Case Study: IBM Watson Analytics

III) Suggestions for resolving above issues

1) Minimizing the cost of using Watson

IBM Watson Analytics is open to company use for free, comprising cognitive analytics and some fundamental features. The free-of-charge service is however only eligible for one user.

The best option for a bank of DBS’s scale would to subscribe to the $80 per month per user professional package.

There are however other possibilities for DBS to reduce its cost in using Watson. To avoid all unnecessary costs, the bank should pay only for the services that it really needs.

It can firstly eliminate unnecessarily added-on features provided by IBM. DBS has invested US$11.8 million in Watson over three years. However, DBS can subscribe back the service as per requested.

These include the complexity of interpreting ambiguity and context, the supervision of the learning process and providing recognizable suggestions. In order to improve Watson’s liability, DBS can turn to specialized groups of analysts to cope with different problems existing in the data retrieval and interpretation process, in terms of wealth management.

They would firstly need to strengthen the research in social computing, and design computational systems that accommodate specific social practices to help provide better insight into wealth management.

Secondly, testing the accuracy of prediction figures, detecting errors and supervising the learning process are required in order to reduce the chance of system errors as much as possible.

Moreover, we suggest that DBS offer training programs for customer relationship managers, in order for them to understand the implications of the suggestions made by Watson, and to help them make the best decisions with the assistance of the

4) Avoiding system failures

In order to minimize the damages caused by potential system failures, DBS should collaborate closely with IBM scientists to maintain Watson’s system.

Firstly, it is important to avoid any disruption of the service by adding redundancy to primary storage systems, such as duplicate storage systems or identical servers. However, one drawback is that this will reduce the storage utilization.

Additionally, DBS should elaborate backup system ‘snapshots’, that can provide checkpoints to restore to should a disaster occur. Relying solely on one restoration protocol could be imprudent. Therefore, DBS should consider multiple layers of restoration pathways to avoid losing essential and confidential data, and to ensure data