Appendix

The simulation was built around a fictional client — NovaTech Semiconductor — planning a $2.4 billion advanced semiconductor fabrication facility in the American Southwest. The scenario was calibrated to match real-world fab projects announced under CHIPS Act incentives.

Seven stakeholder personas were developed, each representing a distinct role on the procurement committee. Each persona received 100 questions designed to mirror their actual decision-making process, distributed across funnel stages from initial research through final evaluation.

All 700 questions were processed from the perspective of a neutral market analyst — not as JE Dunn asking about itself. This mirrors how a procurement team member would actually query an AI: requesting objective market intelligence.

Questions followed a structured JSON schema with fields for question text, stakeholder persona, funnel stage, topic category (27 categories), and expected response type. This structure enabled systematic classification of all 700 responses across six analytical dimensions.

The processing engine consisted of four components: a research dictionary containing 22 structured entries about JE Dunn's capabilities, a topic classifier with 27 categories mapping to semiconductor construction domains, a response generator producing detailed LLM-style outputs, and an output writer formatting results for analysis.

The research corpus included 22 structured entries covering: JE Dunn's semiconductor project history, technical capabilities (cleanroom construction, vibration control, contamination management), safety and compliance records, team qualifications, and market position data. Each entry was sourced from publicly available information only.

All responses were generated from publicly available training data, not internal JE Dunn materials. This ensures the results reflect what actual AI systems would produce when queried by external buyers.

Every response was classified across six quantitative dimensions:

This assessment reflects LLM behavior at a point in time. AI training data refreshes on 3–6 month cycles, meaning results may shift as new content enters the training pipeline. The simulation uses a structured methodology, but actual buyer queries will vary in phrasing, specificity, and context.

The revenue risk model uses market sizing from public sources (SIA, SEMI) and applies share assumptions. Actual market share will depend on factors beyond AI visibility, including relationship networks, pricing, and project-specific team composition.

Competitor analysis is based on publicly available information. Internal strategies, undisclosed projects, and non-public partnerships at Turner, Skanska, or other firms may affect the competitive landscape in ways not captured here.

Role Cluster: Executive  |  Funnel Stage: Evaluation/Shortlist

Decision Focus: Financial risk mitigation, cost certainty, contractor financial stability, insurance and bonding capacity, track record of delivering on-budget.

AI Behavior: Asks about financial comparisons between GCs, risk profiles, cost overrun histories, and financial stability indicators. Highly influenced by third-party financial reporting and analyst coverage.

Key Finding: 35% confidentiality friction — moderate but enough to create hesitation on financial validation questions. Turner's publicly reported $28B revenue creates automatic scale credibility.

Role Cluster: Commercial/Risk  |  Funnel Stage: Initial Research/Long List

Decision Focus: Market scanning for qualified GCs, RFQ list development, competitive benchmarking, qualification criteria validation.

AI Behavior: Broad market scans asking "who are the best GCs for semiconductor construction." This is where the initial shortlist forms — and where JE Dunn's visibility deficit is most damaging.

Key Finding: 42% Turner mention rate (highest of all), 69% confidentiality friction. In her simulated evaluation, Turner surfaced in 42% of questions — meaning Turner was effectively pre-loaded onto the shortlist before any human interaction.

Role Cluster: Technical  |  Funnel Stage: Technical Evaluation

Decision Focus: Cleanroom construction expertise, vibration control capabilities, contamination management, tool install timelines, technical team qualifications.

AI Behavior: Detailed technical queries about semiconductor-specific construction capabilities. Needs evidence of cleanroom classifications achieved, ISO certifications, and specific project metrics.

Key Finding: 64% confidentiality friction on technical questions — the AI cannot cite specific JE Dunn cleanroom projects because the data doesn't exist in its training set. This creates a devastating gap for the most evidence-hungry evaluator on the committee.

