Skip to main content

Everything it takes to publish a number you can defend

Verification is the foundation — but a report isn't finished when it's generated. Bringing data in, editing precisely, reviewing as a team, signing off, exporting and keeping it current are all part of the product.

Verification

Every number checked in code, not by another model

This is the product. A second AI reviewing the first one's output is still a guess — validation here is deterministic, and it runs before anything reaches a slide.

An immutable snapshot of your data

Every numeric value in your file is captured in a canonical snapshot the moment it's parsed — before any AI call runs. Your values are never modified.

Deterministic validation

Numbers the AI returns are diffed against that snapshot in application code. Anything that doesn't match your file is rejected, not published.

Explainable figures

Hover any number in the deck to see how it was derived, and open the actual source rows behind it without leaving the slide.

Derived metrics computed in code

Growth rates, cash runway, net-new ARR and LTV:CAC are calculated deterministically, then handed to the model as verified ground truth — never invented by it.

Your own edits are checked too

Text you type into a slide runs through the same validation. Type a figure that isn't in your data and the edit is refused, with the unverified number flagged as you type.

Understanding before generation

The AI has to show you how it read every column — with confidence scores — and you correct it before a single slide exists.

Bring your data

Start from the file you already maintain

No restructuring, no schema to define. Time series, snapshots and multi-series data are detected during parsing, deterministically.

Up to five files at once

Combine several exports into one deck — revenue from finance, engagement from product. Each source keeps its own lineage.

Multi-sheet workbooks

Every tab in an Excel workbook is picked up as its own data source, including long-format and metric-matrix layouts.

Link a Google Sheet or hosted file

Paste a published Google Sheet CSV link or any hosted CSV/Excel URL. The deck can then be refreshed in place when the source changes.

Import from Google Drive

Pick a file with the Google Picker. Access is granted per file — the narrow drive.file scope, so nothing else in your Drive is readable.

Five framing presets

Board review, QBR, sales review, financial review or investor update. The preset changes persona, narrative arc and which metrics lead — never the numbers.

Straight to a draft when the read is clear

When column inference is confidently correct, the mapping step is skipped and your deck is already generating. You can still review everything.

Shape the story

Edit it like a deck, not like a prompt

Direct manipulation where you want precision, AI where you want speed — and the integrity guarantee holds either way.

Click any text to edit it

Titles, headlines, takeaways, bullets and captions are editable in place. No regeneration, no waiting for a model.

Insert a verified figure

Press ⌘K for a searchable list of numbers that provably came from your data, and drop one straight into the sentence you're writing.

Choose the chart and the series

Switch between bar, line, pie and scatter, and pick exactly which columns are plotted — read straight from your source, with no AI round trip.

Restructure freely

Reorder by dragging, duplicate, remove, or ask for a new slide on a specific topic.

Ask a slide a question

Query the data behind any slide in plain language. Answers that would change one of your numbers are blocked before they reach you.

Undo and redo

Unsaved changes undo instantly in the browser; saved versions step backward and forward without ever destroying history.

Version history you can restore

Every save is a new version with its author. Restoring an older one appends it as the latest — nothing is overwritten or lost.

Templates and your brand

Pick a template or build a brand kit — accent colour, background, typeface and your logo — applied across the editor, presenting, sharing and every export.

Present without exporting

Full-screen presentation mode with keyboard navigation and speaker notes, straight from the browser.

Export

One deck, every format the room expects

The same validated content, rendered for whoever is receiving it — including a workbook built specifically for the person who wants to check your work.

Verification workbook (.xlsx)

Every figure on every slide traced to the cell it came from — and honestly marked unverified where it can't be. Hand it to the sceptic in the room.

PowerPoint (.pptx)

A real, fully editable PPTX with native charts — not an image dump. Optionally append data-lineage backup slides showing how each number was computed.

PDF

A fixed, send-anywhere version for board packs and email attachments.

Word and Markdown memo

The same findings as a written narrative report for readers who want prose, not slides. Rendered directly from the deck — no second AI pass to drift from it.

