Scott Macpherson
Founder, Bancadia
Scott Macpherson is the founder of Bancadia, a data layer that AI agents query through an MCP server to recommend financial products to their own users, starting with US business checking accounts. Before Bancadia, Scott spent five years as a software engineer at a large bank. Scott writes about how product discovery is moving from web pages to agents, and what banks, credit unions and fintechs need to do to be found when an agent is the one doing the research.
Background
Scott spent five years as a software engineer at a large bank before starting Bancadia.
Bancadia grew out of a different project. Scott set out to build an AI wealth advisor agent, and hit a wall that had nothing to do with the model: there was no practical way for the agent to find financial products and recommend one. The product information existed, but not in any form an agent could query.
That gap became Bancadia. Instead of building one more agent that couldn't find products, Scott started building the product-data layer every agent was missing: structured, queryable data served over MCP, starting with business checking. The articles here make the case for why that layer matters.
Articles
- Product discovery is moving from web pages to APIs and MCP servers
Why I think product discovery is shifting from HTML built for people to data that agents query, and what Bancadia is building for it.
- In AI search, the agent is the lead: why affiliate networks will overtake ad networks
Agents arrive at a recommendation already holding the user's full context. That makes them bottom-of-funnel lead sources, and conversion-based pay follows.
- AI agents know everything about their user. Your product data has to keep up.
An agent's context only helps if it can be turned into filters. If your product attributes aren't structured, the agent can't match you to the user you fit best.
- What AI agents need from product data: fresh, structured, token-efficient and accurate
The four requirements for product data that agents can recommend from, with a real token measurement of page scraping versus an MCP query.
- Affiliate content was written for a generic reader. AI agents know exactly who they're helping.
Listicles had to be broad because their author knew nothing about the reader. Agents know everything, and they need a data layer, not an article.