Economies of Data Integration

What are the real issues or 'real time' issues of data integration? While we see everyday ads or
claims from companies about how their software can integrate data, there are still a lot of nitty
gritties that have to carefully considered while making a 'one strategy for all' decision. One of
my favourite quotes is this: 100% automation leads to ineffectiveness.
All organizations, from the smallest of SMBs to the largest enterprise have silos of data, silos
of information across finance, IT Operations, various customer or client data repositories,
internal repositories, internal workflow systems etc. Unless an enterprise is ideally designed,
most of them would have parallel streams of workflow. One stream would be the integration between
internal operations, finance, costing, auditing, compliance divisions. These are largely
Microsoft Excel based, Tally based or based on similar tools. How many of the currently available
Accounting softwares, Network monitoring softwares or Compliance checkers are interoperable with
each other and with enterprise data flow management? The internal Ops data is also highly
sensitive and protected and the typical finance manager does not know or does not believe in SSO
kind of things and he only knows authentication established locally. In 80% of the companies
according to a research site on macro economics, the most sensitive financial data is still
resting on the hard drive of the chief finance person's laptop which is ofcourse well backed up.
The second stream that can be integrated is the inhouse knowledge and information management,
including competence building, training, market gap analysis well tuned with product analysis,
gap based skills acquisition. This stream data is not as hard to integrate as the first stream
but nevertheless difficult. If each division such as Corporate Training, Knowledge Management, IT
Management, Program Management etc have different products controlling them (which is usually the
case) as varied as Microsoft spreadsheets, Siebel based systems, SAP based systems, Tivoli based
data - well, I can hear SOAists screaming, but we still have a long way to go on that.
The least difficult of the data integration streams is the enterprise level data and information
integration. We all know that all major enterprise players have developed various adapters and
stand alone products which will recognize and discover services, understand business rules,
transform business data according to that and 'talk to each other'.

I'm trying to gather information about how well information integration itself is working out. Readers, please pass on any web sites that publish information about the economics of
service-to-service adoption and how do their returns measure up against cost in clear
quantifiable terms.