I attended a demo and workshop on IBM Watson showing off what it can do for the Oil and Gas industry. IBM has assembled a strong team of Oil and Gas professionals to really deep dive into that domain. From what I saw, they are ready to take on the Oil and Gas industry and add real value.

First, IBM Watson has the ability to integrate data from disparate data sources. There's nothing new here as this has been done for many years. The key differentiator with Watson is that it is able to relate data from data sources with free text data so that it can look at the data in context. It can use this historical and contextual data to help make predictions of what will happen.

For example, if you have as a set of morning reports which includes structured and unstructured data about depths, equipment, performance, and free-text reports, Watson will look at the morning report and read and interpret the natural language as the human brain does and use that as context about the recorded events that happened during that time period. Say there was reaming required in a certain type of well with certain characteristics, Watson could go back and find similar wells with similar characteristics where reaming was required.

Next, you can ask Watson questions. Once it has the data from morning reports and has ingested the data into it's cognitive engine, you can ask it natural language queries. For example, "What are similar wells that required reaming in the past with similar lithology?"?? Watson could give you a set of wells that are similar and can even rank the similarity as a percentage.

One of the great things about the demo was the quality of user experience. Knowing that this was a demo of current and future technology, it was quite impressive. The user experience was such that all relevant data was mapped out and related and navigation was customized by persona/role. The use of a radial bar chart to show well similarities was quite interesting and even showed how the chart could be used to show differences by using overlays of bars.

The visualizations of the data shown in this example were to compare analogous wells?? based on key physical characteristics of the wells and basins where they are located. The example also used unstructured data that could be displayed to show well history and relevant information for the similar wells. In this case, they were looking at how to drill a well in a new area of the world by comparing it to wells drilled in similar geological areas in other parts of the world.

The visualization was interactive. The user could tell Watson what to focus on and what to ignore and it would learn from this. This would allow the results to be further refined. Additionally, Watson remembers these choices for the next time the user tries to do a similar exercise. So if they were looking for analogous wells in other areas of the world with different geological properties, Watson would use the choices used in the last exercise.

The visualizations were not limited to the radial bar chart. A map was used and all well sites could be color coded based on specific characteristics. You would see the wells highlighted by color based on the characteristics chosen. The entire demo took 20 minutes and showed that the visualizations available are extremely powerful and allow for querying of data rapidly to get to where you can use Watson's cognitive abilities to further refine the results.

The use of KPIs and heat maps showed great user experience for operations. This matches the work of PAS and their high performance HMI work. You want to only highlight information that is the most critical to operations, especially safety and environmental, so that these high-cost events can be properly mitigated. The KPI interfaces were simple and uncluttered and brought the user straight to the problem that needed to be resolved. Here, once a failure was noted, Watson was able to correlate information about causes from not just internal data sources but also news. For example, a pond being overfull after a severe thunderstorm.

There was another example around project planning and predicting cost overruns. Because Watson can take evidence from multiple sources of data, it can be used to predict time and cost overruns using not just structured data but also reports about related activity across the enterprise.

So as opposed to statistical techniques and using models, using Watson allows a process similar to human thought to connect data from across the enterprise and across all public data sources and put it together into a full picture. It learns as it goes and gains domain knowledge. This can reduce the amount of training required of employees and can even be used to drive best practices across an organization by having Watson learn these best practices where employees can query what to do next. It can be used in conjunction with dashboards to drive users to appropriate information and actions to help make better decisions. The practical implications are impressive and, if used properly, could see huge cost savings by driving better decisions. At economies of scale, it can lead to massive savings for a company. I predict that it will disrupt many of the existing technologies used in Oil and Gas today for information display, dashboards, and KPIs by giving the addition of advisory functionality and historical knowledge to organizations.