David Drai is the CEO of Anodot, a machine learning-based business monitoring solution.
More than 59 zettabytes (ZB) of data will be created, captured, copied and consumed in the world this year, according to IDC. In this fast-paced digital environment, organizations are struggling with traditional dashboards and business intelligence platforms to see the entire picture, as the volume of data isn’t being fully analyzed and opportunities to proactively act on data-driven insights are missed.
When Gartner published its “Top Trends in Data and Analytics for 2020,” highlighting the “Decline of the Dashboards,” they noted that data and analytics leaders need to regularly evaluate their existing analytics and business intelligence tools to offer new augmented and natural language processing (NLP)-driven user experiences beyond the predefined dashboard.
In reality, companies that continue to simply react to problems—as opposed to anticipating and eventually mitigating them—will continue to see a negative impact on their bottom line, and old-school BI dashboards may not solve the challenges of forward-thinking companies. The next generation of industry disruptors needs real-time visibility into the ongoing state of their business.
So, how can businesses transform their monitoring strategies and see significant improvements from business initiatives such as increased revenue, reduced costs and even improved customer satisfaction?
Business leaders must consider leveraging AI to monitor and learn the behavior of data in real-time. With the proper implementation, AI-based solutions can help businesses detect potential anomalies, prevent mishaps and ultimately plan for the future with as much data-driven foresight as possible. In this article, let’s look at some of the challenges and how businesses can solve these challenges in order to reap AI’s benefits.
Challenges In Implementing AI-Based Solutions
Though promised as a business panacea, utilizing AI as a solution to all business problems is never quite as easy as it may seem.
The most frequent failure when implementing AI-based solutions occurs when business management and leadership fail to establish clear and consistent success criteria. To put it simply: “If you don’t know where you’re going, how will you know when you get there?” Innovation leaders, particularly inexperienced ones, sometimes view AI as a “magic wand” for growth. But this mindset often yields success criteria that are amorphous or unreachable, which makes it impossible to achieve business goals—AI or no AI.
Challenges also arise when internal pain points in need of solving are not actually suited to AI in terms of accuracy, speed and/or scale. Similarly, some projects are simply too big or traverse too many internal silos to be handled by a single AI tool. In short, just because there is an issue doesn’t always mean that AI is the solution.
On a cultural level, resistance from within the organization can also be a barrier to AI adoption. Talk of incoming AI often fuels anxiety from certain employees that they will become redundant. Others are anxious that training will be insufficient for processes designed for data scientists rather than regular employees. Over 23% of analytics challenges have been attributed to internal cultural resistance, so the concerns of even a few employees should never be overlooked.
How To Tackle These Challenges
To avoid unclear and inconsistent success criteria, organizations must try to hire or reposition experienced product managers and data scientists—perhaps even a Chief Data Scientist—who understand both business as well as AI. The right product manager will be able to align with data scientists and translate expectations and results to management, bridging the gap between the data science teams and the company at large.
Likewise, data scientists must shift their mindset and prepare to monitor the quality of machine learning (ML) models while they are in process and not just in the planning stages. Why? Because ML does just that: learns and therefore changes as it is applied. So, these models must be monitored closely during the production process so that issues are detected and addressed in real time before resulting in compounded problems.
When determining which pain points to address with AI, innovation leaders must first verify that a given problem is indeed well suited to AI or ML solutions. If research reveals a solution that is more deterministic than AI in terms of speed, accuracy and/or scale, then it’s probably best to go with that solution instead.
For areas where AI may be relevant, but the size of the project poses complications, start by breaking down the issue into a series of smaller problems that can be solved one at a time. The field is still young and evolving, so trying to force the same underlying AI solution onto a variety of projects, i.e., trying to fit the problem to the tool, may lead to wasted resources, frustration and even failure of the project.
Lastly, it is critical that company leadership does their best to assuage employees’ concerns about AI—or at least be transparent about what employees should expect from any incoming changes. Proper training will help dispel any myths and help employees understand that AI processes can be executed by staff other than just data scientists. Training will also demonstrate that, rather than taking away jobs, AI will streamline processes and itself execute many mundane tasks, freeing people up to do more interesting tasks better suited to their skill set.
Steps Toward Successful Implementation
Blindly adopting AI because it’s trending is a recipe for failure. Leaders must first strive to understand how any AI-related transformation will complement existing structures rather than replace or hinder them. Following implementation, teams must continue to monitor these AI solutions to identify any room for improvement (and continuously do so) or to determine when such processes fall short of those solutions promised in the research phase.
Again, don’t expect magic. AI isn’t witchcraft—just numbers, analysis and the critical insights they yield. The magic comes when those who implement AI solutions do so with the end user in mind.