How it works — Netrasya Lens

Step 01

Send your hardest reports

Five to ten of your densest project PDFs — the assessments, hydrogeological and geotechnical reports, borehole logs and spec sheets your current tools choke on.

Step 02

We stand up your search

We parse and index them in a dedicated, single-tenant environment in Canada. Nothing touches ChatGPT, Copilot, or any public AI service at any point.

Step 03

Your team asks, and checks

Ask in plain English, or pull one field across every report into a table. Every answer cites its page — and if it isn't in your reports, it says so.

Why not just use a chatbot

It cites, or it says it can't.

Every claim below carries a citation, the same way an answer does. Open one to see what stands behind it.

	Netrasya Lens	A general assistant
What it reads	Your firm's own archive — project files, data rooms, decades of PDFs.	The public web, plus whatever happens to sit in the connected drive.
Source Point it at years of project files and data rooms, not just what lives in SharePoint. Nothing from the public web enters an answer — the retrieval set is your documents and only your documents.
How it answers	Plain English, with the exact document and page number attached.	Fluent prose, with no way to check where it came from.
Source Every claim carries a filename and page you can open. That inverts the trust question: you verify the answer instead of deciding whether to believe it — which is the difference between a tool you can put behind a stamped report and one you can't.
When the answer isn't there	It says so.	It answers anyway.
Source When your documents don't contain the answer, it returns "not in these documents" rather than composing something plausible. This is the behaviour we guard hardest, because a confident wrong number is worse than no number. The fourth question in the demo on the home page shows it.
Dense tables	Borehole logs, lab results and allocation grids keep their structure.	Flattened into a run of numbers you can't line back up.
Source Layout-aware parsing: nested headers, multi-column allocation tables and lab grids stay tables rather than collapsing into a paragraph of digits. This is the part general tools get wrong on technical reports, and the reason an assistant that reads your email still can't read your ESA.
Where it runs	A single-tenant environment we operate in Canada.	Wherever the vendor runs it, on terms you don't set.
Source AWS ca-central-1. We say stored and processed in Canada rather than "never leaves Canada", because the US CLOUD Act reaches Canadian data centres and no vendor can honestly promise otherwise. What we do promise is checkable: no third-party AI service sits in the path, and your documents are never used to train a model.

Straight answers

Questions we get from principals.

What does it cost?

The pilot is a fixed fee, agreed in writing before anything starts, and credited to your first month if you continue. After that it's one flat monthly fee for the whole firm — never per seat, never per page. We don't publish the number because a twelve-person shop and an eighty-person shop shouldn't pay the same. Tell us your team size on a short call and you'll have a figure inside that call.

We already have Microsoft Copilot.

Copilot is good at email and Office documents, but it routinely flattens the nested tables and spec grids in technical reports, and it only sees what's in SharePoint — not your archives and data rooms. Netrasya Lens is built to preserve those layouts, and it's one flat fee for the firm rather than a per-seat line item.

Couldn't our IT person self-host an open-source tool?

They could. But standing up the infrastructure, tuning the parsing for technical tables, and keeping it running is dozens of hours of senior time at a billable rate — and it becomes one more system your firm owns. We run the whole thing, managed and in Canada, for a flat monthly fee.

How accurate is it, and who else uses it?

We're launching out of Kitchener, so you'd be one of our first local partners. That's exactly why the founder integrates your toughest documents personally during the pilot — so you verify the accuracy on your own reports, cited page by page, before you sign anything ongoing.

Is our data actually safe?

Your documents live in a dedicated, single-tenant environment we operate in Canada, are never sent to any public AI service, and are never used to train a model. If you don't continue after the pilot, they're deleted.

Can I control who sees which documents?

Yes. Access is set document by document rather than all-or-nothing, so confidential client matters stay visible only to the people you choose — and every search returns only what that person is cleared to see.

From a 500-page archive to a cited answer.