Steve Hendrick of IDC kicked off day 2 with his keynote on Intelligent Process Automation: The Key to Business Process Automation.
Steve started with a story about AMR and SABRE. Back in 80s startup low-cost airlines were putting real pressure on them and they could not meet the price. Thomas Cook came and introduced yield management to vary prices of seats. Classic use of Operational Research – predictive analytics for demand and optimization for yield management. More or less saved American Airlines. Steve regards this as a very early example of Intelligent Process Automation or IPA.
He defines IPA as the integration of business analytics with business process management to increase the decision making capabilities within existing and new types of applications and to improve the ability of stakeholders to make faster, more informed decisions.
Four key constructs for IPA:
- Event Processing
Infrastructure and messaging for sense and respond - Decisioning
Capturing rules and identifying actions. Consumes and generates events - Business Process Management
Stateful deployment, workflow and integration of processes - Advanced Analytics
Data mining and model building software
I prefer to group analytics into the decisioning piece as I think the value of decisioning using business rules and analytics is much greater.
Steve’s view of advanced analytics matches mine and the one we use in the book. Analytics range from profiling and exploratory analysis, through predictive analytics to optimization and simulation (see slide 22 in the presentation Neil and I gave yesterday – Steve’s slide was pretty similar). Steve gave a quick intro to optimization and how linear and non-linear programming work (LP and NLP). He made the point that these models are powerful but expensive to develop and can age and degrade over time, requiring a degree of constant maintenance.
Steve views a decisioning component as important within the range of BPM functions and sees the combination of decisioning and BPM as what matters. Not sure I completely agree. He did say, and I agree, that advanced analytics can easily be integrated into BPM through decisioning. Check out this article, for instance, about bringing business intelligence into processes.
Steve’s view of events is that they are becoming more and more relevant as companies move to more real-time systems with ultra-low latency. Like Steve I think events and Event Driven Architecture/Complex Event Processing are valuable adjuncts to decisioning and processes when very fast response is required. Event processing is not an end in itself but a great way to bring the other pieces into a real-time environment.
He presented a slide with a picture of how the pieces fit together (link to previous post) and I have a slightly different view of how this would look. His colleague Maureen Fleming also presented on this topic and I blogged about that here.
His examples of successful IPA included credit card fraud (neural networks, dynamic profiles – like those discussed by Charles Nicholls of SeeWhy yesterday – and business rules), campaign management for precision marketing (such as this Fair Isaac case study), operational efficiency at GM across 30 plants in 10 countries saving billions. By and large these examples come down to profit maximization or cost minimization along with increased precision. He also gave a couple of examples of poor management that could have been solved with better IPA (such as JetBlue’s fiasco) and opportunities such as Netflix’s prize for improvement in their recommendation algorithm.
IPA’s challenges include data availability, the complexity of mathematics, the lack of process context for analytics and the fact that most event processing relies on patterns yet there is little or no help for which patterns to look for. That said, lots of the building blocks are available. In addition the combination of business rules management systems with predictive analytics can handle the more complex decisioning. Organizations are more and more reliant on their technology, making IT departments keen to deliver this support to their business users. Finally advanced analytics can deliver some tremendous value. The component markets are growing fast, especially BPMS which IDC sees as growing at 44%!
IPA Prerequisites:
- Great data
- Great model builders
- Great Tools (BRMS, Analytics)
- Experiencing in Decisioning
IPA remains integration challenged in Steve’s view but is, he thinks, the “final word” in complex decisioning and helps you achieve the most with what you have.
Steve ended by pointing out that Competing on Analytics and Super Crunchers are good background books on analytics (which they are) though he failed to mention Smart (Enough) Systems!
By the way, Steve mentioned PMML as a valid way to bring models and model artifacts into the business rules execution space. You can find more on PMML here.
Finally my pal Sandy K blogged about this session on her blog here in case you want a second opinion.
Comments on this entry are closed.