Debunking Common Myths About Operationalizing AI
October 04, 2021

Alan Young
InRule

Share this

Is your company trying to use artificial intelligence (AI) for business purposes like sales and marketing, finance or customer experience?

If not, why not?

If so, has it struggled to start AI projects and get them to work effectively?

Chances are, you're being held back by one or more operational misperceptions that are causing an overwhelming majority of AI projects to fail. To better understand AI's challenges, InRule Technology tapped Forrester Consulting to explore some common myths about operationalizing AI and suggest ways enterprises can overcome their AI challenges.

The report found that companies believe operationalizing AI can generate real value — helping them gain insights about customers and markets and improve business outcomes. They're just having trouble making it happen; operational silos, data strategy challenges, and a lack of resources are standing in their way.

One commonly held myth suggests that there aren't enough use cases to convince leadership to make AI a priority. Turns out, many companies are overwhelmed by having too many use cases. At least three quarters of AI decision-makers have either a manageable number or too many use cases to manage. This number should grow, since more than two thirds of decision-makers expect their AI and machine learning use cases will increase at least slightly over the next 18 to 24 months.

There's also a wide variety of use cases being exercised across business functions. The most popular involve generating insights into competitors, markets and customer behavior. Others include projects focused on innovation, automation, security, business efficiencies and business automation.

A second myth: AI projects are hard to implement because you can't find enough data scientists with doctorate degrees in statistics. Good data scientists are important, but the truth is, you don't need PhD's to start operationalizing AI. You don't need a PhD to work with most of the machine learning modeling tools in the market today. The real challenge is connecting data scientists to the rest of the ecosystem. Internal silos ranked as one of the top three collaboration challenges firms face, keeping data programmers, gatherers, interpreters and users from communicating with each other. The fact that one in four organizations have cultures that do not encourage data democratization makes the problem worse.

Data is clearly a requisite for AI projects, but the myth that you need lots of data managed by massive data systems is untrue. Regardless of the volume of data available, it's the quality that really matters. Data quality ranked second highest among the top challenges firms encounter when using AI technologies. If your data quality is poor, decisions will suffer, and this likely will impact customer experience and the corporate bottom line.

Another myth: AI learns by itself, so you can set it and forget it. This is where a lot of AI projects fail to live up to expectations. AI models need to be nurtured and continually monitored to make informed predictions and/or recommendations. While 71% of AI decision-makers routinely monitor and retrain models, a surprisingly high 28% build and train models and then leave them alone, creating an incorrect, negative perception about the effectiveness of AI. The most successful AI adopters build models with data feedback loops so they can be continuously retrained. For example, AIOps can enhance IT processes within an enterprise. While AIOps allows for real-time continuous data acquisition, the outcome data is important for model updates and insights as part of an ongoing feedback loop.

What can organizations do to better operationalize their AI?

An important starting point is sharp decision-making. Machine learning algorithms need case-relevant context and decision logic to be successfully operationalized. Decision platforms that incorporate machine learning, human decision logic, and other decisioning technologies and techniques can help scale AI projects, turning them into an integral part of your business strategy. AIOps anchors machine learning, decision automation, digital process and advanced analytics to automate and improve governance of repetitive tasks, freeing teams to focus on new mission critical problems with higher ROI — resulting in faster and more effective completion of projects and higher-impact business outcomes. Forrester data shows that more than two thirds of all enterprises are currently implementing AI and nearly all will be doing so by 2025. Getting up to speed on AI will pay dividends in the future.

Alan Young is Chief Product Officer at InRule
Share this

The Latest

December 08, 2022

Industry experts offer thoughtful, insightful, and often controversial predictions on how APM, AIOps, Observability, OpenTelemetry and related technologies will evolve and impact business in 2023. Part 4 covers monitoring, site reliability engineering and ITSM ...

December 07, 2022

Industry experts offer thoughtful, insightful, and often controversial predictions on how APM, AIOps, Observability, OpenTelemetry and related technologies will evolve and impact business in 2023. Part 3 covers OpenTelemetry ...

December 06, 2022

Industry experts offer thoughtful, insightful, and often controversial predictions on how APM, AIOps, Observability, OpenTelemetry and related technologies will evolve and impact business in 2023. Part 2 covers more on observability ...

December 05, 2022

The Holiday Season means it is time for APMdigest's annual list of Application Performance Management (APM) predictions, covering IT performance topics. Industry experts — from analysts and consultants to the top vendors — offer thoughtful, insightful, and often controversial predictions on how APM, observability, AIOps and related technologies will evolve and impact business in 2023. Part 1 covers APM and Observability ...

December 01, 2022

You could argue that, until the pandemic, and the resulting shift to hybrid working, delivering flawless customer experiences and improving employee productivity were mutually exclusive activities. Evidence from Catchpoint's recently published Site Reliability Engineering (SRE) industry report suggests this is changing ...

November 30, 2022

There are many issues that can contribute to developer dissatisfaction on the job — inadequate pay and work-life imbalance, for example. But increasingly there's also a troubling and growing sense of lacking ownership and feeling out of control ... One key way to increase job satisfaction is to ameliorate this sense of ownership and control whenever possible, and approaches to observability offer several ways to do this ...

November 29, 2022

The need for real-time, reliable data is increasing, and that data is a necessity to remain competitive in today's business landscape. At the same time, observability has become even more critical with the complexity of a hybrid multi-cloud environment. To add to the challenges and complexity, the term "observability" has not been clearly defined ...

November 28, 2022

Many have assumed that the mainframe is a dying entity, but instead, a mainframe renaissance is underway. Despite this notion, we are ushering in a future of more strategic investments, increased capacity, and leading innovations ...

November 22, 2022

Most (85%) consumers shop online or via a mobile app, with 59% using these digital channels as their primary holiday shopping channel, according to the Black Friday Consumer Report from Perforce Software. As brands head into a highly profitable time of year, starting with Black Friday and Cyber Monday, it's imperative development teams prepare for peak traffic, optimal channel performance, and seamless user experiences to retain and attract shoppers ...

November 21, 2022

From staffing issues to ineffective cloud strategies, NetOps teams are looking at how to streamline processes, consolidate tools, and improve network monitoring. What are some best practices that can help achieve this? Let's dive into five ...