October 03, 2026
For: Steve Watts, CEO, Pacific Glazing Corporation Date: October 03, 2026
Agent infrastructure has reached production maturity, fundamentally changing the build-versus-buy calculus for glazing automation. Small, calibrated AI models now run on commodity hardware with verifiable decision outputs, removing the enterprise procurement barriers that previously made automation risky. The convergence of these capabilities with advances in 3D generation and visual inspection creates a narrow window—approximately 18-36 months—where first-mover advantage in glazing automation remains available. The research confirms this window is closing faster than previously assessed.
What it is: Battle-tested orchestration systems (dots, OpenDots, ZCode) that coordinate multi-step AI task execution. These frameworks handle complex workflows like planning, tool use, and memory management out of the box.
Why it matters: PGC no longer needs to build agent infrastructure from scratch. The build-versus-buy decision has flipped. A glazing-specific agent can be deployed on proven foundations, reducing development time from months to weeks.
What it is: 0.8 billion parameter models (like "jeff") that deliver millisecond decisions with calibrated probability outputs on standard hardware. These models provide auditable, verifiable decision trails.
Why it matters: Addresses two PGC concerns simultaneously: data sovereignty (runs on-premise, no cloud dependency) and procurement requirements (verifiable outputs for enterprise contracts). Enables real-time decision support for field technicians without latency or security tradeoffs.
What it is: Reusable knowledge modules (universal-modder patterns) that encapsulate specialized expertise as executable skills. Each pack contains decision logic, best practices, and contextual awareness for a specific domain.
Why it matters: Direct application to glazing. PGC's accumulated knowledge—installation techniques, building codes, client preferences—becomes a deployable asset. Expert knowledge scales to every technician without requiring direct access to experts.
What it is: Advances in Gaussian-based 3D generation, high-resolution modeling (SILSA), and automatic floorplan-to-3D conversion. Architectural visualization tools now produce photorealistic results from basic inputs.
Why it matters: Client presentations transform from static images to interactive 3D walkthroughs showing glass placement, light behavior, and aesthetic impact. Installation planning gains spatial accuracy previously requiring expensive CAD work.
What it is: Multimodal models (VISTA benchmark) that solve complex inspection tasks by combining image understanding with contextual reasoning. Identifies defects, verifies standards compliance, and flags anomalies.
Why it matters: Glass quality inspection, installation verification, and damage assessment become automated. Reduces callback rates and liability exposure while standardizing quality across all projects.
What it is: AI agents with autonomous web navigation capabilities (dots framework). These agents browse, scrape, compare, and synthesize information without human intervention or blocking issues.
Why it matters: Supplier price monitoring, specification research, and compliance verification automate. Procurement research that currently takes days compresses to hours with consistent output quality.
What it is: A standardized integration layer connecting AI agents to enterprise tools and data sources. MCP enables AI systems to read from and write to existing software without custom development.
Why it matters: Integration costs drop significantly. PGC's estimation software, CRM, and project management tools connect to AI agents through a standard protocol rather than bespoke integrations.
What it is: Structured benchmarks (KaliBench) testing AI tool use for security workflows. Enterprise procurement now requires audited security compliance before AI vendor approval.
Why it matters: PGC's automation systems must pass security verification to win enterprise contracts. Early adoption of verification standards positions PGC ahead of competitors facing procurement roadblocks later.
What it is: Mobile-capable AI systems that process images, video, and sensor data in real-time. Guides field technicians through standardized checklists while capturing structured data.
Why it matters: Site surveys become consistent and auditable. Technician expertise requirements lower as AI prompts correct inspection sequences and captures documentation automatically.
What it is: Business video generation with quality standards, sound design, and motion-video-kit capabilities. Creates professional walkthroughs and presentations from static inputs.
Why it matters: Client proposals and installation explanations transform into dynamic visual content. Marketing and sales materials generate automatically from project specifications.
┌─────────────────────────────────────────────────────────────────┐
│ KEY CONNECTIONS │
├─────────────────────────────────────────────────────────────────┤
│ │
│ AGENT FRAMEWORKS ────┬────────→ SKILL PACKS │
│ Provides the │ Encapsulate glazing expertise │
│ orchestration engine │ for agent execution │
│ │ │
│ SMALL MODELS ────────┼────────→ EDGE DEPLOYMENT │
│ Calibrated outputs │ On-premise, low-latency, │
│ enable verification │ data-sovereign operation │
│ │ │
│ MULTIMODAL AI ───────┼────────→ VISUAL INSPECTION │
│ Image understanding │ Quality control, site surveys │
│ + reasoning │ automated │
│ │ │
│ MCP PROTOCOL ────────┼────────→ SYSTEM INTEGRATION │
│ Standard connections │ Links agents to existing │
│ for all tools │ PGC software stack │
│ │ │
│ SECURITY BENCHMARKS ──┴────────→ PROCUREMENT ACCESS │
│ Verified compliance Enterprise contracts require │
│ unlocks large clients security-audited AI systems │
│ │
└─────────────────────────────────────────────────────────────────┘
The convergence creates a compounding effect: agent frameworks provide the execution layer, small models provide trustworthy outputs, skill packs provide domain expertise, and MCP enables integration—all backed by security verification for enterprise procurement.
Objective: Test AI-assisted specification extraction from architectural PDFs
Setup:
Success metric: Reduce specification research time by 50% or more on test documents
Next step if successful: Expand to full PDF library and integrate with estimation workflow
Objective: Test multimodal AI for field documentation
Setup:
Success metric: AI captures inspection data accurately with minimal technician input
Next step if successful: Develop standardized inspection templates for field deployment
Objective: Test browser-based agent for procurement research
Setup:
Success metric: Agent identifies price opportunities or specification changes that manual research would miss
Next step if successful: Expand supplier coverage and integrate alerts into procurement workflow
1. Competitive Window Is Real and Narrow
The 18-36 month window for first-mover advantage in glazing automation is supported by current research. Open-source agent frameworks, small calibrated models, and skill pack architectures are converging now. Companies that build foundational capabilities in 2026-2027 will have sustainable advantages. Those who wait will face established competitors with trained models and integrated systems.
2. Build on Proven Foundations, Not Custom Development
PGC should leverage production-ready agent frameworks rather than building automation infrastructure from scratch. The research confirms these frameworks are battle-tested. Investment should flow toward glazing-specific skill packs and integrations, not underlying AI infrastructure.
3. On-Premise Capability Removes Procurement Barriers
Small calibrated models running on commodity hardware solve the enterprise verification problem. PGC can offer AI decision support with auditable outputs and data sovereignty guarantees—key requirements for large commercial contracts.
4. Expertise Capture Becomes Strategic Asset
The skill pack architecture pattern offers PGC a way to preserve and scale institutional knowledge. Senior glaziers' expertise becomes executable, distributed, and auditable. This is both a competitive moat and a risk mitigation strategy against workforce transitions.
5. Visual AI Applications Have Immediate ROI
Quality control automation, site survey standardization, and 3D visualization for client presentations offer tangible returns within 12 months. These applications do not require full autonomy—human oversight combined with AI assistance delivers value now.
6. Integration Strategy Determines Long-Term Success
MCP protocol adoption positions PGC to connect AI agents to existing systems without持续的 custom development. Prioritizing MCP-compatible frameworks early avoids costly integration rework later.