A Conceptual Framework for Enhancing Workforce Readiness for Smart Manufacturing in the AI Era

Summary

This paper introduces the Workforce Readiness Level (WRL) framework, a nine-stage, rubric-anchored, individual-level assessment model designed to operationalize the evaluation of AI-era smart manufacturing competencies. The framework adapts NASA's Technology Readiness Level (TRL) scale to individual workforce competency, addressing the gap between the competencies required on the shop floor and those delivered by traditional engineering and technology education.

Background and Motivation: The convergence of artificial intelligence (AI), the Industrial Internet of Things, cyber-physical systems, and advanced robotics is reshaping manufacturing faster than engineering curricula can adapt. A joint study by Deloitte and the Manufacturing Institute projects that up to 2.1 million U.S. manufacturing jobs may be unfilled by 2030, with digital, data-analytics, and AI/automation skills among the leading unmet requirements. Existing frameworks each have structural limitations: the MSSC Certified Production Technician (CPT) and CPT+ credentials contain limited explicit AI, ML, IIoT, or digital-twin content; the SME Tooling U-SME Smart Manufacturing Certificate is largely knowledge-based via multiple-choice items rather than competency-demonstrated through performance tasks; European microcredential frameworks are deliberately general-purpose and lack manufacturing specificity; O*NET is observational rather than prescriptive; DigComp 2.2 targets general citizens rather than manufacturing technicians; and the ACATECH Industrie 4.0 Maturity Index assesses organizational rather than individual maturity. None of these tools combines (i) manufacturing domain specificity, (ii) explicit AI/CPS content, (iii) stage-gated progression, and (iv) performance-based assessment.

Framework Design: The WRL construct is grounded in three design commitments: (1) competency must be demonstrated through observable performance on artifacts, not merely asserted through coursework completion or self-report; (2) progression must be stage-gated and monotonic (once a stage is reached, it is not lost by definition); and (3) the construct must be multidimensional, because AI-era manufacturing roles inherently span computing, controls, human factors, and analytics. Three further design choices distinguish WRL: it is an attribute of a person at a point in time (can rise through learning or fall through skill loss); it is structurally similar to NASA's TRL with nine stages arranged in an awareness/lab/production progression; and assessment is artifact-anchored and double-rated.

Nine WRL Stages: The nine stages partition the trajectory from complete beginner to innovation leader into three groups of three stages: awareness (WRL 1-3), applied practice (WRL 4-6), and autonomous leadership (WRL 7-9). Stages 1-3 emphasize awareness and foundational literacy (recognizing terminology, explaining concepts, performing guided exercises under supervision). Stages 4-6 emphasize applied skill in progressively less controlled environments (single laboratory subsystem, integrated multi-system laboratory scenario, real but mentored production setting). Stages 7-9 emphasize autonomous performance, supervision, and innovation in production settings (meeting independent production KPIs, supervising others and driving Kaizen/Six Sigma improvements, leading cross-functional initiatives that change the production system itself). Natural articulation points for stackable credentials exist at WRL 3 (foundational, often a community-college stopping point), WRL 5 (laboratory-competent, often a bachelor's-degree stopping point), and WRL 7 (production-autonomous, the journeyman-equivalent benchmark for incumbent workers).

Four Competency Pillars: Each WRL stage is evaluated along four pillars: P1 (Digital & AI Literacy) covers computational thinking, data structures, applied statistics, the machine-learning lifecycle, foundation-model and prompt literacy, model interpretability (SHAP, LIME), and AI ethics including bias, privacy, and human oversight; P2 (Cyber-Physical Systems Fluency) covers industrial controls (IEC 61131-3 PLCs), SCADA and HMI design, IIoT sensor selection and commissioning, industrial network protocols (OPC-UA, MQTT, PROFINET, EtherNet/IP), OT/IT security hygiene (IEC 62443 awareness), and digital-twin construction; P3 (Human-Machine Collaboration) covers collaborative-robot safety per ISO 10218 and ISO/TS 15066, AR/VR-assisted work instruction, human-in-the-loop AI for inspection and decision support, ergonomics, and teaming etiquette in mixed human-cobot cells; P4 (Data-Driven Decision Making) covers KPI definition and instrumentation (OEE, FPY, cycle time, energy intensity), statistical process control, design of experiments, root-cause analytics, Lean/Six Sigma DMAIC, A3 problem solving, and continuous-improvement portfolio management.

