Search Icon Expand Search Input Field
close
Altair_Blog_hero_1920x225

Featured Articles

Organizational Friction: Examining Data and AI Friction Part One

A comprehensive, seamless data and artificial intelligence (AI) strategy can revolutionize what organizations are capable of. The transformational power of modern data analytics tools – powered by the latest machine learning and AI software – can maximize potential and get the most from teams and individuals all over the world. It’s no wonder why people are hurrying to implement organization-wide data and AI initiatives – but all too often, these projects run into roadblocks that can cause delays, resets, and even outright failure. These roadblocks are commonly known as friction. 

 

The Three Main Types of Data and AI Friction

Earlier this year, we conducted an international, online survey of 2,037 professionals employed throughout several industries who work with data in some capacity to drive valuable insights for their organization. The goal of this survey, among other things, was to see how, where, and why friction is affecting today’s organizations and teams. After examining the data, we discovered that organizations mainly deal with three main types of friction:

  • Organizational friction: Friction that exists between departments, teams, and individuals. Organizational friction can affect organizations both “horizontally” (between different teams and domains) and “vertically” (between departments, teams, and individuals at different levels of seniority or job function).
  • Technological friction: Friction that stems from technology infrastructure — this includes hardware and software resources, cloud and high-performance computing (HPC) resources, appliances and plugins, vendors, and more. Often, technological friction acts as a bottleneck by limiting projects’ speed, scale, and/or scope.
  • Financial friction: Friction that presents itself when budgets are tight, resources are spread thin, and projects need to provide a return on investment. Financial friction can be most apparent when trying to invest in a new initiative or trying to scale efforts with expensive, legacy investments. 

In this first entry of a three-part miniseries that will dive into what our survey data revealed about data analytics and AI friction, we’ll turn the spotlight on organizational friction. 

 

Causes of Organizational Friction

Put simply, organizational friction is a people problem. As with any successful business strategy, people have to be on the same page with one another – they also have to have the right tools and skills to solve any problems they encounter. As our survey data revealed, many organizations deal with, and suffer from, organizational friction. 

To begin, 84% of the survey’s respondents said their organization faces limitations that slow down AI initiatives sometimes or often; 19% said “often,” while 13% said “very often.” Only 10% of respondents indicated they experience limitations rarely or never. In addition, 63% of respondents agreed with the sentiment their organization makes working with AI tools more complicated than needed. And 35% of respondents said that AI literacy is low among the majority of their workforce.

But why do people feel that data and AI strategies are more complicated than they need to be? Why are there so many frequent roadblocks? More than any other data point in this section, the survey revealed the extent of the data science talent shortage worldwide. Overall, 75% of respondents said their organizations struggle to find enough data science talent. Moreover, when asked to list the most prevalent problems in organizational AI strategy adoption, respondents named the shortage of talent and/or the time it takes to upskill current employees’ skill sets as the most common problem at 54%. Not having enough data science-specific talent is a major source of friction, especially when other employees — likely without sufficient data science expertise — are made to oversee and manage organizational data and AI initiatives. Moreover, employees’ lack of data science expertise can cause problems in the other two main areas of friction. 

 

Effects of Organizational Friction

When teams aren’t aligned or are trying to direct projects they don’t have the skills to handle, it can cause severe negative consequences for individuals, teams, and organizations alike. Friction has a nasty habit of denting morale, wasting employees’ time, and hurting the organizations’ bottom lines. Our survey data shows that wasted time/effort and money accounts for organizations’ main organizational friction concerns at 63% and 40%, respectively. 

It’s also important to note that organizational friction affects organizations both horizontally and vertically. The data revealed some insightful trends on how friction is hurting organizations vertically. 33% of executive-level respondents said teams such as data scientists and business leaders are working in silos and therefore communicating ineffectively. User-level respondents said the same at just a 25% rate, an 8% difference from their executive counterparts. Other data points reinforced a vertical disconnect between users and executives as well. For example, while 42% of executives said they’re encouraging others within their leadership team to understand the rationale or benefits of investing in AI solutions, just 32% of users said the same. In addition, when asked if they believe their organization can scale AI projects without training domain experts in data science, 69% of executives said “yes,” while just 51% of users said “yes.” 

 

Attempting to Remedy Organizational Friction

Naturally, organizations dealing with friction are taking steps to minimize it. In terms of organizational friction, this usually revolves around upskilling employees so they’re better equipped with data science skill sets and knowledge. This often happens through dedicated data science enablement programs. But as our data showed, these programs have problems of their own.

To start, most organizations only have a data science enablement program in place in limited areas of the organization (47%). 43% of organizations have data science enablement programs in place throughout the organization, while 10% said “nothing of significance is in place.” The limited nature of many programs is troublesome and prone to cause friction, but there’s another stat that’s more worrisome. 96% of executive-level respondents said there was some sort of structured data science enablement program in place at their organization, while just 83% of user-level employees said the same. This indicates that many users either aren’t aware of programs, or that these programs are so insufficient that users don’t think they’re valuable in helping them develop their data science skill set. 

 

Conclusion

When it comes to harnessing the game-changing power of organizational data and AI strategies, friction is a killer. Friction – in all its forms – is so prevalent that 42% of respondents admit they have experienced AI failure within the past two years. Additionally, 33% of respondents said more than half of their data science projects never made it to production within that same timeframe. Without a doubt, friction is the main reason data and AI projects fail.

As far as organizational friction is concerned, the data shows that a shortage of talent and a lack of data science skill sets is hampering organizations’ data and AI goals. We also see that there’s a lack of alignment between executives and users on multiple fronts, especially as it concerns data science enablement programs. These problems and their effects ripple throughout teams and organizations and can also cause friction in other areas. In attempting to remedy organizational friction, organizations and leaders must ensure that people are aligned in their objectives and equipped with the data science tools they need to succeed. 

Click here to read the next part in this series, "Technological Friction: Examining Data and AI Friction Part Two." And be sure to check out more of Altair’s Frictionless AI resources, including: