Expertise · Perspective
Ask the person who has shipped it.
I am glad to share a view with investors, analysts and enterprise leaders on how large organisations adopt, govern and pay for AI, how data platforms compare from the customer’s seat, and how credit and surety risk is priced. My perspective comes from deployment rather than theory: I lead AI product development in one of the most regulated corners of financial services.
Range
Three areas, one practitioner
The full AI value chain, the data platforms underneath it, and the risk it is meant to price.
How organisations adopt, govern and pay for AI
- LLM adoption patterns and enterprise AI workflow
- Hyperscaler-hosted versus direct model access
- Hyperscalers versus independent platforms
- SaaS and AI pricing models, copilot tiering
- AI governance that real use cases can clear
Snowflake, Databricks and Foundry from the customer’s seat
- Snowflake versus Databricks in a regulated enterprise
- Unity Catalog, Genie and Palantir Foundry in practice
- Production pipelines built on both Foundry and Databricks
- Data governance underwriters and regulators trust
- Migration, cost and team-skills trade-offs
Pricing and portfolio risk in credit, bond and political risk
- Quota share, excess of loss and stop loss structures
- Single risk credit, surety bonds and SRT tranches
- PD, LGD, EAD and Basel III modelling
- Portfolio capital, correlation and the credit cycle
- Where AI changes underwriting, and where it doesn’t
Token cost is the wrong thing to optimise.
The thesis
Most AI business cases fixate on the price of a model call. The real return comes from somewhere else: redesigning workflows so that LLM calls replace human decision points. Get the workflow right and token cost becomes a rounding error. Get it wrong and the cheapest model in the market still won’t pay back.
I see this from inside a live enterprise stack built on hyperscaler infrastructure: a company-wide assistant, a governed agentic platform held to an approved model tier, and a formal governance process that every AI use case must clear.
Try it
Questions people bring me
Click any question and the site’s AI assistant will give you a first answer from my public profile. For a fuller view, write to me.
Get in touch
Ways to reach me
A question or a second view
Weighing a decision on enterprise AI, data platforms or credit and surety? Write to me with the question and the context.
Write to BanuResearch & collaboration
Researchers, founders and builders working where AI meets financial risk. I am always interested in a good problem.
Start a conversationWorkshops & talks
Board, leadership and conference sessions that make AI and risk concrete for mixed audiences.
SpeakingEvery conversation respects my employment obligations. I do not share confidential information about my employer, its clients or its systems.