Turning a Spreadsheet Into a Slide Deck With One Instruction
Drive's cross-app AI turns a sheet into a deck, a sheet into a report, or a doc's assumptions into a table — three specific flows, not an open-ended 'AI does everything' claim.
A common pitch in AI-native productivity tools right now is some version of "our AI can do anything across your files" — which sounds impressive and tends to mean, in practice, an open-ended prompt box with unpredictable results. Piskes Drive's cross-app AI takes a narrower, more specific approach: exactly three AI-assisted flows between document types, each doing one clearly defined transformation, rather than an unbounded "ask it anything" surface.
The three flows
The logic lives in frontend/apps/drive/src/lib/drive/cross-app-ai.ts, and it covers:
Docs to Sheets — pulling assumptions or figures mentioned in a document's prose into a structured table in a spreadsheet, turning "we're assuming 8% growth and a $50 CAC" buried in a paragraph into actual cells you can build a model on top of.
Sheets to Docs — the reverse direction, writing a narrative report from a spreadsheet's numbers, turning a table of figures into prose that describes what they mean.
Sheets to Slides — the flow this piece is named for: taking a spreadsheet's data and producing a deck built around it, so you're not manually copying numbers into slide text boxes and building charts a second time.
Why three, deliberately, and not eight
It would be straightforward to market this as full any-to-any AI conversion across every document type Drive supports. That's not what shipped, and the reason is worth stating plainly: each of these three flows is a well-scoped transformation with a clear, checkable output — take assumptions from prose, take a table, take a chart-worthy dataset. An "AI can convert anything to anything" claim sounds more impressive right up until you actually try to specify what it means for, say, a deck to become a spreadsheet in a way that's actually useful rather than just technically possible. Three specific, well-defined flows is a more credible claim than eight vague ones, and it's also just what got built and tested.
Worth knowing: five other cross-app conversions are deterministic, not AI
Separately from these three AI flows, Drive already ships five other cross-app conversions (in lib/office/cross-app.ts) that don't involve AI at all — deterministic transformations where the mapping from source to destination is exact and repeatable rather than model-generated. That's a meaningful distinction if you're deciding which conversion to trust for something that needs to be precise: the deterministic flows produce the same output every time from the same input, while the AI flows are drafting something new based on your data, closer to a first draft than an exact transcription.
How the AI flows stay safe
The model is never asked to produce a raw file-format snapshot for any of these three flows — it's asked for CSV, plain lines of text, or deck markdown, and Drive's own builders (the same ones used for file import) convert that into a real document. And like the "create from description" flow, all three cross-app AI conversions create a new file rather than overwriting an existing one, which is why they bypass the propose-and-approve queue used for edits to something that already exists — there's nothing at risk of being silently changed, because nothing existing is being touched.