Did we ever do a transfer slab like this?
A graduate is designing a transfer slab above a car park. It seems likely the practice has done something similar before. They ask a senior engineer, who thinks there was a job in Leeds, maybe six years ago, and suggests asking another director. Twenty minutes later someone finds a job number, and the graduate spends the afternoon opening folders trying to find the calc.
Every practice has years of solved problems in its archive. It is organised by job number and client name, which is great if you already know which job you want and useless if you do not.
Why past work is so hard to find
- The archive is a file server or SharePoint organised by job number.
- Project data such as structure type, ground conditions and building use is not recorded anywhere searchable.
- Calcs are PDFs whose text is not indexed, or scans with no text at all.
- Knowledge sits with long-serving engineers, who are the busiest and who will eventually retire.
- Search in Windows or SharePoint finds file names, not engineering content.
The archive also hides the practice's own standards. Over the years, engineers develop preferred details for things like balcony connections or movement joints. When nobody can find the last version, each project reinvents it, and small differences creep in that make checking harder and drawings less consistent.
What that costs the practice
| Effect | Where it shows |
|---|---|
| Problems solved again from scratch | Design time consumed, stage fees burn |
| Senior engineers interrupted | Their own work and checking slow down |
| Inconsistent details between projects | More risk of errors, harder checking |
| Knowledge leaves with people | Practice becomes weaker when staff move on |
How we build archive search
- We index the archive where it sits, on your file server, SharePoint or document system, without moving files.
- Text is extracted from PDFs, including scanned calcs and reports through OCR, and from Word and Excel files.
- Each project is tagged with key characteristics from your project records, and a language model suggests further tags such as structural form or ground conditions from the documents, for a person to confirm.
- Engineers search in plain words, such as 'piled raft on made ground' or 'steel moment frame retrofit', and see matching documents with project, date and author.
- Results show the relevant page and a short summary, with a link to the original file.
- Existing access permissions are respected, so people only see what they could already open.
- New projects are indexed automatically as they are filed.
The search points engineers to past work. It does not decide whether that work applies. Codes, loads and site conditions change, and the engineer still has to design for the job in front of them.
We usually test on a sample of the archive first, a few years of projects across your main sectors, and let engineers try real searches. That shows how well OCR copes with your older material and which project tags are worth adding before the whole archive is indexed.
What it changes for the office
A graduate finds three similar transfer slabs in a few minutes and reads the calcs before asking anyone. When they do ask a senior engineer, it is a better question.
Directors see a practice memory that does not depend on who happens to be in the office.
Is this your archive?
- Finding past work means asking the most experienced engineer.
- The archive is organised only by job number.
- Scanned calcs cannot be searched.
- Similar problems are solved differently on different projects.
- You worry about knowledge leaving when senior staff retire.