Phase 2: Architecture & Operating Model Review
Our Perspective
These insights are synthesized from our deep experience executing M&A technical diligence and optimizing enterprise architecture. They reflect our ground-truth perspective on what investors must prioritize to separate AI hype from defensible, structural value.
The architecture and operating model review determines whether the company’s technical foundation and organisational structure can support the AI opportunities identified in Phase 1. Many AI initiatives fail not because the use case was wrong, but because the architecture could not support it or the organisation was not ready to operate or scale it.
The architecture assessment evaluates the company’s current model strategy: which providers are used, how models are integrated, whether provider-switching is possible, how data flows between systems and models, and whether the architecture supports the cost, latency, security, and scalability requirements of the priority use cases. Companies that have adopted AI ad hoc often have fragmented architectures – multiple providers, inconsistent integration patterns, unclear data boundaries, and no cost monitoring.
Data readiness is typically the most significant constraint. AI systems require clean, accessible, well-governed data. The review assesses data quality, completeness, accessibility, permissioning, and compliance across the datasets required for each priority use case. It identifies gaps that must be addressed before pilots can begin and estimates the effort required to close them.
The toolchain review evaluates the company’s AI development and deployment infrastructure: model hosting, prompt management, evaluation frameworks, monitoring, logging, version control, and CI/CD for AI workflows. Many companies have production AI features with no systematic evaluation, no output monitoring, and no rollback capability. These gaps create quality and reliability risk.
The governance gap analysis identifies missing policies, processes, and controls. This includes approved tool lists, data usage policies, human review requirements, AI-generated code controls, customer data restrictions, incident response procedures, and ownership accountability. The goal is not to get burdened with bureaucracy but drive at a responsible speed. Good governance enables teams to move faster with confidence.
Model strategy connects architecture to economics. The review recommends where the company should use commercial models, where private or local models are more appropriate, where deterministic systems should be preserved, and how routing decisions should be made. This is where the “local where good enough, cloud where powerful enough” principle becomes operational.
The output is a target-state architecture with a clear gap analysis, a recommended governance framework, and a prioritised remediation plan. This ensures that Phase 3 pilots are built on a foundation that can scale.
Apply this thinking to your portfolio.
Thought Source helps investors and operators assess AI architecture, defensibility, and value creation.