70 likes | 83 Views
AAMAS 2004 – Panel on Business Process Management. Tom Wagner, Ph.D. DARPA / IPTO twagner@darpa.mil. BPM. Disclaimer – not to be confused with fact… Broad space / BPM may involve a bundle of ideas: Human task/workflow management.
E N D
AAMAS 2004 – Panel on Business Process Management Tom Wagner, Ph.D. DARPA / IPTO twagner@darpa.mil
BPM • Disclaimer – not to be confused with fact… • Broad space / BPM may involve a bundle of ideas: • Human task/workflow management. • Information flows, e.g., auto-magic information exchange to support human work. • Union of that space -- human work management + information management. • BPM = business process awareness + management. • Progression of focus for practitioners: transaction processing (data in), then business intelligence (information from your data), to process awareness and management. • Many different slants driven by different areas of commercial interest: Process-centric workflow. Optimization, what-ifs, dynamic task management. Information management. Enterprise applications. Software infrastructure. Others. Not drawn to scale. Axis unknown.
BPM and Agents • Sidebar: standards • BPMI – business process management initiative. www.bpmi.org • Nonprofit company w/goal to setup standards. • E.g., BPML (meta biz process modeling language) or BPMN (graphical representations of BPML). • Others: BPEL (“bee pell”) web services + biz proc (MS, IBM, BEA, SAP). UML (OMG) = component view. • Why agents? Places to hook arguments for: • Distribution – large scale, privacy concerns between different companies, etc. • Lose coupling – integrating heterogeneous systems and people in dynamic setting. • Autonomy – efficiency value proposition (automation makes things faster). • Choice – important for dealing with humans (also mixed initiatives). • Process may also have choices that need to be made locally. • Intelligence – pushing autonomy from simple data triggered action to complex task performance or complex analysis. • Adaptive, dynamic, flexible, etc. • Stumbling block for (some) agents: explicit representations of processes. • Commercial folks have the same problem ( template libraries). • Implies look to domains with documented processes. • But often existing models will lack features, e.g., choice nodes, parallelism. • May not be an issue -- depends on your BPM vision and how near term you are operating.
Make Your Own Custom BPM Vision As a researcher… • How you approach BPM is dependent on your needs and goals. • Select a subset (of that space) and clearly define the subset. • Consider being complimentary to (but not competing with) commercial interests. • Shoot high / offer a new capability that dovetails with where they are going. • Hard to compete with development or near term work. • If you need external investment / are working for “real world” impact: • Parallel application spaces, e.g., military. • Or know your customer. Voice-of-customer should modulate: • Your argument for an agent approach (why distributed, etc.). • Your value proposition. • Potentially your long term technical vision. • You decide the balance of customer voice versus research vision. • Most value propositions will include efficiency improvements (faster, cheaper). Research: learning to adjust workflow models. Commercial: Process-centric workflow. Research: distributed dynamic human activity coordination in RT environment.
Why I like BPM as an Application Domain(Unsure about being “the” killer ap) • Amenable to “agent” solutions. • Two ways to motivate investment: • A product for sale. • Internal use / not for sale. • All companies, agencies, universities, etc., can improve efficiency of processes. • Potentially cheaper / smaller barriers. • Some version of the problem space is probably accessible to you. • The problem exists today: • You will be proposing new solution to a known (and familiar) problem. • Les Gasser’s anecdote • i.e., not developing both a new product and a new market.
Process-centric WorkflowAgents vs More Conventional Approaches • Gross categorizations – probably ranges of things. • *If* conventional workflow is often: • Centralized. • Complete / global information. • Top-down – tasks are put at human effectors. • Monolithic / fully integrated. • Tending toward static. • An agent approach might be: • Distributed (implies scalable, implies ability to span enterprises). • Able to operate with partial information. • Blend top-down (direction from management) with bottom-up (performers inputting preferences, choices, and tasks themselves). • Loosely coupled (easier to integrate heterogeneous systems). • Tending toward dynamic / adaptive. • For the predictability question – if you can get a customer to ask that question, they are already interested in your vision. (You have to generate your own actual response to that question.)
Funding Your Custom BPM Vision Incremental Investment / Payoff Hypothetical vision: intelligent systems, stupid systems, and humans that work together both within an enterprise and across enterprises. Problem: research is expensive, payoffs are uncertain, takes a long time to make anything happen. Research Cost Incremental investment. Hypothetical waypoint: centralized management of tasks of two humans who always work on related process / have activities that interact. Catch-22: You still have to shoot high or they don’t need you. n .5 Project Years