Historicals keyed by hand
Years of reported numbers typed from filings, for the company and for every peer.
Models built on sourced numbers
Needl.ai pulls the historicals and comparables you name out of the filings, with a source on every figure. It reads your models however they are laid out, answers questions with the cells cited, and checks each number against the documents behind it.
Most of a model is typing, and typing is where errors get in
Before an analyst writes a word, the numbers go into a spreadsheet: historicals keyed from annual reports and transcripts, comparables gathered from a stack of peer filings. None of it is the analyst's edge. It is where wrong numbers enter, and once one is in, everything downstream inherits it.
A figure nobody can trace
When a reviewer asks where a number came from, the answer is another search through the filings.
Someone else's model
Checking a model you did not build means re-reading every document behind it.
Sourced numbers in, checked numbers out
The analyst keeps the structure, the assumptions and the view. Needl.ai does the work around the numbers: finding them, citing them and checking them.
Historicals from the filings
Name the companies and the metrics, and the figures come back in a table, each with its source.
Any workbook, read properly
Several tables on a sheet, merged headers and subtotal rows are read the way an analyst reads them.
Answers cited to the cell
Ask the model a question, and every figure in the answer names the sheet and cells it came from.
Checked against the documents
Each figure is set against the filing it should match, with the gap shown where they differ.
From the filings to a model you can check
How the numbers move through Needl.ai, from the documents your team trusts to a model any reviewer can trace.
-
01Connect the filings and the models
Annual reports, earnings calls and prospectuses are connected alongside the drives, data rooms and email where your models and broker research live. Access follows the permissions in each source system.
You getThe filings and the models in one place.
-
02Pull the historicals and the comps
Name the companies, the metrics and the years, such as revenue, EBITDA margin and P/E for FY24 to FY26. The figures come back as a table with a source on each, and export as an Excel workbook.
You getHistoricals without the typing.
-
03Ask the model
Needl.ai reads the workbook however it is laid out and answers in figures, each citing the sheet and cells it came from. Margins, coverage and leverage are calculated and shown with their inputs.
You getAn answer you can check in the cell.
-
04Compare versions and files
Put two versions of a model, or a model and its peers, into one question. The answer shows what moved, by how much, and in which cells.
You getEvery change between drafts, with its cells.
-
05Check it against the documents
Each figure is set against the filing or document it should match, with both numbers and the gap. Assumptions are marked validated, contradicted or unvalidated, and cells showing formula errors are called out.
You getThe gaps, found before the review.
Everything a model review needs
Every figure keeps its source, whether it came from a filing, a workbook or a calculation, so a review starts from evidence.
Cited on every figure
Each figure names the filing and page, or the workbook, sheet and cells, it came from.
Any layout
Several tables on one sheet, merged headers, notes and subtotal rows, read without reformatting.
Valuation comps
P/E, EV/EBITDA and margins for the companies you name, side by side in one table.
Ratios with their inputs
Margins, coverage, leverage and growth are calculated and shown with the figures they came from.
Formula errors called out
Cells that show an error such as #REF! or #DIV/0! are pointed out in the answer.
PDFs and workbooks together
A figure from a filing and a figure from a model can meet in the same calculation.
Excel out
Tables export as a structured workbook, so the numbers land where the desk works.
Follow-ups build on the work
A later question about the same workbook starts from the tables already read.
Permission-aware
Only the files each person can already open are read for them.
What changes in a model review
A sourceOn every figure pulled from a filing, where typed numbers used to carry none.
The cellBehind every answer about the model, where finding a figure used to mean scrolling the sheet.
The gapBetween the model and the filings, shown with both numbers, where a check used to mean re-reading the documents.
“We stopped second-guessing the output. Almost every answer comes back with a citation we can open and verify, so a review that used to take hours now takes minutes.”
In your cloud, with your permissions
Needl.ai can be deployed in your own AWS or Azure account, in your region, and every answer follows the permissions already set in each source system.
- Deployed in your cloud account, in your region
- Models read only for people who can open them
- A source on every figure
AICPA SOC 2
ISO 27001 certified
ISO 42001 in progress
GDPR compliant
CASA assessed
Frequently asked questions
What research and investment teams ask before bringing their models to Needl.ai.
Who builds the model?
Your analysts. Needl.ai supplies the sourced historicals and comparables as an Excel workbook, and reads, compares and checks the models your team builds. The structure, the drivers and the forecast stay with the analyst.
Can it read our models as they are?
Yes. Workbooks are read as they come, with several tables on a sheet, merged headers and subtotal rows, whether they are uploaded or sit in the drives and data rooms you connect.
How do we check a figure?
Each figure names where it came from: the filing and page, or the workbook, sheet and cells. Figures that are calculated are shown with the inputs they came from.
What if a number cannot be found?
It says so. A figure with no source to cite is reported as unsourced, and is never given a citation it does not have.
Does it change our workbook?
No. Needl.ai reads a copy of the file, so your workbook stays exactly as you saved it.
Where does it run?
In your own AWS or Azure account, in your region, or in Needl.ai's cloud to start.
See it check one of your models
Bring a model and the filings behind it, and see every figure traced and checked on your own documents.