Google Slides

Push the deck straight into Google Slides when your team lives there.

A drafted executive email

Generate the accompanying note. If any figure in it can't be verified, you get no draft rather than an unverified one.

Collaborate

From draft to signed off

Reporting is a review cycle, not a one-shot generation. The parts that follow the first draft are built in.

Shared workspaces

Invite your team by email, move decks into a shared workspace, and switch between personal and team views.

Comments on a slide

Reviewers leave notes against the exact slide and resolve them when addressed. Comments never quietly regenerate the deck.

Sign-off that can't go stale

Approval is tied to a specific version, so any edit clears the badge automatically. An approved deck is always the deck that was approved.

Know what's waiting on you

An in-app bell separates what actually needs your action from general activity, and re-derives it live so a cancelled request stops nagging.

Batched email digests

Comment notifications arrive as one digest per deck instead of a message per remark. Approval and comment mail are independently mutable.

Nobody overwrites anybody

If a teammate saved while you were editing, you're told before your change lands — instead of silently discarding theirs.

Share

Send a link, see what landed

Share the live deck instead of emailing a file that's out of date the moment it's sent.

Public share links

Publish a read-only view at its own URL, with its own presentation mode. Unpublish at any time.

Password protection

Require a password before the deck opens.

Expiring links

Set a link to stop working after a number of days — useful for diligence and anything time-boxed.

Per-slide attention

See which slides readers actually dwelled on. Your own views are excluded, and no IP addresses or personal details are stored.

A check before you publish

Publishing runs a scan for things that usually shouldn't leave the building — email addresses, phone numbers, identifiers, compensation. It's advisory: it flags, you decide.

Automate

The recurring report that files itself

Monthly reporting is the same work every month. For decks built from a linked source, it can simply happen.

Scheduled refresh

Put a deck built from a linked Google Sheet, hosted file or Drive file on a daily, weekly or monthly cadence. Uploaded files can't be scheduled — there's nothing to re-read.

Delivered, with what moved

Each run emails up to ten recipients the refreshed deck and a PDF, with a summary of what changed since the previous run.

A nudge when the source changes

If the underlying file has been updated since your deck was generated, you're told while you're looking at it.

Refresh in place

Re-pull the source and regenerate at the same URL, so the link you already sent stays the right one.

FAQ

Feature questions

What makes this different from an AI presentation generator?+

The validation layer. Generic AI slide tools ask a model for numbers and print what comes back. InsightNarrative snapshots every value in your file before any AI call, then diffs the model's output against that snapshot in application code. A number that isn't in your data cannot reach a slide — and that check runs on text you type yourself too.

Which file types and data shapes are supported?+

CSV, TSV, XLSX and XLS, up to five files per deck. Multi-sheet workbooks are supported, with each tab treated as its own data source. Time series, point-in-time snapshots, multi-series and metric-matrix layouts are all detected deterministically during parsing — you don't restructure anything.

What formats can I export to?+

PowerPoint (a real editable PPTX with native charts, optionally with data-lineage appendix slides), PDF, Word, Markdown, Google Slides, and an Excel verification workbook that traces every figure on every slide back to its source cell.

Can my team review and approve a deck before it goes out?+

Yes. Invite your team to a shared workspace, leave comments on specific slides, and request sign-off. Approval is tied to a specific version, so any subsequent edit clears the badge automatically — an approved deck is always the deck that was actually approved.

Can a deck update itself when the underlying data changes?+

For decks built from a linked source — a published Google Sheet, a hosted CSV or Excel URL, or a Google Drive file — yes. You can schedule a daily, weekly or monthly refresh that regenerates the deck and emails it with a summary of what changed. Decks built from a one-off file upload can't be scheduled, because there is nothing to re-read.

Do I need an account to try it?+

No. You can upload a file or run a sample dataset and generate a deck without signing up. Signing in is only required to save decks to an account, collaborate in a workspace, or schedule a refresh.

See it on your own numbers

Upload a file and check the generated deck against your spreadsheet. Every figure is validated in code — that's the whole point.