diff --git a/lib/captain/base_editor_service.rb b/lib/captain/base_editor_service.rb
index dec8f741b..efed5b2cc 100644
--- a/lib/captain/base_editor_service.rb
+++ b/lib/captain/base_editor_service.rb
@@ -7,17 +7,11 @@ class Captain::BaseEditorService
   # 120000 * 4 = 480,000 characters (rounding off downwards to 400,000 to be safe)
   TOKEN_LIMIT = 400_000
   GPT_MODEL = Llm::Config::DEFAULT_MODEL
-  CACHEABLE_EVENTS = [].freeze
 
   pattr_initialize [:account!, :event!]
 
   def perform
-    return value_from_cache if value_from_cache.present?
-
-    response = send("#{event_name}_message")
-    save_to_cache(response) if response.present?
-
-    response
+    send("#{event_name}_message")
   end
 
   private
@@ -26,52 +20,10 @@ class Captain::BaseEditorService
     event['name']
   end
 
-  def cache_key
-    return nil unless event_is_cacheable?
-
-    return nil unless conversation
-
-    # since the value from cache depends on the conversation last_activity_at, it will always be fresh
-    format(::Redis::Alfred::OPENAI_CONVERSATION_KEY, event_name: event_name, conversation_id: conversation.id,
-                                                     updated_at: conversation.last_activity_at.to_i)
-  end
-
-  def value_from_cache
-    return nil unless event_is_cacheable?
-    return nil if cache_key.blank?
-
-    deserialize_cached_value(Redis::Alfred.get(cache_key))
-  end
-
-  def deserialize_cached_value(value)
-    return nil if value.blank?
-
-    JSON.parse(value, symbolize_names: true)
-  rescue JSON::ParserError
-    # If json parse failed, returning the value as is will fail too
-    # since we access the keys as symbols down the line
-    # So it's best to return nil
-    nil
-  end
-
-  def save_to_cache(response)
-    return nil unless event_is_cacheable?
-
-    # Serialize to JSON
-    # This makes parsing easy when response is a hash
-    Redis::Alfred.setex(cache_key, response.to_json)
-  end
-
   def conversation
     @conversation ||= account.conversations.find_by(display_id: event['data']['conversation_display_id'])
   end
 
-  def event_is_cacheable?
-    # self.class::CACHEABLE_EVENTS is way to access CACHEABLE_EVENTS defined in the class hierarchy of the current object.
-    # This ensures that if CACHEABLE_EVENTS is updated elsewhere in it's ancestors, we access the latest value.
-    self.class::CACHEABLE_EVENTS.include?(event_name)
-  end
-
   def api_base
     endpoint = InstallationConfig.find_by(name: 'CAPTAIN_OPEN_AI_ENDPOINT')&.value.presence || 'https://api.openai.com/'
     endpoint = endpoint.chomp('/')
diff --git a/lib/captain/label_suggestion_service.rb b/lib/captain/label_suggestion_service.rb
index d959fdbab..b06514721 100644
--- a/lib/captain/label_suggestion_service.rb
+++ b/lib/captain/label_suggestion_service.rb
@@ -1,10 +1,14 @@
 class Captain::LabelSuggestionService < Captain::BaseEditorService
-  CACHEABLE_EVENTS = %w[label_suggestion].freeze
-
   def label_suggestion_message
+    # Check cache first
+    cached_response = read_from_cache
+    return cached_response if cached_response.present?
+
+    # Build content
     content = labels_with_messages
     return nil if content.blank?
 
+    # Make API call
     response = make_api_call(
       model: GPT_MODEL, # TODO: Use separate model for label suggestion
       messages: [
@@ -14,13 +18,41 @@ class Captain::LabelSuggestionService < Captain::BaseEditorService
     )
     return response if response[:error].present?
 
-    # LLMs are not deterministic - sometimes response includes "Labels:" prefix
-    # TODO: Fix with better prompt
-    { message: response[:message] ? response[:message].gsub(/^(label|labels):/i, '') : '' }
+    # Clean up response
+    result = { message: response[:message] ? response[:message].gsub(/^(label|labels):/i, '') : '' }
+
+    # Cache successful result
+    write_to_cache(result)
+
+    result
   end
 
   private
 
+  def cache_key
+    return nil unless conversation
+
+    format(
+      ::Redis::Alfred::OPENAI_CONVERSATION_KEY,
+      event_name: 'label_suggestion',
+      conversation_id: conversation.id,
+      updated_at: conversation.last_activity_at.to_i
+    )
+  end
+
+  def read_from_cache
+    return nil unless cache_key
+
+    cached = Redis::Alfred.get(cache_key)
+    JSON.parse(cached, symbolize_names: true) if cached.present?
+  rescue JSON::ParserError
+    nil
+  end
+
+  def write_to_cache(response)
+    Redis::Alfred.setex(cache_key, response.to_json) if cache_key
+  end
+
   def labels_with_messages
     return nil unless valid_conversation?(conversation)