Archive for March 26th, 2008

In praise of swivel chairs

Wednesday, March 26th, 2008

Sean McGrath wrote an interesting piece at IT World suggesting that use of swivel chair interfacing is often the best integration strategy available to an organisation.

I am not sure how many swivel chairs it would take before Sean would see an automated approach as more suitable.  I’ve certainly seen banks of people doing little more than this type of business process, inconsistently and with plenty of errors – in itself a very expensive way of doing things.  I do, however, take his point and this is not a criticism of what is essentially an excellent post that raises a number of important points.

For me, the most important point Sean raises is that “if you cannot build a new end-to-end business process that covers both systems manually then you do not understand the requirements sufficiently to start coding it.”

These thoughts are also picked up by Reg Braithwaite in an equally excellent post Is software the documentation of business process mistakes?  Reg argues that if your code represents the “user manual” of the business process, then if the code is too complex, this may well be because the process is too complex.  In other words, get the manual process right before you automate it.

The only reason swivel chair integration exists is because of the inflexibility afforded to the business user by the existing systems.  Business users are left with manual as the only option available to them.  It is the quickest, cheapest and most reliable way of getting a business process done – which is a sad indictment on the state of affairs in IT.

My interest in this subject is that innovative solutions like Blue Prism are trying to address these issues, bringing agility to the business, but without damaging the integrity of the underlying systems, or creating the need for new code.

There will always be some place for swivel chair integration but let’s keep it to a minimum.  Once a process can be done manually, the only things that should prevent it from being automated are the need for human interaction, or the need for human intelligence (e.g. expert judgement).