proprietary data as AI moat
Databricks: proprietary data, not model choice, is the moat in agentic financial services DatabricksTL;DW
- Proprietary data, not frontier models, creates competitive advantage—less than 1% of enterprise data is publicly available on internet.
- Bring models to your data, not data to models; contextual information about business-specific scenarios is unavailable to any general AI model.
- Most AI deployments today add zero revenue impact; they improve efficiency 20-30% but lack proprietary context for critical business decisions.
- Build agentic systems left-to-right: first define data needs, then governance, then pick a reasoning engine—not right-to-left by starting with model choice.
- Three critical questions for agentic future: Who wins the model race? What could go wrong with agent sprawl? What's my unique competitive moat?
- AI without governance is risk; centralize enterprise data from all systems (Salesforce, Workday, trading tools) into single lakehouse with role-based access control.
- Financial services competitive advantage maturity curve: operational productivity tasks → automated workflows → reimagined business models leveraging proprietary data.
- RBC Capital Markets built equity research agent reducing report time at 99% accuracy; MasterCard monetizes proprietary real-time transaction data via intelligent services.
- Enron case study: unified data access reveals fraud red flags—mismatches between reported valuations and internal models, suspicious methodology changes, CFO-ordered calculation changes.
- Start with business challenge, not technology; too often teams pick models first then search for problems—reverse this pattern and define use cases before model selection.
TL;DW
- Proprietary data, not frontier models, creates competitive advantage—less than 1% of enterprise data is publicly available on internet.
- Bring models to your data, not data to models; contextual information about business-specific scenarios is unavailable to any general AI model.
- Most AI deployments today add zero revenue impact; they improve efficiency 20-30% but lack proprietary context for critical business decisions.
- Build agentic systems left-to-right: first define data needs, then governance, then pick a reasoning engine—not right-to-left by starting with model choice.
- Three critical questions for agentic future: Who wins the model race? What could go wrong with agent sprawl? What's my unique competitive moat?
- AI without governance is risk; centralize enterprise data from all systems (Salesforce, Workday, trading tools) into single lakehouse with role-based access control.
- Financial services competitive advantage maturity curve: operational productivity tasks → automated workflows → reimagined business models leveraging proprietary data.
- RBC Capital Markets built equity research agent reducing report time at 99% accuracy; MasterCard monetizes proprietary real-time transaction data via intelligent services.
- Enron case study: unified data access reveals fraud red flags—mismatches between reported valuations and internal models, suspicious methodology changes, CFO-ordered calculation changes.
- Start with business challenge, not technology; too often teams pick models first then search for problems—reverse this pattern and define use cases before model selection.
Jamin Nakai argues frontier models are commoditizing fast, so financial institutions' edge shifts to data strategy—grounding agents in proprietary context via a governed lakehouse. Covers agentic sprawl risks, RBC Capital Markets and Mastercard as real examples, and a hypothetical Enron fraud-detection case to show what context-aware reasoning unlocks.