diff --git a/RELEASE_NOTES_v2.2.64.md b/RELEASE_NOTES_v2.2.64.md
new file mode 100644
index 0000000..55b8917
--- /dev/null
+++ b/RELEASE_NOTES_v2.2.64.md
@@ -0,0 +1,18 @@
+# Release Notes v2.2.64
+
+Dato: 18. marts 2026
+
+## Fixes
+
+- Forbedret QuickCreate robusthed når AI/LLM er utilgængelig.
+- Tilføjet lokal heuristisk fallback i `CaseAnalysisService`, så brugeren stadig får:
+  - foreslået titel
+  - foreslået prioritet
+  - simple tags
+  - kunde-match forsøg
+- Fjernet afhængighed af at Ollama altid svarer, så QuickCreate ikke længere ender i tom AI-unavailable flow ved midlertidige AI-fejl.
+
+## Berørte filer
+
+- `app/services/case_analysis_service.py`
+- `RELEASE_NOTES_v2.2.64.md`
diff --git a/app/services/case_analysis_service.py b/app/services/case_analysis_service.py
index 4b00ef6..d923ad6 100644
--- a/app/services/case_analysis_service.py
+++ b/app/services/case_analysis_service.py
@@ -67,12 +67,12 @@ class CaseAnalysisService:
                 
                 return analysis
             else:
-                logger.warning("⚠️ Ollama returned no result, using empty analysis")
-                return self._empty_analysis(text)
+                logger.warning("⚠️ Ollama returned no result, using heuristic fallback analysis")
+                return await self._heuristic_fallback_analysis(text)
                 
         except Exception as e:
             logger.error(f"❌ Case analysis failed: {e}", exc_info=True)
-            return self._empty_analysis(text)
+            return await self._heuristic_fallback_analysis(text)
     
     def _build_analysis_prompt(self) -> str:
         """Build Danish system prompt for case analysis"""
@@ -470,6 +470,73 @@ Returner JSON med suggested_title, suggested_description, priority, customer_hin
             confidence=0.0,
             ai_reasoning="AI unavailable - fill fields manually"
         )
+
+    async def _heuristic_fallback_analysis(self, text: str) -> QuickCreateAnalysis:
+        """Local fallback when AI service is unavailable."""
+        cleaned_text = (text or "").strip()
+        if not cleaned_text:
+            return self._empty_analysis(text)
+
+        lowered = cleaned_text.lower()
+
+        # Priority heuristics based on urgency wording.
+        urgent_terms = ["nede", "kritisk", "asap", "omgående", "straks", "akut", "haster"]
+        high_terms = ["hurtigt", "vigtigt", "snarest", "prioriter"]
+        low_terms = ["når i får tid", "ikke hastende", "lavprioriteret"]
+
+        if any(term in lowered for term in urgent_terms):
+            priority = SagPriority.URGENT
+        elif any(term in lowered for term in high_terms):
+            priority = SagPriority.HIGH
+        elif any(term in lowered for term in low_terms):
+            priority = SagPriority.LOW
+        else:
+            priority = SagPriority.NORMAL
+
+        # Basic title heuristic: first non-empty line/sentence, clipped to 80 chars.
+        first_line = cleaned_text.splitlines()[0].strip()
+        first_sentence = re.split(r"[.!?]", first_line)[0].strip()
+        title_source = first_sentence or first_line or cleaned_text
+        title = title_source[:80].strip()
+        if not title:
+            title = "Ny sag"
+
+        # Lightweight keyword tags.
+        keyword_tags = {
+            "printer": "printer",
+            "mail": "mail",
+            "email": "mail",
+            "vpn": "vpn",
+            "net": "netværk",
+            "wifi": "wifi",
+            "server": "server",
+            "laptop": "laptop",
+            "adgang": "adgang",
+            "onboarding": "onboarding",
+        }
+        suggested_tags: List[str] = []
+        for key, tag in keyword_tags.items():
+            if key in lowered and tag not in suggested_tags:
+                suggested_tags.append(tag)
+
+        # Try simple customer matching from long words in text.
+        candidate_hints = []
+        for token in re.findall(r"[A-Za-z0-9ÆØÅæøå._-]{3,}", cleaned_text):
+            if token.lower() in {"ring", "kunde", "sag", "skal", "have", "virker", "ikke"}:
+                continue
+            candidate_hints.append(token)
+        customer_id, customer_name = await self._match_customer(candidate_hints[:8])
+
+        return QuickCreateAnalysis(
+            suggested_title=title,
+            suggested_description=cleaned_text,
+            suggested_priority=priority,
+            suggested_customer_id=customer_id,
+            suggested_customer_name=customer_name,
+            suggested_tags=suggested_tags,
+            confidence=0.35,
+            ai_reasoning="AI service unavailable - using local fallback suggestions"
+        )
     
     def _get_cached_analysis(self, text: str) -> Optional[QuickCreateAnalysis]:
         """Get cached analysis if available and not expired"""