Evaluation Model: The evaluation model uses a behaviorally anchored rubric where si,j ∈ 0, 1, 2, 3 denotes a learner's demonstrated proficiency on pillar j at WRL stage i. The composite stage score is Si = (1/4)Σwj si,j with default weights wj = 1. A learner is certified at WRL k if both (a) Si ≥ τ (default τ = 2.25) and (b) minj si,j ≥ 2 (the no-thin-pillar rule) hold for every stage i ≤ k. Under integer scores and default weights, the joint rule reduces to no pillar below 2 and at least one pillar at 3 at each stage. Programs seeking a stronger high-stakes gate may set τ = 2.5, requiring at least two pillars at 3. The framework also yields diagnostic tools: the pillar profile vector identifies relative strengths and remediation targets; the gap-to-next-stage indicator Gk+1 = max(2 - sk+1,j)+ identifies the pillar most in need of remediation. At the cohort level, the workforce-readiness index WRI = (1/N)Σkn is used as a comparative benchmark across cohorts and pillars, while acknowledging that averaging ranked levels has no strict interpretation as a competency state.

Implementation at IDEELab: The framework is instantiated at the Innovation, Design, and Engineering Education Laboratory (IDEELab) at Mississippi State University, a reconfigurable smart-manufacturing teaching and research facility with four interconnected cells: Robotics & Assembly Cell (two UR10e cobots, FANUC industrial arm, AGV, vision-guided pick-and-place stations), Process & Control Cell (Allen-Bradley CompactLogix and Siemens S7-1500 PLCs, conveyor with photoeye/RFID stations), Additive & Subtractive Cell (Markforged Metal X, fourteen Bambu Lab printers, Haas CNC machining centers, Flow waterjet, metrology equipment), and Digital-Twin & Analytics Cell (NVIDIA Omniverse, Siemens Plant Simulation, JupyterHub with PyTorch and scikit-learn). All cells publish to a shared OPC-UA/MQTT data backbone streaming live into the Digital-Twin & Analytics Cell.

Case Studies: Four case studies are analyzed in depth from 89 sponsored capstone projects delivered over four semesters (Fall 2024-Spring 2026). Case 1 (AI-Driven PoDFA Rating for Aluminum Dynamics, Spring 2025) involved four students building an SEM-image processing pipeline that classifies inclusion components by morphology, achieving a validation macro-F1 of 0.89; all four students met the WRL 5 joint rule, with two subsequently satisfying WRL 6. Case 2 (Human-Detection Sensor Suite for a Hol-Mac Automatic Side-Loader, Fall 2025) involved five students performing a comparative evaluation of AI cameras, LiDAR, and ultrasonic sensors for personnel detection; four of five satisfied WRL 5, with the fifth reaching WRL 6 after a supplemental on-site exercise. Case 3 (Python-Driven End-of-Line Test Station, Refrigerated Solutions Group, Fall 2024-Fall 2025) involved two successive teams; 7 of 9 students satisfied WRL 5, with three of the five Fall 2025 students additionally satisfying WRL 7 by virtue of the on-site production install. Case 4 (Machine-Learned Actuator-Line Model for Marine Propellers, IDEELab/U.S. Navy, Spring 2026) involved five students developing a deep-neural-network actuator-line model achieving a held-out R2 of 0.94 on integrated thrust and 0.88 on torque; all five satisfied WRL 6, with two reaching WRL 7 through subsequent Navy placements.

