Would this deck survive a managing director's red pen?
Paste the slide text or drop the .pptx. Your browser reads every figure on every slide and checks them against each other as you type: the same metric on two slides, growth rates against their end points, margins, table totals, "#1" claims, unit and date drift, sources and language. All free, before you sign in. Then refresh the deck to new numbers with a change list you approve line by line, and run the QC pass that writes the deck check report. Every quote it writes is checked against the slide it cites.
Each example comes with a saved model run, one per verdict and one for the refresh lane, so you can see the whole page for free.
Your recent runs
What this does, and what it does not
The prescan is arithmetic over the text. Every figure is read with its currency, scale and unit, so $485M, $485MM and $485,000,000 are one number. Each is filed under the metric named nearest to it and the period next to it (or its table column's header), and the same metric for the same period is compared across slides at the precision each figure is written to ($0.5B and $485M agree; $318M and $305M do not). Growth "from A to B" is recomputed, and so are CAGRs with both years, margins with their revenue and EBITDA, and table Total rows. A "#1" or "leading" claim is set against revenue over the stated market. The refresh finds each old figure across scale, precision and unit style, including bare numbers under a $M header or on a chart axis, and writes the new value in that instance's style.
It reads text, not pixels: overlaps, overflow, fonts and contrast still need a person looking at the slides, and it cannot write your .pptx back. It knows nothing the deck does not say, so a figure that is wrong everywhere is not caught. The QC pass and the refresh review explain and challenge. They do not give investment advice. Derived from the agent skills @anthropics/ib-check-deck and @anthropics/deck-refresh (anthropics/financial-services-plugins, Apache-2.0). The example companies are fictional.