Direct answer
Can AI extract quantities or line items from project documents?
A practical answer for Singapore AEC firms deciding whether this workflow is worth testing.
Where it helps
Good automation starts where rules and repetition are visible.
These are the parts of the workflow that can usually become a controlled first pass with review gates.
Comparing document versions
Flagging missing fields
Preparing a review spreadsheet
Separating confident and uncertain extraction
Matching items against a template
Where it should stay human
Judgement and accountability stay with the firm.
Praeco's pattern is preparation, checking and review support. The accountable professional still decides what goes out.
Measurement judgement
Ambiguous scope interpretation
Contractual reliance
Final BOQ approval
What to test first
Use one anonymised document set with the expected output spreadsheet so accuracy can be checked directly.
The workflow diagnostic maps the current process before any separately scoped paid proof.
What Praeco would need to see
Praeco would need sample documents, the target spreadsheet format, known edge cases, and the review rules.
Hands-on testing with these files belongs to a separately scoped paid proof. Client details can be removed or anonymised; the important part is seeing the real inputs, output format and review rule.
Next step
Start with the workflow diagnostic.
Send one repeated workflow and Praeco will check whether it is a good candidate before proposing any build.