Couple of weeks ago, I blogged here on SCN Not your father’s analytics, search and AI….”What is Watson? for 2000 please” about the promise of Watson. I spoke to every one who would listen – from my friends at work and my clients to my 6 year old daughter, and my two dogs – that Watson will radically change enterprise computing . Dennis Howlett and I spoke about this afterwards, and he posted a great blog on ZDNET http://www.zdnet.com/blog/howlett/ibm-watsons-jeopardy-play/ . Since then, I have had a lot of discussions – again with any one who would lend me an ear, about how SAP ecosystem would benefit if HANA and Watson join hands.
SAP has tonnes of information stored in tables, cubes etc. With HANA, this information is even better organized for faster retrieval. Watson can understand English, the way it is spoken, and zero in on the actual questions – then it can analyze structured and unstructured data using parallel algorithms – and come back with excellent answers. And it can give the confidence it has on the answer.
Here are a few scenarios. I am assuming Watson is changed appropriately – and not in its current form where it only knows how to play Jeopardy.
Let me start with one that I chatted with Dennis
Data in an SAP shop sits on CRM, SCM, ECC , BW , emails etc. Now, a customer service person gets a call about placing a service order. In theory, all the information needed to answer anything that gets asked during this call sits readily available in one of these systems – except, some one should know where to look. If HANA is in the picture – the data can be retrieved really fast. But still – you need to know where to look. Or, some one should make reports for several scenarios and train the customer rep on how to use them. But we already know this is hard to pull off. But what if Watson was also working behind the scenes with HANA? Watson can get the question, and parallely check many different things, and give the rep all she needs to answer the question. Even for an open ended system like jeopardy, Watson could come up with answers in couple of seconds. Enterprise data is much more constrained – and this should be easier to accomplish.
There is a lot going in favor of this scenario – like
1. The rep needs less complex training
2. Watson can learn along the way, and will get better with time
3. All kinds of information can be fed into Watson behind the scenes, so that it is always upto date.
4. Unless the rep has to create or update something, she would never need to touch SAP screens.
If Watson and HANA works together, do we need a lot of pre-built reports and dashboards as we do today? probably not – HANA already makes computing on the fly a breeze, and with Watson working in tandem – we might even be able to bypass manual datamodelling in HANA in some scenarios.
How about fraud detection? Most fraud detection software needs a large number of rules to identify fraud. Since crooks have a full time job trying to by-pass these rules, it is kind of difficult to keep up. Here, I think the combination of Watson’s parallel processing abilities, in-memory technology of HANA and predictive capabilities of software like SPSS will deliver something far superior to what we have today in the SAP world.
Sentiment analysis, text analysis and other techniques in social media side of computing should also make significant leaps with such a combination of two awesome technologies.Most current search software goes against keyword searches to find answers. Obviously, this very rarely leads to useful insight.
How about alerts? with data sitting in RAM, just by HANA alone – we should be able to improve how alerts work. Now combine it with Watson’s abilities, and we could potentially see a fundamental change in how alerts work. We might even reach a scenario where we don’t need a lot of pre-defined alerts – system can learn along the way, and alert the right business user. Same would be the case with workflows – where such a combined system can intelligently route processes to keep running at an optimum efficiency.
There are probably many other applications – the general idea being the whole is much better than the sum of parts.