Message brokering across different systems, with transactionality and the ability to have fine tuned control over what happens using Java (or other languages), instead of a heavy, proprietary languages. One situation that it doesn't fit very well (as far as I have experienced) is when your workflow requires significant data mapping. While possible when using Java tooling, some other visual data mapping tools in other integration frameworks are easier to work with.
Drools is well suited for big projects where business logic and rules must be separated from program code. So they can evolve when business evolves without being tied to code evolution and deployment.
Some of the documentation is a little sparse. In particular, its TCP-based routes use an underlying Netty server, and the interactions between Netty's decoder capabilities and Apache Camel's routing/handler capabilities can be a little muddy at times. In general it is clear which routes and endpoints are the more frequently used and which haven't been given as much attention.
Fusion doesn't support persistence of working memory, which brings some extra high availability risk to our business.
Guvnor still has a lot room to be implemented, it is not so user-friendly for non-technical people, so a lot of business users complain it is hard to master.
Rule execution server doesn't even have JMX implemented, hard to be monitored.
Drools is still lacking support for key Web services standards.
Apache Camel has been the integration framework of choice, but I was not the person to make the decision to use it. Compared to other competing products like Tibco Business Works, etc., it is free and open source and its licensing policy is acceptable to the management of Cox.
OpenRules provides the non-technical Excel way for a business user to easily modify and manage the rules. Sometimes we found Drools is a little bit overkill for some small and quick projects and we found Roolie is a not bad option as Drools alternative.
Very fast time to market in that so many components are available to use immediately.
Error handling mechanisms and patterns of practice are robust and easy to use which in turn has made our application more robust from the start, so fewer bugs.
However, testing and debugging routes is more challenging than working is standard Java so that takes more time (less time than writing the components from scratch).
Most people don't know Camel coming in and many junior developers find it overwhelming and are not enthusiastic to learn it. So finding people that want to develop/maintain it is a challenge.
The IT department quickly adopted Drools as it is a very good java-based rule engine, which saves a lot of time to meet the project timeline and balanced our business requirements.
Recently we start considering the OpenRules, which may be more business user-friendly.
