diff --git a/process_report/process_report.py b/process_report/process_report.py
index 7553e4d..2624276 100644
--- a/process_report/process_report.py
+++ b/process_report/process_report.py
@@ -60,12 +60,30 @@ def load_old_pis(old_pi_file):
     return old_pi_dict
 
 
+def dump_old_pis(old_pi_file, old_pi_dict: dict):
+    with open(old_pi_file, "w") as f:
+        for pi, first_month in old_pi_dict.items():
+            f.write(f"{pi},{first_month}\n")
+
+
 def is_old_pi(old_pi_dict, pi, invoice_month):
-    if pi in old_pi_dict and old_pi_dict[pi] != invoice_month:
+    first_invoice_month = old_pi_dict.get(pi, invoice_month)
+    if compare_invoice_month(first_invoice_month, invoice_month):
+        sys.exit(
+            f"PI {pi} from {first_invoice_month} found in {invoice_month} invoice!"
+        )
+    if compare_invoice_month(invoice_month, first_invoice_month):
         return True
     return False
 
 
+def compare_invoice_month(month_1, month_2):
+    """Returns True if 1st date is later than 2nd date"""
+    dt1 = datetime.datetime.strptime(month_1, "%Y-%m")
+    dt2 = datetime.datetime.strptime(month_2, "%Y-%m")
+    return dt1 > dt2
+
+
 def get_invoice_bucket():
     try:
         s3_resource = boto3.resource(
@@ -103,7 +121,7 @@ def main():
     parser.add_argument(
         "--upload-to-s3",
         action="store_true",
-        help="If set, uploads all processed invoices to S3",
+        help="If set, uploads all processed invoices and old PI file to S3",
     )
     parser.add_argument(
         "--invoice-month",
@@ -159,16 +177,20 @@ def main():
     parser.add_argument(
         "--old-pi-file",
         required=False,
-        help="Name of csv file listing previously billed PIs",
+        help="Name of csv file listing previously billed PIs. If not provided, defaults to fetching from S3",
     )
     args = parser.parse_args()
 
     invoice_month = args.invoice_month
 
     if args.fetch_from_s3:
-        csv_files = fetch_S3_invoices(invoice_month)
+        csv_files = fetch_s3_invoices(invoice_month)
     else:
         csv_files = args.csv_files
+    if args.old_pi_file:
+        old_pi_file = args.old_pi_file
+    else:
+        old_pi_file = fetch_s3_old_pi_file()
 
     merged_dataframe = merge_csv(csv_files)
 
@@ -192,7 +214,7 @@ def main():
 
     billable_projects = remove_non_billables(merged_dataframe, pi, projects)
     billable_projects = validate_pi_names(billable_projects)
-    credited_projects = apply_credits_new_pi(billable_projects, args.old_pi_file)
+    credited_projects = apply_credits_new_pi(billable_projects, old_pi_file)
 
     export_billables(credited_projects, args.output_file)
     export_pi_billables(credited_projects, args.output_folder, invoice_month)
@@ -211,9 +233,10 @@ def main():
 
         upload_to_s3(invoice_list, invoice_month)
         upload_to_s3_HU_BU(args.HU_BU_invoice_file, invoice_month)
+        upload_to_s3_old_pi_file(old_pi_file)
 
 
-def fetch_S3_invoices(invoice_month):
+def fetch_s3_invoices(invoice_month):
     """Fetches usage invoices from S3 given invoice month"""
     s3_invoice_list = list()
     invoice_bucket = get_invoice_bucket()
@@ -290,7 +313,7 @@ def remove_billables(dataframe, pi, projects, output_file):
 def validate_pi_names(dataframe):
     invalid_pi_projects = dataframe[pandas.isna(dataframe[PI_FIELD])]
     for i, row in invalid_pi_projects.iterrows():
-        print(f"Warning: Project {row[PROJECT_FIELD]} has empty PI field")
+        print(f"Warning: Billable project {row[PROJECT_FIELD]} has empty PI field")
     dataframe = dataframe[~pandas.isna(dataframe[PI_FIELD])]
 
     return dataframe
@@ -316,6 +339,8 @@ def apply_credits_new_pi(dataframe, old_pi_file):
             for i, row in pi_projects.iterrows():
                 dataframe.at[i, BALANCE_FIELD] = row[COST_FIELD]
         else:
+            old_pi_dict[pi] = invoice_month
+            print(f"Found new PI {pi}")
             remaining_credit = new_pi_credit_amount
             for i, row in pi_projects.iterrows():
                 project_cost = row[COST_FIELD]
@@ -329,9 +354,23 @@ def apply_credits_new_pi(dataframe, old_pi_file):
                 if remaining_credit == 0:
                     break
 
+    dump_old_pis(old_pi_file, old_pi_dict)
+
     return dataframe
 
 
+def fetch_s3_old_pi_file():
+    local_name = "PI.csv"
+    invoice_bucket = get_invoice_bucket()
+    invoice_bucket.download_file("PIs/PI.csv", local_name)
+    return local_name
+
+
+def upload_to_s3_old_pi_file(old_pi_file):
+    invoice_bucket = get_invoice_bucket()
+    invoice_bucket.upload_file(old_pi_file, "PIs/PI.csv")
+
+
 def add_institution(dataframe: pandas.DataFrame):
     """Determine every PI's institution name, logging any PI whose institution cannot be determined
     This is performed by `get_institution_from_pi()`, which tries to match the PI's username to
diff --git a/process_report/tests/unit_tests.py b/process_report/tests/unit_tests.py
index 1aa1428..baa58f3 100644
--- a/process_report/tests/unit_tests.py
+++ b/process_report/tests/unit_tests.py
@@ -328,6 +328,16 @@ def test_apply_credit_0002(self):
         self.assertEqual(0, credited_projects.loc[4, "Balance"])
         self.assertEqual(800, credited_projects.loc[5, "Balance"])
 
+        updated_old_pi_answer = "PI2,2023-09\nPI3,2024-02\nPI4,2024-03\nPI1,2024-03\n"
+        with open(self.old_pi_file, "r") as f:
+            self.assertEqual(updated_old_pi_answer, f.read())
+
+    def test_apply_credit_error(self):
+        old_pi_dict = {"PI1": "2024-12"}
+        invoice_month = "2024-03"
+        with self.assertRaises(SystemExit):
+            process_report.is_old_pi(old_pi_dict, "PI1", invoice_month)
+
 
 class TestValidateBillables(TestCase):
     def setUp(self):