It's no secret that making data science, machine learning, automation, and (eventually) AI pervasive across teams and industries will be a critical factor to success for the enterprise as a whole. Yet it also means deep-rooted change, especially at the tactical and organisational levels. So how can companies cope and move forward?

In 2018, Dataiku surveyed more than one hundred data professionals, and they ranked organisational change as their third biggest data challenge (behind data cleaning and model productionalisation). In 2019, structural change moved up to the second place slot, suggesting that companies have not yet overcome the culture change obstacles to data integration.


Why AI Necessitates Organisational Change

Let’s take a step back and understand why businesses looking to be AI-driven must undergo structural change. Today's line-of-business teams (including marketing, but also human resources, finance - you name it) have no shortage of business questions they want to solve, yet they run into all kinds of challenges when trying to make AI a reality, including:

  • Lack of data or incomplete data.
  • Data projects that rely on limited statistical models (instead of more sophisticated machine learning models).
  • Difficulty deploying and automating models due to complex links with frontend systems.
  • Lack of tools or easy access that allow them to dig into data themselves.

Plus, even when projects do end up being a success, their ongoing maintenance can present continual challenges for non-technical teams (in case you haven't heard, machine learning models aren't like software - they do need quite a bit of upkeep).

The bottom line is that long term success – that is to say, innovation - resides in the transformation of putting data and analytics (or, going a step further, data science and machine learning) at the very core of the company. And this is no easy task.


People, Processes, and Technology

All of this to say that it’s not enough to simply hire a few data scientists or business analysts and put them on a team together to create AI projects; instead, it’s important that data use be pervasive and democratised throughout the business, giving all staff a baseline ability to use data to make decisions. This is the piece that requires huge organisational change: the reality is that most teams and larger companies are not set up to be data driven from the ground up.

But achieving data democratisation doesn't mean all employees need to become data scientists overnight. Instead, the answer is:

  • Accessibility between people, for example instead of the expectation that all business people should be experts in data science (or the dream that each business team has its own data scientist or other data expert on staff), it's critical to build an environment that allows people of different data abilities to work together to solve business problems and build AI solutions.

  • Building processes that allow people at all levels of the business to use data to make decisions. That involves easy (but controlled) connection directly to data sources - no more back-and-forth asking data for IT and sending around spreadsheets! - as well as simple ways to share projects between employees for validation and cross-checking.

  • Investing in technology and tools that not only enable these people and processes, but also are sustainable, ensuring a solid investment in the future. That means not only going with open source, which is cutting-edge, but also considering proprietary software on top to provide a simple user experience no matter what level of technological skill the person has.


How to Get Ahead of Organisational Change Management

OK, so companies are going to be changing with AI. Fast. What are the next steps that both business professionals - and wider businesses - can take now to adapt to the data-driven times?

  1. Education: Be able to speak the basics about machine learning, deep learning, and AI. Or, if you are an executive or manager, empower your teams to have a baseline level of knowledge on these subjects. A few suggested resources are Machine Learning Basics - an Illustrated Guide for Non-Technical Readers, Introduction to Deep Learning, and Data Architecture Basics.

  2. Foster collaboration: Invest in technology (like a data science, machine learning, or AI platform) that can be used not only by data experts, but data beginners, for everything from managing data projects to connecting to data themselves. Check out the white paper Why Enterprises Need Data Science, Machine Learning, and AI Platforms to learn more about what they can provide.

  3. Dive in: Start driving change by choosing at least two or three simple data projects that would help provide more insight or efficiency, and partner with data experts to get started. Why not just one project? Well, data science isn't really an exact science, so it's possible that for one project the right data doesn't exist. Or the project gets executed, and the results aren't helpful. Getting started with a few low-hanging fruits will up the chances that at least one is a success. Get more tips for running a data science POC.


Written by CEO and Founder of Dataiku, Florian Douetteau.

PrivSec Conferences will bring together leading speakers and experts from privacy and security to deliver compelling content via solo presentations, panel discussions, debates, roundtables and workshops.
For more information on upcoming events, visit the website.

comments powered by Disqus