Guha Athreya Bhagavan, Director, Data Science, Grainger
In their current forms, Artificial Intelligence (AI) and Business Intelligence (BI) automate the mundane and aid decision-making respectively. Their confluence presents opportunities to do more. It offers two transformational opportunities:
1.Accelerated Data Democratization
2.Embedded Analytics and Business Processes Transformation
Accelerated Data Democratization
The democratization of data makes it accessible and facilitates self-serve analysis by the user. While many organization shave already started this journey, there is an opportunity to accelerate.
• Why: The most significant upside is a shift towards fact-based decision making. With all the hype around sophisticated machine learning, it’s easy to forget that even today, managers frequently argue over facts. Sometimes this creates a culture of compromise masked by collaboration. Once this problem is solved, driving the adoption of more sophisticated technologies becomes easier.
• How: The key is to define ease of access from the user’s point of view. The user may know how to navigate a spreadsheet but may not like the task. They may not know SQL or pivot tables. More importantly, all of this may seem to distract in the context of their roles and priorities. There can also be strong preferences for specific formats ranging from printing every attachment to only audio-visual content.
• What: The current state may be that too many bulky reports land up in everyone’s mailbox. The desired end state is often a single source of truth for rich data available on demand in an interactive and intuitive format. AI can help users query data through a voice or touch interfaces and present relevant information in audiovisual formats. This goes a long way in making data accessible to the non-technical user.
Managers who frequently use data to make better decisions are likely to understand the value of sophisticated algorithms.
Opportunity 2: Embedded Analytics and Business Process Transformation
Embedding predictive models in business processes and tools helps large user groups in decision-making. AI can help solve the harder prediction and classification problems involved. It can also help generate predictive insights at an industrial scale.
• Why: Delivering personalized value to customers requires decentralized decision making. This diffusion requires surfacing relevant insights seamlessly at each step of the process.
• How: Executives often complain that AI-based solutions are hard to integrate with existing business processes. They are right. AI is most potent when you allow it to transform operations. The root cause is often a poorly defined problem statement. Sometimes, it may be necessary to redefine misclassified BI and AI efforts as Business Process Transformation programs.
• What: Business process transformation involves looking at each step of the process to ask which actions can be executed better, faster or simply automated. Often this eliminates some steps or changes the sequence of execution. In some cases, the new process can look significantly different from the current version. Enablers - People, Products, the elusive 3rd
Past efforts to realize such opportunities have resulted in a healthy mix of successes to celebrate and opportunities to learn from failures.
People:Data Scientists and Systems Designers bring complementary skills to the table. Scientists know how to design experiments, examine empirical evidence and generate insights from data. Systems designers and developers understand process and data flow, entity relationships and state transitions. More importantly, they also have different ways of thinking and soft skills. Scientists are known for their intellectual curiosity, logical reasoning and the kind of rigor that is evidenced by methods like recursion. These are entirely different from the systems and design thinking mindset of using empathy to understand user requirements and innovation to overcome technical challenges.
From recruiting to project staffing, it is critical to ensure a good mix of skills from these disciplines. Most important of all is the collaboration that helps people with these diverse backgrounds to work together effectively. The acronym C.I.R.C.L.E is a simple way to remember these ingredients of success. Products: Both proprietary products and the open source ecosystem offer an ever-increasing array of options.
Mature users of technology can sidestep false dichotomies (e.g., build vs. buy) and strengthen foundations to then integrate a variety of micro-services. Put another way, when you have the right people, this problem solves itself.
The elusive 3rd P is Process. It’s hard to define a sequence of steps to follow when the goal is to transform other processes. We can, however, build a repository of guidelines and best practices.
As a first step towards researching this, eight best practices were shown to a group of about 100 CXOs, BI professionals,and Data Scientists. They were asked to identify the ones they thought were relevant. Here are the best practices and their perceived relevance.