | Horizon 3: Predict & Optimize (Months 13-24) — enterprise optimization, predictive intelligence |
| PRJ-XX | Corporate project ID — cross-site programs that bundle site-level projects. See Chapter 5 of the main report for full detail. |
Site Summary¶
Cleveland Works is where this roadmap was born. As CLF's headquarters-city integrated mill, Cleveland operates under a single defining constraint: the No. 1 Steel Producing facility (1SP) must hit 28 heats per day, every day, with all casters booked through May and no redirect capacity anywhere in CLF. Every missed heat is lost revenue with no recovery path. The root cause is not equipment age or workforce skill — it is a 25-year-old information fracture created when the 2001 ISG restructuring merged three maintenance roles into one, broke the feedback loop between operations and maintenance, and left Cleveland running 70% reactive, 30% planned. Cleveland is where the corporate thesis — that CLF has an information flow problem, not an AI problem — was first validated, by every stakeholder in every conversation. It is the entry site for the two foundational corporate projects (Ops-Maintenance Data Integration and Predictive Maintenance), and the place where the PdM Proof-of-Value is now in progress, launched in early April on Chad Asgaard's direct authority.
The First 90 Days (Quick Wins)¶
1. Ops-Maintenance Data Integration (CLV-P01) — Start immediately. Jamie Betts and Dan Hartman already describe the same problem from opposite sides: a 14-minute operational delay has no matching work order in Tabware because operations and maintenance use different naming conventions in different systems to describe the same event. The fix is a semantic matching layer that maps operational delay categories to the Tabware asset hierarchy. Both datasets are approximately 95% structured. Deliverable in 8-12 weeks: a unified dashboard that lets anyone trace a production delay to its maintenance root cause — or prove that no maintenance action was taken. This is the foundation every downstream project depends on.
2. PdM Proof-of-Value at 1SP (CLV-P02) — ACTIVE. Charter submitted, contract in final approval. Chad Asgaard greenlit the PdM PoV at the March 16 corporate check-in ("get started right away"). Multi-asset eight-week charter (~$236K) covering three target assets spanning the data-richness spectrum: BOF bag house ID fans (primary — 4 ID fans, richest data profile, environmental compliance threshold), BOF scrubbing system (secondary — 3-week degradation cycle with RUL prediction, -5 heats/day when degraded), and Crane 300 (tertiary — single point of failure, near-zero instrumentation, MCSA from substation data). Plus a multi-asset data readiness assessment scoring every asset reachable through 1SP subsystems, a 1SP health monitoring dashboard prototype, and an IE-led data foundation blueprint. The PoV is not a single-asset experiment — it is the first working section of an integrated 1SP health monitoring system. Paul Aaron Dash, Jamie Betts, John Messi, Brian Thompson, and Phil Thorman as site champions.
3. Procurement Fast-Track Rules (CLV-P03) — Parallel quick win. A $300 power supply takes 2 months to order. Purchasing silently cancels requisitions after 60 days. Every item above $500 goes to John Stubna's desk. Risk-stratified approval rules — routing low-risk, low-cost items to automatic approval — can cut procurement lead time from weeks to hours. This requires a corporate policy change, not just technology, but the rules engine itself is a bounded 4-6 week deliverable that Dan Hartman and Paul Aaron Dash can champion locally.
Project Sequence¶
| ID | Project | Horizon | Corporate PRJ | Value ($/yr) | Champion |
|---|---|---|---|---|---|
| CLV-P01 | Ops-Maintenance Data Integration | H1 | PRJ-01 | $2-5M | Jamie Betts |
| CLV-P02 | Predictive Maintenance Platform | H1-H2 | PRJ-03 | $3-12M | Paul Aaron Dash |
| CLV-P03 | Maintenance Workflow & Copilot | H1 | PRJ-06 | $3-11M | Dan Hartman |
| CLV-P10 | Environmental Compliance | H1 | site-specific | $0.5-1M | Andrew Mullen |
| CLV-P04 | Process Risk & Cobble Prediction | H2 | PRJ-05 | $6-23M | Dan Hartman |
| CLV-P07 | Intra-Plant Logistics | H2 | PRJ-07 | TBD | Shipping team |
| CLV-P08 | Caster Chemistry Optimization | H2 | PRJ-08 | $2-8M | TBD |
| CLV-P09 | Utility Management | H2 | site-specific | $7-15M | Paul Aaron Dash |
| CLV-P05 | Quality & Yield Traceability | H2-H3 | PRJ-04 | $2-5M + enabler | Stephen Palmer |
| CLV-P06 | Production Scheduling & S&IOP | H3 | PRJ-02 | $20-65M | Andrew Mullen |
| CLV-P11 | Enterprise Data Platform Strategy | H2-H3 | enabler | enabler | Andrew Mullen |
Horizon 1: Bridge the Gap (Months 0-6)¶
Four projects launch in parallel. CLV-P01 (ops-maint integration) is the foundation — it produces the attribution data that CLV-P02 (PdM) and CLV-P04 (cobble prediction) both need downstream. CLV-P02 is in progress via the separate PdM PoV SOW on bag house, scrubbing, and cranes (launched early April). CLV-P03 (copilot + procurement) addresses the deepest visceral pain: green technicians going to repairs blind, paper PM sheets crumpled and thrown away, and the "Where's my Ask Jeeves?" problem Dan Hartman described. Wi-Fi dead zones must be solved for the copilot to work — "metal box inside another metal box" is a real infrastructure constraint. CLV-P10 (environmental compliance) is a low-cost, rules-based automation that can demonstrate value quickly.
