Correspondence, submittals and drawing revisions
Identify project, document type, revision, attachments, owner and deadline without replacing the CDE.
From inbox to a review-ready record.
Test one high-volume case type with anonymised data. Measure routing, missing evidence, deadlines and cited guidance before connecting any live system.

Success is not a feature count. We agree the operating outcome first, then decide whether the evidence supports a wider rollout.
Start with the repeated work between email, attachments and approved procedures, then output a reviewable record for your existing register or system.
Identify project, document type, revision, attachments, owner and deadline without replacing the CDE.
Turn property, urgency, missing evidence and ownership into a review-ready case.
Answers cite approved documents. When evidence is insufficient, the system says so.
The company, volume and organisation are fictional. The n8n executions, extracted fields, RAG citations and review routing below come from the working local demo.
Public-works design, drawing submittals and contract-document control
Shared mailbox, Excel register and project folders
50–120 messages per week across document control, PMs and contract administration
5n8n cases executed
5 / 5acceptance cases passed
5versioned SOP files
0external actions performed
Accept test email and attachment metadata and retain an execution record.
Extract case, type, project, deadline and attachments; route incomplete, sensitive or low-confidence cases to people.
Search five versioned SOPs and show file, version, section, text and relevance. Stop when evidence is insufficient.
Modify, approve or reject before writing SQLite and CSV records. No external action is allowed before approval.


Portfolio boundary: this is a de-identified simulated business case, not a client engagement. Live Gmail/Outlook, CDE, SharePoint, CRM and ERP systems are not connected.
A smaller, fixed-scope starting point. Resolve column differences, duplicates and invalid values across two CSV or Excel files before considering scheduled imports, n8n or RAG.
Retain source context while standardising names, companies, email, phone and dates.
Merge on stable identifiers and never overwrite an ambiguous value silently.
Missing, invalid and conflicting values retain their reason and source file.
Deliver clean CSV, an Excel report, a change log and the Python workflow.


Up to two CSV or Excel files and 5,000 rows: consolidation, deduplication, common-field standardisation, review queue, clean master and change log.
Ask about data cleanupThis execution-tested case uses fictional data and contains no customer records. A live CRM or ERP import is outside this starting scope.
This is not a page with hard-coded outcomes. The local demo processes a test email, attachments, SOP retrieval, citations and human-review routing.
Before approval: zero outbound messages, zero deletion, zero production-system changes.
Confirm case types, current register, decision rules, exceptions and actions that must remain manual.
Use 3–20 de-identified examples in a fixed scope. No production account or live data is needed.
Use accuracy, missing-evidence detection, corrections and handling time to decide what happens next.
Your first email only needs the company, workflow, current tools and approximate weekly volume. Please do not attach confidential or customer data.
Send an enquirytimmyworkflow@gmail.com