Jerome Boyer of IBM presented a methodology – Best Practices for Managing a Cognitive Business Operations (CBO) Journey Proven Method. Cognitive, in this context, is about extracting intent from unstructured text using natural language and thus improve process execution. By adding Watson services to a BPM/ODM installation, organizations can improve interactions, improve assignments, improve advice and deliver faster reactions. For instance, using Cognitive to triage emails and then decide how route them appropriately to the right process. The challenge is to do this kind of thing quickly so IBM uses a design-thinking approach and create a “garage” feel so a small team works incrementally.
The Bluemix Garage Method is based on a lean startup method focused on strong collaboration, iteration, design thinking etc:
- Ignite with an innovation workshop
- Design with a design-thinking workshop to create a quick win/minimum viable product (MVP)
Move directly to a broader design
- Realize with a prescriptive project
- Scale with a lean program
He focused on the early steps in this process. First the initial Innovation Workshop. This is designed to deliver:
- Business Goals
- Business Entities – nouns or objects
- Business Decisions including mapping to data types (unstructured or structured)
- Data Science Feasibility to see how it will be to acquire, process and use the data
A good CBO opportunity will almost always have a process context. Business drivers for CBO include engagement, scaling expertise or processing new “dark” data – especially unstructured data. Common Watson services for these opportunities include natural language and speech handling, visual recognition and sentiment/tone/personality identification.
Second the Design Thinking workshop is focused on user outcomes and on very rapid iteration, accepting failures but measuring and assessing progress – even using A/B testing to formalize this testing. The workshop is short and goes through understanding the problem, exploring the area and defining the solution. Example deliverables:
- Understanding: Empathy map for each persona in terms of what they see, hear, feel, say and do. This identifies potential problems that can be addressed, ways to support the persona or eliminate something that is painful for them.
- Understanding: Add some detail to how decisions are made with the identified data, specifically the cognitive services and logic needed to make each of the decision (I would use a decision model).
- Exploration: A model of the to-be process
- Exploration: Feasibility matrix for unstructured data
- Exploration: MVP goals and validation for those – Hills in design thinking terms – and no more than 3 of these saying who will be able to do what when the solution is done
- Realization: MVP efforts based on 2 week efforts: many iterations, rapid feedback and assessment against the goals identified. Learn, Build, Measure, repeat. Each component has its own cross-functional squad.
The end result generally involves a coordinate set of cognitive and decisioning services on Bluemix, connected to the organization’s business processes. Several of these services will be based on machine learning technologies that will also need to trained with data in the usual analytic way – whether unstructured, semi-structured or structured data. The detailed methodology is available from IBM.