Cross-Case Synthesis: The framework transferred across sectors (aluminum smelting, heavy-truck manufacturing, commercial refrigeration, and defense/naval research) without change to the pillar rubric. The no-thin-pillar rule helped flag gaps in Cases 1 and 2, surfacing P2 or P3 pillars sitting at the floor behind otherwise strong P1 profiles; the rule was operationally binding in Case 3, blocking two Fall 2024 students at WRL 5 for P2 below the floor. The WRL 6→7 step was reached only after industry-embedded exposure, indicating that advancement to the highest stages is gated by industry-embedded experience rather than additional coursework. Across the highlighted cohorts, the workforce-readiness index ranged from 5.2 to 6.4.

Accreditation Implications: The four pillars map onto specific ABET Engineering Accreditation Commission Student Outcomes: P1 and P4 provide strong evidence for SO1 (complex problem solving), SO6 (experimentation and data analysis), and SO7 (acquiring and applying new knowledge); P2 anchors SO2 (engineering design) and contributes strongly to SO1; P3 carries SO4 (ethical and professional responsibility, largely via functional-safety practice) and SO5 (teamwork in mixed human-machine cells). Every ABET outcome receives at least supporting evidence from the pillar set. Aggregating rubric scores by pillar across the four highlighted cohorts (N = 23 students), P1 is strongest (mean 2.65), while P4 (2.05) and P2 (2.20) sit closest to the no-thin-pillar floor, identifying data-driven decision making and cyber-physical fluency as the two competencies where targeted investment would most raise cohort WRI.

Workforce-Development Implications: Approximately one third of the 89 projects were sponsored by manufacturers with primary operations in Mississippi. The pillar-level patterns imply a division of labor for the regional workforce ecosystem: four-year programs can carry students furthest on P1 and P4, while the persistent P2/P3 gaps are best closed through employer-embedded experiences (co-ops, MMEP projects) and stacked technician credentials. P2 sits just above the floor (mean 2.20), making it simultaneously the highest-demand and least-satisfied competency, the clearest case for a shared university-community-college credential at WRL 5. P4 is the pillar nearest the floor (mean 2.05) and the one most likely to block certification, identifying data-driven continuous improvement as the highest-leverage target for incumbent-worker training.

Conclusion and Future Work: The paper frames the reported numbers as illustrative of framework mechanics on a single-institution pilot rather than a psychometric validation. Five lines of future work follow: (i) calibrate pillar weights through a Delphi study with regional manufacturers; (ii) develop an open-source learner-record schema interoperable with IMS Caliper and Open Badges; (iii) extend the rubric to generative-AI competencies such as prompt engineering; (iv) replicate the framework across multiple institutions to test reliability and predictive validity at scale; and (v) test the certification-weighting recommendation by scoring learners who hold hands-on-evaluated versus knowledge-only certifications.

Improvements to AI Systems Based on This Paper

1. AI-Driven Competency Assessment Engine

• Improvement: Build an AI system that automatically evaluates learner performance on artifacts (e.g., capstone project reports, code, sensor logs) against the WRL rubric’s four pillars (P1–P4) and nine stages. Use NLP to parse written deliverables, computer vision to assess physical task execution (e.g., cobot programming), and time-series analysis on production data (e.g., OEE, FPY) to score proficiency levels (0–3) per pillar.

• What it can do: Generate real-time, artifact-anchored WRL stage certifications without human double-rating, flag thin-pillar gaps (e.g., P2 below 2) automatically, and produce diagnostic pillar profiles for each learner, reducing assessment bias and scaling to large cohorts.

2. Adaptive Learning Path Recommender

• Improvement: Use the WRL’s stage-gated progression and gap-to-next-stage indicator (Gk+1) to train a reinforcement learning agent that recommends personalized training modules (e.g., PLC programming for P2, SPC for P4) based on a learner’s current pillar scores. The agent optimizes for monotonic stage advancement (WRL 3→4→5…) while respecting the no-thin-pillar rule.

