AI and Analytics: Coming to a Process Near You
Enterprises are increasingly pushed for faster insights from their ever-increasing data volumes. A TDWI senior analyst looks at how some vendors are responding.
- By David Stodder
- April 17, 2020
Accelerating speed to insight from data is critical to nearly all types of organizations, especially as managers seek to develop strategies for responding to unexpected and rapidly changing circumstances such as the global coronavirus outbreak. TDWI's recently published Best Practices Report, Faster Insights from Faster Data, takes an in-depth look at practice and technology issues that matter most in reducing delays in data life cycles and putting well-prepared and relevant data in the hands of users sooner.
Not long after the publication of the report, I had the opportunity to visit with some technology providers exhibiting at the TDWI Las Vegas Conference and Strategy Summit in February. It was interesting to see how issues brought up in the report are being addressed by vendors.
Technology trends are increasing the opportunities for faster insights and the challenges of deriving them. Artificial intelligence (AI) techniques such as machine learning and software automation are advancing fast and will reduce the need for specialized hand-coding. Yet, these technologies require careful direction, maintenance, and good data. The flow of data today is not only voluminous but also faster thanks to the adoption of technologies such as the Internet of Things (IoT) and data streaming.
In the Best Practices Report, we discuss the role of data streaming and real-time analytics. However, as important as these and other technology developments are, factors such as project organization, governance, business and IT leadership, self-service capabilities, and the use of methods such as agile and DataOps are just as critical. The report examines these aspects as well.
Delivering Actionable Dashboards
To understand some of the key drivers, TDWI asked the report's research participants what would improve if their organizations increased investment in technologies that support "fast batch" and/or delivery of true real-time data, including data streams. Topping the list for about two-thirds of participants was actionable information in dashboards (67 percent), followed by operational decisions and management (56 percent) and performance management (42 percent).
These results show organizations' strong level of interest in delivering the most up-to-date information possible to operational personnel so they can increase efficiency and effectiveness of operations and align daily decisions with strategic performance objectives.
Dashboards help us visualize data and in many cases gain easier access to underlying data through graphical interfaces. However, a challenge many operational managers and frontline personnel confront is that they have too many dashboards, each associated with a different business intelligence (BI) system, application, or specific report. They have to go from one to another to get a complete view of data relevant to a specific situation.
Metric Insights, exhibiting at TDWI Las Vegas, addresses this problem by providing "push intelligence" that gives users a consolidated BI portal. The portal employs a connectivity layer to draw data from multiple BI, database, and operational systems such as Salesforce into a single view. Rather than require users to view dashboards constantly, the push system can be personalized so each user receives only relevant updates through notifications and alerts via email, instant message, or other communications system about data changes, anomalies or outliers detected, and exception reports.
Solving Business Problems with AI Augmentation and AutoML
AI can help organizations process fast-arriving volumes of data to enable both humans and applications to make better decisions. In the Best Practices report, we asked research participants to identify the most important ways their organizations currently use (or plan to use) AI to augment BI, analytics, and data integration and management.
The most prevalent choice is to automate discovery of actionable insights (52 percent). Technologies are indeed trending toward enabling organizations to set up algorithms and models that do not require regular human intervention with the ultimate purpose of supplying personnel in operations with insights that improve daily decisions. This "augmentation" will primarily come to users in the form of recommendations; 41 percent of research participants say their organizations want to augment users' decision making by giving them recommendations.
Many leading BI and analytics technology vendors are engineering solutions to deliver AI augmentation. SAS, for example, discussed at its industry analysts conference in February the goal of "hiding AI and analytics in plain sight" to enable users throughout organizations to focus on solving business problems rather than figuring out how to program algorithms -- or (more likely) depending on data scientists to do so. SAS Intelligent Decisioning focuses on automating and embedding AI and analytics integrated with governing business rules to enable greater scale and quality of real-time, operational decisions.
At TDWI Las Vegas, exhibitor dotData explained how applying automation to more stages of data science and machine learning (ML) development processes can accelerate operationalization of analytics and increase the impact of data insights. The company provides an autoML solution that applies automation to the spectrum of phases in data science and ML workflows, from data acquisition and preparation to feature engineering, target selection, training, and model evaluation. AutoML is a major emerging technology trend; it is about automating ML model development and related steps so organizations can move faster toward applying insights to decisions.
Some exhibitors showed how they can customize automation and augmentation to address particular business challenges. MResult, for example, provides cloud services and analytics and process support teams to tackle a range of vertical business problems. It recently introduced "Opptymized," which focuses on the campaign analytics, web and application development, e-learning, and e-commerce needs of small and midsize advertising agencies. These firms -- as well as advertising and marketing departments within small and midsize companies -- are often hesitant about devoting large budgets to analytics and AI but also fear falling behind as these technologies become more prevalent. MResult's solution is an example of how services are incorporating new technologies to help smaller organizations gain data insights.
Dataflix, also exhibiting in Las Vegas, addresses how to use chatbots more effectively. Chatbots are now widely used, but too often their implementation is divorced from the flow of insights from analytics and AI. They can become too canned and unresponsive to increasingly frustrated customers. Dataflix's conversation analytics brings insights to bear on human-to-bot interactions. Dataflix can analyze the interactions in conversation logs and retrain bots using ML and natural language processing techniques. The solution could enable organizations to use analytics to improve and more effectively scale call centers, online customer support, and related operations.
Tightening Analytics and Business Process Integration
Moving forward, we will see more technology solution providers increase their response to demands for faster insights and for more mainstream use of automation and AI augmentation to operationalize the flow of data insights into business processes. More than two-thirds of the report's research participants (69 percent) say their organizations find it either very or somewhat important to automate decisions in operational or process systems. Nearly the same percentage see it as a priority to enable analytics to guide real-time process optimization (67 percent).
Whether to respond quickly to crises such as the coronavirus outbreak or to address more mundane matters such as a shift in customers' buying preferences, organizations need to examine how they can close latency gaps in data and analytics workflows so that critical insights have the biggest impact.