Artificial intelligence, business intelligence and content intelligence allow organizations to collect data and then make sense of it in meaningful ways. How are these different types of intelligence being used to make things smarter, faster and better? Our panel this month offers some solid advice.
What new trends can we expect to see in BI this year?
Sam Babic: Increase in technologies that support data democratization. Things like cloud storage, federation and visualization will come together in a more meaningful way. Augmented with AI technologies, it will allow business users to make more informed decisions.
Chris Huff: Security is front and center, because it’s not just a technology issue, it’s a people issue. Lives are impacted by security breaches and I think there’s a new level of responsibility that’s being legislated on technology companies to secure customer data. Look at GDPR and the level of fines that are being levied against large technology companies. Business intelligence will be used alongside intelligent automation platforms and infrastructure monitoring tools to identify “enterprise vulnerabilities” before they become breaches. BI will be the “face of automation” by leveraging adjacent data ingest, process orchestration and advanced analytics to create compelling visualizations that depict a holistic posture of enterprise security.
Kevin Kern: This year we will continue to see the rapid evolution of BI. The development of augmented analytics platforms that use machine learning to generate more contextual information from the various sources of data within the organization and externally will continue to move BI beyond data and visualization, essentially bringing the data to life. And with new players like Amazon entering the market users will provide more options than ever to deploy cutting edge BI tools.
Terry Simpson: From my vantage point I see organizations “wiring” up BI reports and dashboards to the source of the data more and more. Historically reporting has been done from the database or record or an intermediate location that is synced up periodically (hourly or daily). I see process owners sending data to the BI dashboards directly from the workflow and forms that originate the information. Having this real time data is becoming more relevant as reporting engines are becoming easier to manage and configure.
Give us one good example of BI in action
Babic: A university leveraging BI to understand student dropout rates. This example touches on many of the points here. Federation is required to get a view of data across multiple systems. Everything from transcripts, both current and historical, to attendance, to meal card usage. Looking at the data from multiple sources can be used to see new patterns and determine the likelihood that a student will continue their education. Leveraging BI to increase student retention improves revenue, graduation and even placement rates. This example is one that also requires careful anonymization and anti-bias practices. Conclusions could be drawn based on traits such as gender, race and socioeconomic status that would be detrimental to those groups.
Huff: Financial service companies are finding themselves building robust intelligent automation solutions to detect fraud. This is critical as companies seek more straight-through processing. BI tools can be used as an effective “gate guard” where transactions that exceed certain thresholds are aggregated and held for “human in the loop” review. In one instance, we had a customer that was struggling to make payments within 30 days to avoid interest payments. To improve throughput they automated and optimized processes, but found themselves making more misdirected payments (and seven-figure payments). To address this, we built a predictive algorithm/model using historical transactions and applied the model with preset thresholds to current transactions. Robotic Process Automation (RPA) tools were able to identify the flagged transactions and feed them into the BI tool for analysis and adjudication. Instead of clawing back misdirected payments, the customer was able to identify potential problem transactions before they were disbursed.
Simpson: The world I live in on a daily basis revolves around automating processes in the business. A lot of effort is placed on gathering requirements around that process and ultimately automating it. Once the process is automated, testing is done and users are adopting it, most organizations stop at that point. I see really good use of BI when the organization builds dashboards around that process to then keep improving it. The onboarding of employees is a great example of this. From the job application all the way through an employee having a 30-day review post hire, organizations have a huge opportunity to automate and, more importantly, run reports on the effectiveness of this automation. BI is all about making better decisions in the business. Analyzing and optimizing your processes is critical. Automating them isn’t enough.
How has the cloud affected BI?
Babic: Things like cloud storage make the data more available, separating it potentially from on-premise silos.
Kern: The cloud has enabled organizations of all sizes to access powerful insights into data from any location on any device with a far lighter load on the IT department while providing the flexibility to scale to changing business requirements. It also opened up sophisticated BI tools to smaller companies in a reasonably priced subscription model.
Zachary Jarvinen: The cloud has been transformative for enterprise BI. It provides several “force multiplier” benefits. First, its elastic storage capacity supports very large data sets and on-the-fly expansions. This big data capacity is important because organizations nowadays can both easily amass more information to analyze (including text and other forms of unstructured data) and derive more accurate predictions and useful insights from larger data sets. Second, the cloud speeds up time-to-value with rapid deployment and implementation, while reducing IT investment by eliminating expensive on-premises hardware and associated staff – both of which reduce the risk in starting or expanding ambitious BI projects. Finally, versatile data input and data engineering techniques provide more flexible output and distribution options, with better, more on-target results.
Simpson: The cloud and BI is a familiar story. I recall when platforms like SharePoint and other BI tools were 100% on-prem. As offerings like SharePoint Online within Office 365 and other cloud based platforms came available, adoption was slow. BI tools are becoming more and more feature rich and accessible. Just like SharePoint, adoption of cloud based BI is a slow progress as well. As comfort with cloud security increases, the adoption of cloud-based BI is increasing as well.