• What it can do: Dynamically sequence coursework, lab exercises, and industry-embedded projects (e.g., co-ops) to close specific pillar deficits, predict time-to-next-stage, and suggest stackable credentials (e.g., community-college WRL 3 → bachelor’s WRL 5) tailored to individual profiles.

3. Predictive Workforce Readiness Forecaster

• Improvement: Train a supervised model on historical WRL scores (e.g., from the 89 capstone projects) with features like project sector, team size, prior coursework, and lab exposure to predict a learner’s final WRL stage and pillar scores. Use SHAP/LIME (as referenced in P1) for interpretability.

• What it can do: Forecast cohort-level workforce-readiness indices (WRI) for manufacturers, identify which projects or training interventions most likely yield WRL 6–7 (industry-embedded) outcomes, and proactively recommend employer partnerships (e.g., Navy placements) to close P2/P3 gaps before graduation.

4. AI-Powered Rubric Calibration and Validation Tool

• Improvement: Implement a Bayesian calibration system that learns optimal pillar weights (wj) and thresholds (τ) from expert Delphi study data and multi-institution outcomes. The AI can simulate certification decisions under different weightings to test robustness (e.g., τ=2.25 vs. 2.5).

• What it can do: Automatically update the WRL rubric as new competency data emerges, provide confidence intervals for stage certifications, and generate evidence for ABET accreditation by mapping pillar scores to Student Outcomes (SO1–SO7) with quantified contributions.

5. Generative-AI-Enhanced Assessment Content

• Improvement: Extend the framework to generative-AI competencies (future work item iii) by training an LLM to generate scenario-based assessment tasks (e.g., “Given a digital twin of a conveyor line, design a prompt to detect anomalies”) and auto-grade responses against P1/P4 rubrics.

• What it can do: Produce unlimited, validated assessment items for WRL stages 1–3 (awareness) and 4–6 (applied), enabling low-cost, high-frequency testing of AI literacy (e.g., model interpretability, ethics) without human proctoring.

6. Cross-Institutional Benchmarking AI

• Improvement: Develop a federated learning system that aggregates de-identified WRL scores across universities and community colleges to train a shared model for predicting certification success, while preserving data privacy.

• What it can do: Identify best practices (e.g., which lab configurations or teaching methods yield highest P2 gains), benchmark institutional WRI distributions, and provide early-warning signals for programs where learners consistently fail the no-thin-pillar rule (e.g., P4 deficits in technician-track students).

The convergence of artificial intelligence (AI), Industrial Internet of Things, cyber-physical systems, and advanced robotics is reshaping manufacturing faster than engineering curricula can adapt, widening the gap between the competencies required on the shop floor and those delivered by traditional engineering and technology education. This paper proposes a Workforce Readiness Level (WRL) framework, which adapts the Technology Readiness Level scale into nine progressive competency stages and a four-pillar rubric, digital and AI literacy, cyber-physical systems fluency, human-machine collaboration, and data-driven decision making, aggregated through a composite stage score and a cohort-level workforce-readiness index under a ``no-thin-pillar'' rule. The framework is instantiated at a university smart-manufacturing teaching laboratory and draws on 89 sponsored capstone projects delivered over four semesters, four of which are analyzed in depth. Four pillars jointly span the relevant ABET student outcomes. Across the highlighted cohorts the workforce-readiness index ranged from 5.2 to 6.4, and the no-thin-pillar rule was diagnostically informative in three of the four cases and the binding certification constraint in one, repeatedly surfacing cyber-physical and data-driven-decision gaps concealed behind strong analytics profiles; advancement to the highest stages was gated by industry-embedded experience rather than additional coursework. WRL offers educators, accreditation bodies, and regional workforce systems a common, evidence-based instrument for diagnosing and advancing workforce readiness; future work will calibrate pillar weights and test reliability and predictive validity.