The Core Decision

Three Adoption Models

Build vs. Buy vs. Blend — the single most consequential architectural decision your organization will make.

MODEL I

Fully External

e.g. OpenAI API · AWS Bedrock · Azure AI · Claude

Data ControlLow

Setup CostLow

ScalabilityHigh

CustomizationLow

Privacy RiskHigh

Vendor-dependent — fastest path to value

MODEL II

Hybrid

e.g. Local LLM + GPT-4 for public queries

Data ControlMedium

Setup CostMedium

ScalabilityMedium

CustomizationMedium

Privacy RiskMedium

Balanced — the pragmatic sweet spot for most banks

MODEL III

Fully In-House

e.g. Bloomberg FinBERT · JPMorgan LLM Suite

Data ControlHigh

Setup CostHigh

ScalabilityLow

CustomizationHigh

Privacy RiskLow

Full sovereignty — for large banks and regulated institutions

Which Model by Firm Size?

Small <100 employees

Model: Fully External

Budget: $10K–$100K

Start with: ChatGPT, Claude for Work, Copilot

⚠ Barrier: AI literacy & budget

Mid-Size 100–1,000

Model: Hybrid

Budget: $500K–$5M

Start with: API integrations + workflow tools

⚠ Barrier: Legacy system integration

Large 1,000+

Model: Hybrid or Internal

Budget: $10M+

Start with: Enterprise licenses + AI teams

⚠ Barrier: Change management at scale

Banks All sizes

Model: Hybrid (best practice)

Budget: Varies + compliance premium

Start with: SR 11-7 compliant model inventory

⚠ Barrier: Model risk & fair lending