<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/"><channel><title>Fintech on A naive bidder</title><link>https://anaivebidder.com/tags/fintech/</link><description>Recent content in Fintech on A naive bidder</description><language>en-us</language><lastBuildDate>Tue, 10 Nov 2026 06:00:00 +0100</lastBuildDate><atom:link href="https://anaivebidder.com/tags/fintech/index.xml" rel="self" type="application/rss+xml"/><item><title>Emails, spreadsheets, SQL: can AI agents connect the dots?</title><link>https://anaivebidder.com/posts/llm-reconciliation-benchmark/</link><pubDate>Tue, 10 Nov 2026 06:00:00 +0100</pubDate><dc:creator>Ruben Afonso</dc:creator><guid>https://anaivebidder.com/posts/llm-reconciliation-benchmark/</guid><description>&lt;p&gt;Can an AI agent open an Outlook email, follow it into an Excel workbook, query a database it has never seen and come back with the right number?&lt;/p&gt;
&lt;p&gt;In my last post, &lt;a href="https://anaivebidder.com/posts/build-your-own-llm-benchmark/"&gt;How to Build your Own LLM Benchmark&lt;/a&gt;, I shared tips for building your own. I kept two things back: the model names and the benchmark itself. Today you get both.&lt;/p&gt;
&lt;p&gt;Meet &lt;a href="https://github.com/rubenafo/finrec-bm"&gt;&lt;strong&gt;finrec-bm&lt;/strong&gt;&lt;/a&gt;, my benchmark for financial reconciliation, now on GitHub. It hands AI agents a number to verify and a pile of evidence: emails, spreadsheets, screenshots and a database. The answer has to be right to the cent, with the paper trail to prove it.&lt;/p&gt;</description></item></channel></rss>