Glasshouse
Your organization's documents are unsearchable. Glasshouse extracts, indexes and makes them findable — with AI that actually reads the pages.
The problem
Organizations accumulate documents across dozens of systems, and most become unfindable within months.
- Scanned documents — contracts, letters and forms scanned as images. No search engine can read them, so they might as well not exist.
- Email attachments — critical documents buried in inboxes. When someone leaves, their attachments leave too.
- SharePoint and shared drives — the document graveyard. Uploaded with good intentions, never found again.
What it does
Glasshouse is a layer that sits on top of wherever your documents already live. It connects to SharePoint, email, network drives and cloud storage; extracts text with GPU-powered OCR that copes with tables, handwriting and bad faxes; indexes it twice, for exact matches and for meaning; and lets people search and ask in plain language, with every answer citing the page it came from.
Proven on BC government FOI records
Thousands of Freedom of Information pages that the government's own search cannot find. Glasshouse finds them — because the content existed only inside scanned images that no search engine was reading.
Features
- Multi-tier OCR — digital text, GPU OCR, table-aware extraction, and cloud vision reserved for the hardest pages
- Dual search — keyword for exact matches, semantic for meaning. "End of life care" finds MAID documents.
- Conversational Q&A — grounded answers with citations to specific pages in specific documents
- On-demand extraction — admins trigger cloud vision per page, with cost estimates up front
- Table export — structured tables out as HTML or CSV, turning scanned spreadsheets back into data
- Cross-references — related documents surfaced across departments and time periods
- Admin dashboard — coverage, extraction methods and spend at a glance
- Bulk ingest — point it at a source and pull everything in, with live progress
What it isn't
Glasshouse does one thing well, and it helps to be clear about the rest.
- Not document backup — it indexes and searches; it doesn't replace where files live
- Not a primary repository — keep using SharePoint, Drive or your file server. It sits alongside.
- Not customer-facing — this is an internal tool for your own team
- Not a generic chatbot — every answer is grounded in your documents, with citations
- Not cloud-only — deploy on-premises if your data can't leave the building
How it's sold
You licence the tool and load your own documents — Glasshouse doesn't host or supply any content. The BC FOI corpus is a proof point, not a product.
There's a fuller walk-through in the deck, including what the search results actually look like against a real corpus.
Deployment can be hosted or fully on-premises. If you need something bespoke on top, that's the other half of what this studio does.