Bruce Silver always makes me break the rules, and tonight I’m breaking the “everything is off the record after the bar opens” rule since he scheduled sessions after dinner and with an open bar in the back of the room. Rules, as they say, are made to be broken.
Roger King of TIBCO attempted to start this demo during the earlier session but there were problems with the fancy projector setup. He’s back now to talk about model-driven work management. TIBCO’s core customer base (like mine) is traditional enterprises such as financial services, and they’re seeing a lot of them retiring legacy enterprise apps now in favor of process-centric apps built on platforms such as TIBCO. They see specific problems with work management in very large, branch-network organizations like retail banks; by work management and resource management, they mean the way that work is distributed to and accessed by end users, one of the things that BPMN doesn’t do when you define processes. With tens of thousands of participants, just a small increment in productivity through better work management can cause a significant ROI in absolute terms, but traditionally this has been done through custom user interfaces and distribution/matching. There are a number of resource patterns that have been studied and developed, e.g., separation of duties, round robin; Roger demonstrated how these are being incorporated into TIBCO’s AMX BPM (modeled within their Business Studio product) through organizational models, where you can find the resources in the organization, groups and custom organizational units that you need to bring your business vocabulary to determining how work is distributed within your organization. The idea is that once you have this defined, you can then use very fine-grained rules for determining which person gets which piece of work, or who has access to what. This now becomes something that you can attach to an activity in a process model using simple assignments or with a resource query language that assigns it dynamically, including based on process instance variables – essential when you have 100’s or 1000’s of branches and can’t realistically administer your organizational model and work distribution methods manually. Furthermore, you need to be looking at having people go to the work rather than having work sent to the people. This is the only type of work distribution approach when you’re creating declarative processes, where configuration needs to be much more dynamic than what might be drawn in the process model.
We finished off the short opening day of bpmNEXT with a keynote by Jim Sinur, late of Gartner (but not hesitant to use the materials that he helped to create there) and now an independent analyst, on how his processes are smarter than him. Processes based on machine learning, however, can only go so far: although machines are more accurate and consistent (and never complain when you ask them to work overtime), people are better at unexpected situations. The key is to have computers and people work together within intelligent processes: let the computers work on the parts that they do best, including events, analytics standardized decisions, pre-defined processes and the resulting actions from combining all of these; exploit emerging technologies such as cognitive systems, what-if scenarios via simulation, intelligent business operations, visualization and social analytics. Intelligent agents are a big part of this, but we need to have goal-directed processes to really make this work, or abandon the concept of processes at all except for the footprints that they leave behind.
Rule-breaking done. Back tomorrow for a full day of bpmNEXT 2014.