diff --git a/CHANGELOG.rst b/CHANGELOG.rst
index 7601678597..969adcd7e4 100644
--- a/CHANGELOG.rst
+++ b/CHANGELOG.rst
@@ -17,6 +17,10 @@ Unreleased
 ----------
 * nothing unreleased
 
+[4.21.0]
+---------
+* refactor: replace openai usage with xpert api
+
 [4.20.14]
 ---------
 * chore: Update requirements
diff --git a/enterprise/__init__.py b/enterprise/__init__.py
index 39d3c34b3a..77b47b0188 100644
--- a/enterprise/__init__.py
+++ b/enterprise/__init__.py
@@ -2,4 +2,4 @@
 Your project description goes here.
 """
 
-__version__ = "4.20.14"
+__version__ = "4.21.0"
diff --git a/enterprise/api_client/open_ai.py b/enterprise/api_client/open_ai.py
deleted file mode 100644
index c4c15af1e4..0000000000
--- a/enterprise/api_client/open_ai.py
+++ /dev/null
@@ -1,31 +0,0 @@
-"""
-OpenAI client
-"""
-
-import openai
-
-from django.conf import settings
-
-openai.api_key = settings.OPENAI_API_KEY
-
-
-def chat_completion(prompt, role, model='gpt-3.5-turbo'):
-    """
-    Use chatGPT https://api.openai.com/v1/chat/completions endpoint to generate a response.
-
-    Arguments:
-        prompt (str): ChatGPT prompt
-        role (str): ChatGPT role to assume for the prompt.
-        model (str): ChatGPT model to use for the prompt completion.
-
-    Returns:
-        (str): Prompt response from OpenAI.
-    """
-    response = openai.ChatCompletion.create(
-        model=model,
-        messages=[
-            {'role': role, 'content': prompt},
-        ]
-    )
-
-    return response['choices'][0]['message']['content']
diff --git a/enterprise/api_client/xpert_ai.py b/enterprise/api_client/xpert_ai.py
new file mode 100644
index 0000000000..9686e004d1
--- /dev/null
+++ b/enterprise/api_client/xpert_ai.py
@@ -0,0 +1,40 @@
+"""
+Xpert client
+"""
+
+import json
+
+import requests
+
+from django.conf import settings
+
+CONNECT_TIMOUET_SECONDS = 5
+READ_TIMEOUT_SECONDS = 20
+
+
+def chat_completion(prompt, role):
+    """
+    Generate response using xpert api.
+
+    Arguments:
+        prompt (str): ChatGPT prompt
+        role (str): ChatGPT role to assume for the prompt.
+
+    Returns:
+        (str): Prompt response from OpenAI.
+    """
+    headers = {
+        'Content-Type': 'application/json',
+        'x-api-key': settings.CHAT_COMPLETION_API_KEY
+    }
+
+    body = {'message_list': [{'role': role, 'content': prompt},]}
+
+    response = requests.post(
+        settings.CHAT_COMPLETION_API,
+        headers=headers,
+        data=json.dumps(body),
+        timeout=(CONNECT_TIMOUET_SECONDS, READ_TIMEOUT_SECONDS)
+    )
+
+    return response.json().get('content')
diff --git a/enterprise/models.py b/enterprise/models.py
index ae05e2fc34..67566c8f1c 100644
--- a/enterprise/models.py
+++ b/enterprise/models.py
@@ -44,8 +44,8 @@
 from enterprise.api_client.ecommerce import EcommerceApiClient
 from enterprise.api_client.enterprise_catalog import EnterpriseCatalogApiClient
 from enterprise.api_client.lms import EnrollmentApiClient, ThirdPartyAuthApiClient
-from enterprise.api_client.open_ai import chat_completion
 from enterprise.api_client.sso_orchestrator import EnterpriseSSOOrchestratorApiClient
+from enterprise.api_client.xpert_ai import chat_completion
 from enterprise.constants import (
     ALL_ACCESS_CONTEXT,
     AVAILABLE_LANGUAGES,
diff --git a/enterprise/settings/test.py b/enterprise/settings/test.py
index c2c9032b14..b98da6eef5 100644
--- a/enterprise/settings/test.py
+++ b/enterprise/settings/test.py
@@ -353,7 +353,8 @@ def root(*args):
 # disable indexing on history_date. Otherwise it will add new alter migrations.
 SIMPLE_HISTORY_DATE_INDEX = False
 
-OPENAI_API_KEY = ''
+CHAT_COMPLETION_API = 'https://example.com/chat/completion'
+CHAT_COMPLETION_API_KEY = 'i am a key'
 
 LEARNER_ENGAGEMENT_PROMPT_FOR_ACTIVE_CONTRACT = 'Create learner engagement report for active contracts.'
 LEARNER_ENGAGEMENT_PROMPT_FOR_NON_ACTIVE_CONTRACT = 'Create learner engagement report for non active contracts.'
diff --git a/tests/test_enterprise/api/test_views.py b/tests/test_enterprise/api/test_views.py
index 4435bf2e2f..e7ee9063db 100644
--- a/tests/test_enterprise/api/test_views.py
+++ b/tests/test_enterprise/api/test_views.py
@@ -7423,12 +7423,16 @@ def test_view_with_admin_user_tries(self):
         assert response.status_code == status.HTTP_403_FORBIDDEN
         assert response.json() == {'detail': 'Missing: enterprise.can_access_admin_dashboard'}
 
-    @mock.patch('enterprise.models.chat_completion')
-    def test_view_with_admin_user(self, mock_chat_completion):
+    @mock.patch('enterprise.api_client.xpert_ai.requests.post')
+    def test_view_with_admin_user(self, mock_post):
         """
         Verify that an enterprise admin user having `enterprise.can_access_admin_dashboard` role can access the view.
         """
-        mock_chat_completion.return_value = 'Test Response.'
+        xpert_response = 'Response from Xpert AI'
+        mock_response = mock.Mock()
+        mock_response.json.return_value = {'content': xpert_response}
+        mock_post.return_value = mock_response
+
         self.set_jwt_cookie(ENTERPRISE_ADMIN_ROLE, self.enterprise_uuid)
 
         self.client.login(username=self.user.username, password=TEST_PASSWORD)
@@ -7436,8 +7440,8 @@ def test_view_with_admin_user(self, mock_chat_completion):
 
         response = self.client.post(self.url, data=self.payload, format='json')
         assert response.status_code == status.HTTP_200_OK
-        assert 'learner_progress' in response.json()
-        assert 'learner_engagement' in response.json()
+        assert response.json()['learner_progress'] == xpert_response
+        assert response.json()['learner_engagement'] == xpert_response
 
     @mock.patch('enterprise.models.chat_completion')
     def test_404_if_enterprise_customer_does_not_exist(self, mock_chat_completion):