diff --git a/debezium-embedded/src/main/java/io/debezium/embedded/async/AsyncEmbeddedEngine.java b/debezium-embedded/src/main/java/io/debezium/embedded/async/AsyncEmbeddedEngine.java
index da91c7517..f9b7a6af0 100644
--- a/debezium-embedded/src/main/java/io/debezium/embedded/async/AsyncEmbeddedEngine.java
+++ b/debezium-embedded/src/main/java/io/debezium/embedded/async/AsyncEmbeddedEngine.java
@@ -148,7 +148,13 @@ private AsyncEmbeddedEngine(Properties config,
 
         // Create thread pools for executing tasks and record pipelines.
         taskService = Executors.newFixedThreadPool(this.config.getInteger(ConnectorConfig.TASKS_MAX_CONFIG, () -> 1));
-        recordService = Executors.newFixedThreadPool(computeRecordThreads(this.config.getString(AsyncEmbeddedEngine.RECORD_PROCESSING_THREADS)));
+        final String processingThreads = this.config.getString(AsyncEmbeddedEngine.RECORD_PROCESSING_THREADS);
+        if (processingThreads == null || processingThreads.isBlank()) {
+            recordService = Executors.newCachedThreadPool();
+        }
+        else {
+            recordService = Executors.newFixedThreadPool(computeRecordThreads(processingThreads));
+        }
 
         // Validate provided config and prepare Kafka worker config needed for Kafka stuff, like e.g. OffsetStore.
         if (!this.config.validateAndRecord(AsyncEngineConfig.CONNECTOR_FIELDS, LOGGER::error)) {
diff --git a/debezium-embedded/src/main/java/io/debezium/embedded/async/AsyncEngineConfig.java b/debezium-embedded/src/main/java/io/debezium/embedded/async/AsyncEngineConfig.java
index 875dc13a8..24cf289a7 100644
--- a/debezium-embedded/src/main/java/io/debezium/embedded/async/AsyncEngineConfig.java
+++ b/debezium-embedded/src/main/java/io/debezium/embedded/async/AsyncEngineConfig.java
@@ -24,7 +24,9 @@ public interface AsyncEngineConfig extends EmbeddedEngineConfig {
      */
     Field RECORD_PROCESSING_THREADS = Field.create("record.processing.threads")
             .withDescription("The number of threads to be used for processing CDC records. If you want to use all available threads, you can use "
-                    + "'AVAILABLE_CORES' placeholder.");
+                    + "'AVAILABLE_CORES' placeholder. If the number of threads is not specified, the threads will be created as needed, using "
+                    + "Java 'Executors.newCachedThreadPool()' executor service.")
+            .withDefault(""); // We need to set some non-null value to avoid Kafka config validation failures.
 
     /**
      * An optional field that specifies maximum time in ms to wait for submitted records to finish processing when the task shut down is called.
diff --git a/documentation/modules/ROOT/pages/development/engine.adoc b/documentation/modules/ROOT/pages/development/engine.adoc
index 55bd7965d..97ff1d88d 100644
--- a/documentation/modules/ROOT/pages/development/engine.adoc
+++ b/documentation/modules/ROOT/pages/development/engine.adoc
@@ -528,9 +528,9 @@ The default is a periodic commity policy based upon time intervals.
 
 |`record.processing.threads`
 |The number of available cores.
-|The number of threads to use for processing CDC records. 
-The default value is the number of available machine cores, with an upper limit of 16 threads.
-If you want to use all available threads without any limitation, use the `AVAILABLE_CORES` placeholder.
+|The number of threads to use for processing CDC records.
+If you want to use all available cores on given machine, you can use the `AVAILABLE_CORES` placeholder.
+If the value is not specified, threads are created as needed, using Java https://docs.oracle.com/en/java/javase/17/docs/api/java.base/java/util/concurrent/Executors.html#newCachedThreadPool()[cached thread pool].
 
 |`record.processing.shutdown.timeout.ms`
 |1000