The Great Debate: Business Intelligence - Opening Speeches

Hi JC,

I appreciate your interest in the ideas I present around perceptive processing. I think your comment lays out a good way forward. My only caveat is that I don't want to build another IT edifice on top of BI - and that is what I fear would end up happening if we added AI to BI. We need to find ways to add external data to our transaction systems, and for that metatables that look up that info could well do the job. However, we also need to find ways to capture ad hoc information (for example, our own gut instinct) for which lookup tables will not work at all.

The problem is not with the software but rather with expectations. At its most fundamental the software is just an enabling technology to bring data together in a meaningful format so that decisions can be made. The issue is that too many still think that implementing software will prove to be a panacea.

Too often people are so enthused by the capabilities that they just end up producing a vast acreage of reports that are 'interesting' but not actually useful.

The key is to get back to first principles and say a) what are we trying to achieve b) what do we need to measure that will tell us how well we are performing and then c) start using the technology.

Leaving aside information required for regulatory reasons, if you can't make a decision from a report, why bother with it? Challenge yourself by asking these tough questions and you've more chance of generating ROI from your investments.

This is an interesting debate although one would have to take issue with laying the blame entirely at the software developers door.

To date it has proved to be almost impossible to define or obtain the requirements even for basic 'learned processes' - i.e. universal standard reports. With accounting (ERP) packages the Profession would be ideally suited to define standard reporting requirement because they are best positioned to determine what is required. Unfortunately they have never engaged and instead of uniform reports each software vendor has re-invented the wheel rather than just drawing upon approved standard layouts.

Jyoti

Surely you are talking about introducing a form of AI (artificial intelligence). Including the ability to learn by capturing and driving the system using external 'user defined' information held in associated rules metadata tables. Thus providing the means of achieving a 'perceptive process' (as per your soft drinks vendor example:
http://www.kiteblue.net/jyoti/2008/02/perceptive-proc.html

Therefore the system would learn by building a picture of other factors which are not directly connected although they can have a material influence on the end picture

Coupling the above with 'learned processes' would make a reasonably cognitive system capable of providing a fairly extensible BI interface. Still not true AI but a passable approach providing a better competitive edge than current BI systems and what's more it could be a generic model not limited simply to accounting systems (i.e. uses for Forex, Trading etc)

While I have sympathy with both points of view presented here, I am increasingly of the view that extant business intelligence systems are built on the wrong foundation. As a result, they are not intelligent, and certainly fail to provide their users with intelligence, real or contrived. Let me explain myself.

I like to distinguish between two different types of processes in an organisation: learned processes and perceptive processes. Learned processes are about what the company does and all its day-to-day activities are based on these. Our ERP systems do their best to help a company look very skilled in dispensing learned processes.

But learned processes are usually very transactional in nature. They don't give room for new ideas, new concepts, new advances in understanding. For these, we need what thinker Don Norman calls the "effort of reflection." But we need to go beyond reflection - my own name for this is perceptive processing. For perceptive processes to work, we need to take into account an organisation's recent past (our BI systems usually work here), but also our estimates of the future, and our own insights and perceptions of the environment in which we are operating in. BI systems fail in their droves when it comes to perceptive processes precisely because they are singularly one-dimensional in looking simply at transactions.

I am not sure we can make our managers look more intelligent, but I will certainly welcome them appearing more perceptive.

By the way, I have written a paper on perceptive processes, mostly for my own benefit, but I will be happy to share this with anybody who is interested in taking the discussion further.