What are some of the biggest issues affecting BI today?
Babic: Data privacy and bias are potentially some of the biggest issues. Putting data in the hands of additional roles within the organization may require anonymization practices to cleanse the data of any personally identifiable information. In the hands of a human, such anonymization not only protects individuals but also helps combat bias. Anonymization practices aren’t necessarily required for AI, although there have to be considerations for bias. One of the biggest challenges will be preventing bias in both humans and machine algorithms.
What role does visualization play in BI?
Babic: Visualization is key to data democratization, allowing the average business user to visualize insights in a meaningful way. It is also necessary to communicate value up the organizational chain. When making a proposal that is supported by data, visualization is useful in communicating the business decision.
Jarvinen: Data visualization is an essential part of BI. Humans evolved to depend on sight, and can perceive findings more quickly and intuitively through visualizations than tables of numbers. Visualization (including shape, position, color, size and motion) allows you to vividly and convincingly demonstrate the relationships, trends, and patterns in data. And it helps to communicate a concise narrative, uncovering insights that drive results.
Kern: Visualization was and is key to BI as it essentially liberated the data to tell its story. Not too long ago data was basically reports pulled from the data warehouse, ERP Excel spreadsheets or file cabinets. Visualization gave users the ability to aggregate and graphically represent the data providing far better insights into the data and drive better business decisions.
How can businesses make the best use of AI?
Huff: I’ve always advocated for a crawl-walk-run approach to AI. I’m also for practical AI that can be implemented at scale. Examples of practical AI include natural language processing (NLP) and machine learning. NLP can be used today, with a high degree of effectiveness, to quickly ingest unstructured data and transform it into meaningful insights. These insights can improve operations, reduce inefficient spend and enhance customer engagement. I also encourage business and IT to collaborate on AI. AI is a business-led and IT-supported initiative. While AI can be applied across IT operations, it’s going to be scaled in the various lines of business, so collaborating to scale AI is critical to unlocking the value needed to justify the investments required.
Jarvinen: The ongoing digital revolution means that businesses have more data than humans can possibly manage unaided. Those that can manage this information overload, and maximize the value therein, are what we call intelligent and connected enterprises. At the enterprise level, organizations should continue experimenting with specific AI and machine learning use cases that make use of the structured and unstructured data they already have. They should also invest in AI tools that enable a wide range of user personas to run projects and benefit from AI-driven insights – not just data scientists but scaled across the company to analysts, line-of-business managers and beyond.
Do the benefits of AI outweigh the risks?
Babic: Yes. It just needs some checks and balances. On the input side, for example, it’s necessary to have an understanding of the data points that you may not want to include in the assessment — things like gender, race, etc. On the output side, it’s important to have an understanding of what potentially went into the assessment to better interpret data that may have bias.
Huff: If you look across all regions of the world there are macro trends that demand AI solutions to augment and empower the human worker so that organizations can work like tomorrow, today. Some call this the new digital workforce where humans and machines collaborate to drive enterprise value. If you look to Japan, you’ll see the world’s third-largest economy, yet an aging population that can’t provide the necessary workforce to maintain GDP output levels required to generate tax revenue needed to support the healthcare system for the older population. In the United States you have the unemployment level at 3.6%, which is the lowest on record since 1969. The United States simply doesn’t have available workers, yet companies require capacity to grow. In Europe, you have new user data privacy and security requirements in the form of GDPR that necessitate automated data protection technology and solutions. AI solutions are no longer a “nice to have.” The needs of the world have caught up with the technology, so we’ll continue to see increased AI adoption, which will force greater governance.
Jarvinen: Without a doubt. AI is one of the most consequential technologies of today and will be in years to come. The ability to generate insights from massive sets of data will amplify our skills, improve decision making, and make us more productive. AI is poised to help us tackle society’s most difficult problems by providing a deeper understanding of the information we have available to us. In addition, the automation of many routine tasks will allow workers to focus on more strategic projects that drive growth and job creation.
Kern: It’s a tricky question. I think the rapid advancements in AI, machine learning and technologies like natural language processing are already revolutionizing all aspects of science, business and government. We see the evolution of cognitive technology providing even more insightful BI tools. But like with any powerful tool it comes with the intrinsic need to understand the limitations of the technology and the understanding that AI should be used in the service of humans, not to replace them.
What do you wish we could do with BI that we are not able to yet, but you see it coming?
Babic: BI and AI need to be more verticalized. Tools to aggregate and visualize data aren’t enough. Even pushing power into the hands of the business user may not be enough if the business user needs a data science degree to understand how to configure the tools or to interpret the results. Vendors need to provide point solutions that leverage analytics and machine learning to target specific industries — solutions tailored not only for an industry but meant to answer the “question.” Using my previous higher education example, a point solution that targets improving student retention rates. That point solution allows for inputs to support the use of dashboards, and visualization that support the conclusion. The solution could inherently account for anonymization and bias so that each organization does not have to learn from their own mistakes.