A Service Oriented Architectural initiative that concentrates on taking all the existing systems and wrapping their API in web services is missing the point.
I am a strong proponent of Service Oriented Architecture and in fact I am the Lead Architect for Smart Communications SOA COE, so the following is not an indictment of SOA.
A non Service Oriented Architecture initiative that concentrates on Re-Use of existing resources is preferable to a so called SOA that doesn't concentrate on re-use.
For example, let's say you have a back end application that has an API that is used by several other applications. If you have code in all the other applications that accesses that API then you have multiple point to point integrations. One approach that is not SOA but encourages re-use, would be to create a "helper" class that simplifies the various API accesses and provides one class that has multiple operations in a single or small number of interfaces. Then adding a consumer of that API is just a matter of interacting with the "helper" class and its simplified interfaces.
Now if you do that first, go through all the systems you have and look at how you can re-use them by simplifying the access and if you also look to standards based interfaces, THEN you can wrap that "helper" class (aka adapter) in a web services or message oriented middleware (such as MQ OR Message Broker) THEN you will have a SOA basis and your odds of success go up greatly.