AI that does the work
A closer look at how Midas turns plain-English questions into sourced dashboards, reports, and alerts you can trust.
Midas does more than answer prompts. It decomposes complex questions, chooses the right data and visuals, explains its reasoning, and keeps your views up to date. This page walks through the key AI systems behind that experience.
What the AI actually does
Midas is built around a research pipeline rather than a chat box. A question is decomposed into sub-questions, each one routed to the right source — SEC filings, transcripts, fundamentals, or the documents in your own workspace — and answered independently before the results are reconciled into a single view. Where sources disagree, Midas says so instead of averaging the difference away.
Every figure carries a citation back to the sentence, table cell, or filing line it came from, so a number can be verified in one click rather than rebuilt from scratch. The reasoning behind a conclusion is exposed as well as the conclusion: which comparables were chosen, which periods were normalised, what was excluded and why. That is the part an investment committee asks about.
Outputs are not read-only. Dashboards, screens, and AI-generated columns stay live, re-running against new filings and prices, and monitoring agents watch the names and themes you care about and alert you when a disclosure changes materially. Reports export in your own format and language.
Your content is never used to train foundation models, workspace data is isolated, and the analyst remains the decision-maker throughout. The sections below walk through each system in turn.
Keep up with
all your companies.
Free plan. 150 credits a month, forever.