What you can ask for.
You state the outcome. The question. The memo. The deck. Steamshovel does the hunting, the cross-referencing, the slide-by-slide assembly. The part only people can do, deciding what matters, stays yours.
| You ask for | Who runs it | What you get |
|---|---|---|
| Find the answer across hundreds of files | Paralegal · analyst · GC · journalist · researcher | Ranked list with verbatim quotes. Click any to open the source at the passage. |
| Draft the memo from these documents | Associate · GC · policy researcher | One-page memo with every claim cited back to its source rows. Edit, don't assemble. |
| Update the model with this month's actuals | Founder · FP&A · controller | Your Sheet, updated in place. New inputs overlaid, formulas recompute, derived charts refresh. Diff visible before save. |
| Run a sensitivity on these assumptions | FP&A · strategy · risk · founder | Scenario grid (low, base, high), the deltas at each cell, exportable as a Sheet tab or a slide. |
| Forecast Q4 reserves at 95% confidence | Actuary · FP&A · support-staffing manager | Forecast with the method named (GARCH, Cox, GLM as appropriate). Confidence band, assumptions, reproducible a year later. |
| Predict churn and explain the drivers | Risk analyst · CRO · revenue ops · actuary | Gradient-boosted model trained on your Sheet, calibrated probabilities per account, SHAP attributions ranking the features that drove each score. |
| Build the deck from these numbers | Founder · PM · FP&A · consultant | 12-slide deck with real charts, real tables, designer-grade layouts. Every slide cites the source data. |
| Add a slide showing customer churn YoY | Analyst · founder · operations | One slide. Chart and caption inserted at the cursor. Data linked to the source so it refreshes. |
| Capture data from your script | Ops · analyst with a one-off pipeline | Python client with an idempotency key. Batch sync, ad-hoc analysis. We don't run your production pipeline. |
Each row is a real workflow from legal, finance, actuarial, defense, and journalism trial sessions.
Sample output, FP&A: runway 14.3 months, monthly burn $487K, net new ARR $124K, churn 4.2%. Each number opens to its source row on click.
What it saves you
Two hundred files. One question that needs all of them read. Vendor contracts. Monthly status reports. Customer feedback. Design docs. The folder that lands on your screen when someone wants an answer by Monday. Here is the difference.
| The old way | With Steamshovel | |
|---|---|---|
| Time to do it | About twenty hours of careful work (five minutes per document) | About three seconds, end to end |
| Cost in billable time | Roughly $10,000 | Roughly $0.05 |
| What you get back | An annotated list and a paragraph | A ranked list with body text and a link to every source |
| How to check the answer | Re-read the source documents | Click the citation. The source opens to the row, the clause, the figure. |
| If you need to redo it next year | Hope the same person is still here | Same citation, same documents, same fingerprint. The answer reproduces. |
| Tasks you can complete per hour | About a dozen | Tens of thousands, each with a link back to the source |
What it doesn't do
- It doesn't replace your judgment. It finds the documents and writes the first draft. You decide which matter and what the memo argues. The hours saved are the hunting hours, not the thinking hours.
-
It doesn't see what you don't show it.
Each install is scoped to the files you point it at. Your files stay in your own cloud. How we handle your data.
Google Drive Microsoft OneDrive Zoho WorkDrive
- It doesn't pretend. Every claim has a citation to the rows or sentence that back it. If the citation doesn't support the claim, you see it before you ship.
Install Steamshovel.
Free. No salesperson, no waitlist, no account to create. Permission to read and write the file you open it in. Nothing else.
From the Google Workspace Marketplace. Microsoft 365 and Zoho coming.