This is the surface for anything your AI puts on a slide. It runs a deterministic donkey that checks every chart against its data and every slide for one clear claim, so the point lands instead of hiding in the deck. It reviews the slide; it does not redraw it for you.
A slide can be polished and still mislead, or say nothing at all. The donkey names which.
A truncated axis that exaggerates a trend, a chart type that hides the comparison, a number with no source. The chart says more, or less, than the data supports.
The title is a label ("Q3 Revenue") instead of the finding ("Q3 revenue grew on one product"). The reader has to do the work the slide was supposed to do.
The insight exists in the data on the slide but nothing surfaces it: no callout, no contrast, no ordering. The audience never finds what you wanted them to see.
Inkwell, with a slide-hygiene check. Deterministic: same chart and data in, the same review out. In preview.
Why: a model will happily draw a chart that the underlying numbers do not support.
How: it reviews the chart against the data behind it (axis honesty, right chart type,
sourced numbers) and checks the slide for a single stated claim and a title that carries the finding.
An AI-built slide with a flattering chart and a label for a title. The donkey calls it.
$ check slide.json
verdict: REVISE chart: bar claims: 0
AXIS y-axis starts at 80, not 0. The 4% rise looks like 60%.
fix: start at zero, or switch to an indexed line.
SOURCE no source on the revenue figures.
fix: cite the row the numbers came from.
TITLE "Q3 Revenue" is a label, not a finding.
fix: state the point, e.g. "Q3 revenue grew, on one product."
Same slide in, the same review out, every time.
Call the donkey on a slide, or run the surface inside a machine that remembers. The difference is state.
Send a chart and its data, get the review back. Stateless and simple: same slide in, the same findings out. Run it on a deck whenever you want. Nothing to set up, nothing remembered.
Connect your own AI to the doloop machine in presentation mode and the donkey runs inside the loop: your AI drafts a slide, the donkey reviews it, your AI fixes the axis or the title, and only honest slides ship. The machine learns your house chart rules and holds them.
Want this on your reporting pipeline? Talk to us, or see the other surfaces.