Insight
AI economics for financial institutions
Analysis for people deciding whether to commit capital to an AI use case. Every piece carries a live model: change the assumptions and the numbers move.
The State of AI Economics in Financial Services
A recurring flagship report. Edition one: what public research establishes, what it does not, and modelled scenarios for the four workloads financial institutions are funding most.
Read the report →The Real Cost of AI Agents in Financial Services
Model inference is usually the smallest line in a regulated AI workflow. Here is the full cost stack, and the two assumptions that decide whether the case survives a capital committee.
Written for COO, CFO, Head of Transformation
AI Agents in Banking: From Tokens to Cost per Customer Outcome
A client's message is a hundred tokens. What the agent processes is several thousand. Here is how to get from a vendor's price list to a number your finance function can use.
Written for Head of Client Operations, Finance Business Partner
How to Build a Business Case for AI in a Bank
A capital committee does not need to be told AI is transformational. It needs a denominator, a sensitivity table and an exit condition.
Written for CFO, capital committee sponsor, programme director
RAG for Financial Services Leaders
Retrieval is how a model stops guessing about your product terms. It is also a data-governance programme, a refresh pipeline and a permanent cost line.
Written for Head of AI, Head of Data, COO
AI Agents for Investment Research: What Actually Drives Cost?
Research agents look cheap per note and expensive per defensible conclusion. The driver is analyst review time, not inference.
Written for Head of Research, CIO, COO of an asset manager
Build vs Buy AI in Financial Services
The comparison is not licence fee versus engineering cost. It is total cost per outcome, under your control requirements, over a realistic horizon.
Written for CTO, Head of AI, procurement sponsor
Why Human Review Can Cost More Than the LLM
Add a reviewer to every AI output and you have not built automation. You have built a drafting tool with a labour cost attached, sometimes rightly.
Written for COO, Head of Controls, operations leadership
AI Model Risk for Business Leaders: What Actually Matters?
You do not need to validate a transformer. You need to know what the system is allowed to do, how you would prove what it did, and what happens on a bad day.
Written for Board member, CRO, business sponsor
AI Agent Economics for Hedge Funds
At a 30-person fund, the AI question is not cost-to-serve. It is whether a marginal unit of analyst attention is worth more than the spend, and what a wrong number costs.
Written for COO, CIO, head of research at a hedge fund
What Does an Enterprise AI Agent Actually Cost?
A complete cost template for a regulated enterprise agent, with every line named, its behaviour described, and a model you can move.
Written for Finance, architecture and programme leadership
Model your own workload
The models in these pieces use illustrative assumptions. The assessment replaces them with your figures and tells you which ones your business case actually depends on.