Scenario Planning for Enterprise Risk Management Shirin Sohrabi and Octavian Udrea and Anton V. Riabov IBM T.J. Watson Research Center 1101 Kitchawan Rd, Yorktown Heights, NY 10598, USA {ssohrab, udrea, riabov}@us.ibm.com Abstract We present Scenario Planning Advisor (SPA), that takes as input the relevant news and social me- dia trends that characterize the current situation, as well as the domain knowledge and generates mul- tiple plans explaining the observations and project- ing future states. The resulting plans are clustered and summarized to generate the scenarios for use in scenario planning for enterprise risk management. 1 Introduction and Motivation Scenario planning is a commonly used method for strategic planning [Schoemaker, 1995]. Scenario planning involves an- alyzing the relationship between forces such as social, techni- cal, economic, environmental, and political trends in order to explain the current situation in addition to providing insights about the future. A major benefit to scenario planning is that it helps businesses or policy-makers learn about the possible alternative futures and anticipate them. While the expected scenarios are interesting for verification purposes, surprising scenarios to the users (e.g., policy-makers) are the most im- portant and significant [Peterson et al., 2003]. Risk management is a set of principles that focus on the outcome for risk-taking [Stulz, 1996]. A variety of meth- ods and standards for risk management under different as- sumptions have been developed [Avanesov, 2009]. We ad- dress scenario planning for risk management, the problem of generating scenarios with a significant focus on identifying the extreme yet possible risks that are not usually considered in daily operations. Our approach is different from previous work in that we reason about emerging risks based on obser- vations from the news and social media trends, and produce scenarios that both describe the current situation and project the future possible effects of these observations. Furthermore, each scenario we produce highlights the potential leading in- dicators, the set of facts that are likely to lead to a scenario, the scenario and emerging risk, the combined set of conse- quences or effects in that scenario, in addition to the business implications, a subset of potential effects of that scenario that the users (e.g., policy-makers, businesses) care about. For example, given an observation of a high inflation rate in a certain country, economic decline followed by a decrease in government spending can be the consequences or the ef- fects in a scenario, while decreased client investment in the company offerings is an example of a business implication. Furthermore, an increase in the price of a commodity can be an example of a leading indicator for such a scenario. The main idea of the approach in SPA is to view the sce- nario planning problem for enterprise risk management as a problem that can be translated to an AI planning problem [Sohrabi, Riabov, and Udrea, 2017a]. An intermediate step is a plan recognition problem, where the set of given business implications forms the set of possible goals, and the obser- vations are selected from the news and social media trends. The domain knowledge is acquired from the domain expert via a graphical tool and is then automatically translated to an AI planing domain. AI planning is in turn used to address the plan recognition problem [Ram´ ırez and Geffner, 2009; Sohrabi, Riabov, and Udrea, 2016; 2017b]. Top-k planning is used to generate multiple plans that can be grouped into a scenario [Sohrabi et al., 2016]. The set of plans is then clus- tered and summarized to generate the scenarios. 2 Key Ideas Implemented in SPA 2.1 Planning Formulation The scenario planning for enterprise risk management prob- lem is defined akin to a plan recognition problem, where the set of observations is selected from news and social media, the set of possible goals is a set of business implications, and the domain knowledge is captured by the domain experts. We define the solution to the problem as a set of scenarios, where each scenario is a collection of executable plans that explain the observations and considers the possible cascading effects of the actions to identify potential future outcomes. 2.2 Data Transformation SPA continuously monitors multiple real-world sources (e.g., news channels, social media posts) to identify the set of ob- servations. To this end, several text analytics are implemented to find the information relevant for a particular domain in the vast amount of information available to crawl. We define a set of predefined relevant news sources, topics, and organizations in order to refine and filter the information. Users/analysts can also add a set of keywords that is important for a particular domain. Analysts then review the generated results and select the observations that are the most relevant and important for them. Note that SPA can deal with unreliable observations (i.e., noisy and missing observation) as it exploits previous work on plan recognition as planning that addresses unreli- able observations [Sohrabi, Riabov, and Udrea, 2016].