THANK YOU FOR SUBSCRIBING
How to Become an AI-Driven Company?
By Christoph Windheuser, Global Head of Artificial Intelligence, ThoughtWorks, Inc
When the term “Artificial Intelligence” or “AI” was born in 1956, everybody believed that a fully intelligent machine would be built in less than 20 years. A couple of “AI winters” and “AI summers” later, there is again a huge optimism about the capabilities of AI. This time, there are clear signs that AI will stay and will have a growing impact on everything from the way we do business to our daily life.
Industries around the world are spending billions on developing AI applications. The research firm IDC has estimated the global spend in AI at 35.8 Billion Dollars in 2019, which is a 44 percent jump compared to 2018. IDC is expecting that this number will more than double in three years (IDC Worldwide Semiannual Artificial Intelligence Systems Spending Guide 2019).
The reason for this hefty investment is that AI algorithms have proven in numerous applications and products that they are able to dramatically create value and improve user experience (just look at your phone in your pocket!). This trend is expected to accelerate in the coming years due to faster hardware, more data and new scientific breakthroughs due to the huge global research investments into AI. So, AI is probably here to stay!
How to Escape the “Ripe-for-Disruption” Quadrant?
Google was the first company who declared itself “AI First” back in 2017. Many companies have since followed, and a lot have experienced that it is easy to announce but hard to actually implement. So what does it mean to become an “AI-Driven” company? Companies have to improve their capabilities along two fundamental axes:
First, they must excel in their ability to create insights out of their data. “Data-savvy” companies are using mathematical data science and machine learning approaches to model the reality based on data. They are able to forecast important business parameters like customer demand, market behavior, and price developments continuously in real-time.
Along the other axis, companies have to improve their ability of continuous adoption based on insights. Agile enterprises have installed continuous delivery (CD) frameworks for their software development processes to almost instantly adapt their IT systems to new insights.
The Cycle of Intelligence
If we look at the overall process of how data and information flow through an enterprise we recognize a cycle: the cycle of intelligence.
In the first step, data is acquired from the outside world. This is raw data in technical stream or file formats.
In the second step, this data becomes useful information by means of storing, cleaning, curating and featurizing the data.
In the third step, the information is modeled with mathematical models based on data science and machine learning techniques. By finding models which describe the underlying distributions and rules of the data, it is possible to predict the data which gives valuable information about the future behavior of our environment and business.
Becoming an AI-driven company requires some fundamental changes in the culture of the company
We are getting new insights into our business.
In the next step, we have to “productionize” the insights. The productive IT systems have to be changed to make decisions and run automated processes based on these insights.
In the last step, the decisions and processes based on the data insights are executed. This has an impact on the real world. It creates new data which is again collected and the whole cycle starts again.
In most companies, this circle is broken and full of friction: Teams are working in isolation, data is managed in silos, most of the described steps are done manually and ad hoc. Running through the whole circle can take months or even years. AI-driven enterprises have automated the whole circle with techniques like CD for Machine Learning (CD4ML). They are able to change their business processes based on new data and insights in hours and several times a day.
Four Steps to Become an AI-Driven Company
Becoming an AI-driven company requires some fundamental changes in the culture of the company. Four steps, which are based on each other, are mandatory for this journey.
Defining a Data Governance
Companies have to agree on how to use their data. They have to define who is responsible for the data, who has access, what are the rights and duties of data suppliers and consumers and how to describe the data in a data dictionary. In larger organizations, defining data governance can be a very time-consuming process.
Setting up a Data Platform
The data must be easily accessible and discoverable for everybody involved. Many companies are setting up so-called “Data Lakes” which are (usually cloud-based) storage locations to store all kinds of data. Please take a well-meant advice: Big IT-driven company-wide data infrastructure projects without clear business values are very often doomed to failure.
Create Data Science and Machine Learning Capabilities
To create Data Science and Machine Learning capabilities, teams of data scientists, data analysts, and machine learning experts are setup. It is important that these teams do not work in isolation and develop a “throw-over-the-fence” mentality. Building mixed-teams together with the business on one side and IT, software development and operations on the other side is usually a good idea to prevent this. It is also important to keep in mind that data scientists have a different way of working than developers.
Build Intelligent Applications and Products
Building productive intelligent applications and products is the place where all comes together. Data governance and the data platform ensure that the right data is at the right time, at the right place, and in the right format. Data Scientists and Machine Learning experts are working hand in hand with developers and IT Ops in mixed teams to include machine learning models into productive applications and products. A frictionless automatically working CD4ML framework is managing the versioning of the different artifacts (source code, data, models and model parameters), the automatic testing and releasing of the software and the monitoring of the software “in the wild”. New machine learning models trained on new incoming data can be brought into production in hours.
If you have reached this point, you are a truly AI-driven company! If you don’t care, your competitors might follow this path, leaving you eventually in the quadrant of “ripe for disruption”.
Benefits of Test Independence within a Software Testing Team
Tudor Brad, Director of Quality Assurance, BetterQA
Living up to Expectation: AI,Scalability and Automation
David Gonzalez, Head of Big Data and Advanced Analytics,Vodafone Business