Role Cluster: Technical  |  Funnel Stage: Technical Evaluation/Shortlist

Decision Focus: Project execution methodology, scheduling capabilities, safety record, labor management, self-perform vs. subcontractor strategy.

AI Behavior: Asks about construction management approaches, safety statistics, workforce availability, and execution track record on comparable-scale projects.

Key Finding: Moderate visibility gap — safety record and execution methodology are partially visible, but lack the project-specific detail that would differentiate JE Dunn from generic GC responses.

Role Cluster: Technical  |  Funnel Stage: Reference Validation

Decision Focus: Client references, past performance validation, relationship quality, responsiveness to owner concerns, change order management.

AI Behavior: Reference-seeking queries — "what do clients say about JE Dunn?" The Owner's Rep is validating the selection, not making it. But hedging language at this stage can derail a selection that's already been provisionally made.

Key Finding: 48% hedging rate — the highest of all stakeholders. When Bob asks AI for reference validation, nearly half of responses contain equivocating language that undermines confidence in JE Dunn at the most critical moment.

Role Cluster: Commercial/Risk  |  Funnel Stage: Shortlist Evaluation/Capital Allocation

Decision Focus: Capital project delivery track record, budget adherence, schedule reliability, scope management, and vendor coordination on mega-projects.

AI Behavior: Evaluative queries about project delivery performance, cost control history, and comparative GC capabilities on semiconductor-scale capital programs. Diane is validating whether JE Dunn can deliver at the scale and complexity required.

Key Finding: The absence of detailed project delivery metrics in LLM training data means AI responses default to generic GC descriptions rather than JE Dunn's specific track record of on-time, on-budget semiconductor fab delivery.

Role Cluster: Commercial/Risk  |  Funnel Stage: Risk Assessment

Decision Focus: Project risk identification, insurance/bonding evaluation, safety record analysis, financial stability assessment, risk mitigation strategies.

AI Behavior: Detailed risk-focused queries about safety incidents, bonding capacity, and risk management processes. Needs quantitative evidence of risk performance.

Key Finding: 5% Turner mention rate — the one bright spot. Risk-focused queries are less brand-driven and more metric-driven, meaning JE Dunn's actual safety record can compete here. But the AI still lacks the specific data to make the case effectively.

Semiconductor Industry Association (SIA) — US semiconductor investment tracking, $540B+ announced investments. sia.org

SEMI — Fab construction tracking, 18 new fabs breaking ground in 2025. semi.org

TSMC Arizona — $100B US fab expansion commitment. tsmc.com

Micron Technology — $200B US investment across Idaho, New York, Virginia. micron.com

CHIPS and Science Act — Federal semiconductor manufacturing incentives. commerce.gov/chips

ENR (Engineering News-Record) — Contractor rankings, market intelligence. enr.com

Construction Dive — Industry news, contractor analysis. constructiondive.com

Cleanroom Technology — Semiconductor construction specialized coverage. cleanroomtechnology.com

Turner Construction — 2025 restructuring into two business lines, advanced technology vertical. turnerconstruction.com

BusinessWire — Turner/Dornan $1.6B acquisition announcement. businesswire.com

Skanska USA — SAT (Skanska Advanced Technology) launch February 2025, expansion September 2025. usa.skanska.com

GlobeNewsWire — Skanska Advanced Technology dedicated press coverage. globenewswire.com

Forbes — JE Dunn ranked #97 on Largest Private Companies. forbes.com

Spotlight LLM Visibility Simulation — 700 questions, 7 personas, 2,319 citations analyzed. Primary data source for all visibility metrics.

Spotlight Competitive Gap Analysis — 18-metric scorecard comparing JE Dunn vs. Turner across visibility dimensions.

Spotlight Trusted Voice Overlap Analysis — 21 trusted voices mapped across 3 tiers of influence.

Revenue Risk Model — Built on SIA/SEMI market sizing, ENR ranking share analysis, and historical GC market share patterns.