diff --git a/RESUME.md b/RESUME.md
index 7778b83..d4d93bb 100644
--- a/RESUME.md
+++ b/RESUME.md
@@ -195,9 +195,17 @@ Homebrew. No Access ODBC driver, `node_modules` not installed, staging Parquet n
      SCOTHIA DATOS I/E as signed amounts (income +, expense -; net +899,375.77), 66
      `business_line_categories`. No customer FK; categoryId left null (concept->ramo classifier
      is a later enhancement).
-   - **NEXT:** document extraction (step 4) — pull the LONGBINARY blobs (DATMEX doc_1/2, policy
-     docs_1/2/foto1, DATOS docs) to object storage + write *_documents rows. Then the Customer
-     module API/web (Spanish-first).
+   - **Documents — DONE** (`migration/blob_extract.py`, migration step 4): carves embedded files
+     out of the Access OLE wrapper (magic-byte detection) and uploads to MinIO on cubex (stack
+     `jorgecuadros-dev-minio`, S3 at 192.168.4.212:9100, bucket jorgecuadros-documents), writing
+     service_documents/policy_documents pointer rows. Loaded 70 documents (3 service bills +
+     67 policy foto/docs, 0 orphans, ~290 MB). **Data finding:** the LONGBINARY columns are
+     almost entirely empty — DATMEX's real bill-scan columns are ILUZ/IAGUA/IPREDIAL/ITEL (not
+     doc_1/doc_2), but only 3 cells populated across 1520 rows; the *_MENS tables are mail-merge
+     templates (correctly excluded). The 538MB/882MB source files are mostly Access bloat, not
+     documents. **Migration steps 1-4 COMPLETE.**
+   - **NEXT:** the Customer module API/web (Spanish-first) — all relational data + documents are
+     loaded, so build the unified customer list/search/detail view.
    - Full pipeline reproducible in one command: `run_all.py --env <env>` runs customers ->
      properties -> policies -> transactions -> bank in order (all idempotent); add `--stage`
      to re-extract from the Access files first. Verified end-to-end against dev.
diff --git a/migration/blob_extract.py b/migration/blob_extract.py
index 4464063..e2e099d 100644
--- a/migration/blob_extract.py
+++ b/migration/blob_extract.py
@@ -42,8 +42,13 @@ SOURCE_ROOT = Path.home() / "Downloads" / "JorgeCuadros-Legacy"
 
 # (key, access_file, access_table, staged_table_name, blob_cols, model)
 SOURCES = [
+    # DATMEX's real scanned bills live in the ILUZ/IAGUA/IPREDIAL/ITEL invoice-
+    # image columns (per-service bill scans), not doc_1/doc_2 (which are empty).
+    # In practice only a handful are populated — the .accdb is mostly bloat.
     dict(key="datmex", file="UTILITIES.accdb", table="DATMEX", staged="DATMEX",
-         blobs=["doc_1", "doc_2"], model="service"),
+         blobs=["iluz", "iagua", "ipredial", "itel", "doc_1", "doc_2"], model="service",
+         doctypes={"iluz": "ELECTRIC_BILL", "iagua": "WATER_BILL",
+                   "ipredial": "PROPERTY_TAX_BILL", "itel": "PHONE_BILL"}),
     dict(key="mult", file="SEGUROS 16_be.mdb", table="MULT", staged="mult",
          blobs=["foto1", "docs_1", "docs_2"], model="policy"),
     dict(key="autos_ampl", file="SEGUROS 16_be.mdb", table="TABLA AUTOS AMPL",
@@ -171,7 +176,7 @@ def main():
                 prefix = "service" if src["model"] == "service" else "policy"
                 key = f"{prefix}/{parent}/{src['staged']}_{ri}_{col}.{ext}"
                 s3.put_object(Bucket=bucket, Key=key, Body=data, ContentType=ct)
-                dtype = col.upper()
+                dtype = src.get("doctypes", {}).get(col, col.upper())
                 if src["model"] == "service":
                     svc_rows.append((str(uuid.uuid4()), parent, dtype, key))
                 else: