Complex Adaptive Business Process
On Nov. 3rd, I will be attending a Thought Leader Summit of the Workflow Management Coalition. The focus of discussion will be Case Management or Dynamic BPM. Keith Swenson, who felt I could add to the discussion, was so friendly to invite me. Given my critical stance towards BPM, I feel like walking into the Colosseum with lots of hungry lions waiting. Should this be my last post, you know what happened … just a little joke! Keith asked participants to offer a position statement upfront to enable others to prepare themselves.
Here is my Position Statement for WfMC Thought Leader Summit:
Businesses have to succeed in an economy that is a ‘complex adaptive system’ rather than a complicated system that can be decomposed into building blocks. Adam Smith’s invisible hand at work. A system of independently acting agents is not complicated but complex and is considered a complex adaptive environment that resists decomposition, see Anderson, Arrow, Pines (1988). From a planning perspective it seems ideal if businesses would not consist of independent acting agents, but would act coherently, hence BPM to the rescue. I propose that such perfect businesses would be so foreign in a dynamic free market economy that they would fail. Thus also large companies are complex adaptive systems, in which (like in most natural systems) successful structures emerge, adapt and evolve rather than being artificially created (think about in-house politics for a moment!). In such systems, functions cannot be created by a methodology of analysis and design, mostly because for the lack of realistic abstraction. That does not mean that there is no process perspective but it is much less detailed than a step-by-step flowchart model. Models are only necessary to define the real-world entities for process handoff between process owners.
Before jumping into the technology aspect of complex adaptive process I need to discuss the business element of my approach. I agree that there are businesses with different management styles regarding customers service and HR. I propose that even businesses with a more ’employee control’ approach would possibly benefit more from less process analysis and more frequent process adaptation through process owners, because of the huge amount of control structure.
I arrived at my above perspective not from an expansion of BPM to dynamic process management or from the needs of case file collaboration, but from – non manufacturing – business and human resource management for service processes. I am opposed to the idea that a business can be improved by analyzing and designing rigid processes, except for an irrelevant small percentage (as an addition to ERP). I fail to find the proof that flowchart-designed processes and the related change management bureaucracy do make a business more agile. People are (if at all) agile but never analyzed and designed processes. Frequent process changes will simply destroy whatever initiative employees might have left. I have yet to see a business that does actually automate more than 20% of its processes and currently assume that this is the maximum that can be achieved. Rather than the 80% often proposed. The more processes are controlled, the more bypass and add-on activities exist. I fail to see the sense in creating rigid processes and then define everything that does not map into it as an exception. The process spectrum is rather from straight through processing to social networking with case management in between.
Process management in manufacturing was always mostly focused on quality to reduce cost, while it seems that for business services it is reducing manpower, needed skills and thus cost. I suggest that process management must always focus on customer service quality first. But not process makes people happy, but people make people happy. Therefore we need to empower people! I disagree that it is only the knowledge workers who need more dynamics and back office clerks who need rigid BPM. Also a simple activity within a case can be strictly guided by a complex adaptive process, so why bother to waste time with expensive and inaccurate process analysis?
The most likely approach to succeed is to empower the process owner to create processes on the fly as needed and empower the users to adapt those on the job. Here the similarity to Human Focused Process, Case Management and Dynamic BPM approaches becomes apparent. A complex adaptive process is however much closer to the dynamics of social networking rather than to case file collaboration.
Another important element is ‘goal definition’. Goals are a set of RULE correlated parameters that can be verified at certain intervals or be triggered by thresholds. In difference to KPI or key performance indicators, goals can be verified during real-time processing. KPI’s tend to be accumulated, consolidated data warehouse numbers. A goal mismatch event should trigger some corrective process. Goals can be anything from simple SLAs to revenue targets.
I do not want to belabor the current technology issues with roundtrip from BPM model to execution. While it may be true that most BPM systems can offer some case management aspects, the analysis, model, implement, simulate, deploy, monitor and improve cycle remains as a huge obstacle to the promised agility.
Key functions of complex adaptive (CA) process systems:
- Distributed (similar to grid computing) communication enablement of process systems
- The exchange of process definitions is less relevant than meta-data and methods
- Consolidated central repository (not archive) for change management
- Create a business architecture model of data entities in repository.
- Utilize an object model with state/event modeling (implementing i.e. pi-calculus)
- Enable a 100% change roundtrip for ALL elements of a process
- Empower users to write boundary rules in natural language accessing the object data
- Strong distributed security with embedded access control on method and attribute level
- Enable authenticated users to create virtual organizations of collaborators
- Empower business users to access and use defined data entities from repository
- Simple mapping of backend business systems to data entities. (SOA or not)
- Simple linkage to external event creating systems
- Empower business users to create their own data entities.
- Enable business users to create all the necessary content linked to business objects
- Seamless integration with all inbound and outbound content (backend archives)
- Enable business users to define role-specific user interfaces as needed.
- Use timeline graphs to show past activities for documentation
- Use organization charts to show authorized roles and departments or virtual organizations
- User authorization (role/policy) and workgroup/queue assignment are independent
- Real-time business data are mapped into the process to measure business results.
- Customer focused processes are setup to survey customer satisfaction.
- Consolidating process, business and satisfaction data for the process owner
- Business results are accumulated and dashboards and reports create transparency
- Enable offline and mobile participation for all CA processes
- … I certainly missed a few things!
- Business trained document classification and routing
- Business trained content capture and data extraction
- Machine learning agents discover user activity patterns in state space
Let me note that it is not simple to ‘utilize neural networks’ on top of an existing BPM engine, because it cannot expose the state space for time-stamped user activities. It is also difficult to empower business users to define GUIs and write rules without a deeply embedded security layer as otherwise they can define themselves unauthorized data or content access.
I look forward to the summit to meet so many interesting people and will report on it right here.