Horizon 2: Build the Foundation (Months 7-12)¶
H1 outputs enable everything here. CLV-P04 (cobble prediction) needs the ops-maint attribution data from CLV-P01 and L2 access from the PdM infrastructure built in CLV-P02. Dan Hartman is the champion: "Some people just know the right call to make" — but those people are retiring. CLV-P09 (utility management) is Paul Aaron Dash's 20-year mission: no utility management system exists at Cleveland, compressed air alone is a $7M/yr savings opportunity backed by an engineering study, and new Emerson water meters coming online create the instrumentation foundation. CLV-P08 (caster chemistry) requires investigating why the Prime Metals model CLF spent millions on sits unused.
Horizon 3: Predict & Optimize (Months 13-24)¶
CLV-P06 (production scheduling and S&IOP) is the largest single opportunity at any site: $20-65M/yr. But it is H3 for a reason — it requires cross-stage visibility (from P01), quality traceability (from P05), and enterprise data architecture (from P11) to function. The bounded entry point is a cross-stage visibility dashboard (3-4 months), followed by roll change sequencing optimization, then the full scheduling optimizer. "1% improvement equals tens of millions" — John Stubna and Andrew Mullen validated this, but scope must be decomposed carefully to avoid the quicksand Palmer warned against.
Investment & Return¶
Cleveland-Relevant Corporate Projects¶
| Project | Cleveland Role | Cleveland Phases | Project Investment | Cleveland Value |
|---|---|---|---|---|
| PRJ-01 Ops-Maint Integration | Entry site | Ph1 ($540K), Ph2-3 (shared) | $1.84M | $2-5M/yr |
| PRJ-03 PdM Platform | Entry site | Ph1 ($236K active), Ph2 expansion | $2.08M | $3-12M/yr |
| PRJ-05 Cobble & Process Risk | Ph2 cobble at Cleveland HSM | Ph2 (shared with MDT) | $2.63M | Portion of $8-35M |
| PRJ-07 Logistics | Ph3 slab slotting | Ph3 (shared with Tilden) | $2.59M | Portion of $22-60M |
H1 investment at Cleveland: ~$776K (PRJ-01 Phase 1 + PRJ-03 Phase 1). Both launch immediately — PRJ-03 is already in contract approval.
Key value drivers: - 1SP throughput protection — every intervention that reduces unplanned downtime at 1SP has an outsized return because all casters run at 100% utilization with no redirect capacity - Phil Thorman's slab optimization — already saving $3M/month, proof that data-driven optimization works at Cleveland and that the culture can adopt it - PdM PoV as the proving ground — $236K investment in 8-week multi-asset PoV. Results create the playbook for cross-site expansion ($2.08M total program)
Champions & Accountability¶
| Project | Champion | Role | Why This Person |
|---|---|---|---|
| CLV-P01 | Jamie Betts | 1SP Maintenance Manager | Owns Tabware side, validated problem as "#1 do this now" |
| CLV-P02 | Paul Aaron Dash | Maintenance/Utilities | PdM PoV site lead, deep asset knowledge across 1SP |
| CLV-P03 | Dan Hartman | HSM Maintenance (25 yrs) | "Where's my Ask Jeeves?" — most articulate champion for workflow pain |
| CLV-P04 | Dan Hartman + Chad Helms | HSM + Iron Producing | Cobble domain expertise + BF thermal knowledge |
| CLV-P06 | Andrew Mullen | AI/Innovation Coordinator | Cross-functional visibility, corporate bridge |
| CLV-P09 | Paul Aaron Dash | Last Utility Engineer | 20-year champion, knows every system, irreplaceable knowledge at risk |
Executive sponsor: John Stubna. Young leadership team, "full support of corporate," open to change. The strongest GM-level buy-in of any steel site.
How This Site Feeds the Corporate Story¶
Cleveland is the entry site for two of the three highest-priority corporate projects:
-
PRJ-01 (Ops-Maintenance Data Integration) — Cleveland has the strongest signal (7+ stakeholders, 5 transcripts), the most structured data (Tabware + SQL both at ~95% hierarchy), and the natural first champion in Jamie Betts. The pattern proven here adapts to Middletown's Teams/SWAMI, Tilden's ELLIPS/Modular, and Burns Harbor's fragmented landscape.
-
PRJ-03 (Predictive Maintenance Platform) — The PdM PoV is in progress at Cleveland, launched in early April after Chad greenlit it. The bag house/scrubbing targets have the richest sensor data. Early results from this PoV will create the playbook for multi-asset PdM expansion to all four sites.
Cleveland also provides the cobble prediction baseline for PRJ-05 — cobbles destroy drive spindles, increase as experienced operators retire, and are the most frequently cited HSM pain point across three steel sites. Cleveland's L2 data and Dan Hartman's domain expertise make it the natural starting point.
Phil Thorman's slab cut optimization ($3M/month savings) is the proof point the corporate story needs: Cleveland has already demonstrated that data-driven optimization works in a unionized, legacy-system environment. The roadmap builds on that precedent, not against it.