Module 4: AI and Business Support Functions
Video Transcript
Video 4.5: Managing Systems: AI in IT Operations
The third enabling function where your organization spends a lot of its money is IT operations or
managing systems. Now, here again, we have created an operations life cycle for IT operations.
Now, while your IT workflows will vary significantly based on your industry, but we can describe
them as a set of standardized workflows at a higher, at a more abstract level. And these
workflows again consist of five stages.
Design, which is about defining a system architecture, the resources that you need and making
sure that your IT infrastructure is aligned with your business strategy.
Develop is actually writing the code, building your user stories, creating requirements and
writing the code.
Deployment, which is to run test scenarios, fix the bugs and deploy your applications at scale.
Monitor, which involves managing, securing and optimizing the performance of the
infrastructure, your tools, your networks, your cloud providers and your products.
And support, which is provide end-user support, respond to incidents such as security threats or
breakdowns or failures and to continually enhance your features and performance of your
applications. So, that is the IT Ops life cycle from an application standpoint.
Let's look at some use cases here. At the front end, designing and ideating.
Airbnb is using computer vision and classification to automate the product ideation and design
process. Literally, now what you can do is to draw out what your wireframe needs to look like
and generative AI tools will design the wireframe for you. So, they have training data that is fed
into a computer vision algorithm that is then connected to a component library so that you know
which component you're going to use for what application. The technology teams are able to
work with business stakeholders, literally, take a sketch, an idea into a wireframe and into an
application, so design and ideation.
Data processing and integration, which is a very tedious task and a very expensive task. So
here, intelligent process automation allows your developers and data administrator to automate
repetitive workflows related to data collection, data extraction, data transformation, loading and
logging.
So, whether it's an application development where you have automated tools to collect input,
read documents and to track reports, data administration, integrating data across systems and
automating what we call ETL, extract, transform and load workflows and then, logging and
compliance, making sure that you are monitoring the performance, generating the reports,
© Kellogg School of Management | Northwestern University. All rights reserved Page 1 of 2
looking for any discrepancies. The data processing and integration can be automated through
AI.
Then what about bugs? Identifying and managing bugs, this is again fascinating use case for
generative AI now. So, a company like Google, which has billions of lines of code, think about
the fact that nearly 60% of the Google code base changes monthly as its systems grow in
complexity.
So, Google is actually now using language model, LLMs, to analyze coding patterns, historical
bug data and developer behaviors to identify likely hotspots and predict where bugs might come
even before the review happens. So, these machine learning algorithms are forecasting where
bugs are likely to occur, the severity of the bugs and to prioritize the fixes, particularly for
mission-critical systems.
So, Google is combining generative AI tools, like their internal copilots to suggest code fixes
automatically for identifying bugs, reducing the manual debugging process. So, this entire
process of identifying, predicting and fixing bugs is being automated with AI and generative AI.
Then once you've implemented your applications, you can maintain and run your IT operations
autonomously. So, here we're giving you an example of a case company Arago that has built a
platform called Hiro that allows your IT operations to run almost autonomously, 75% automation
across the IT stack. By understanding when, which is the situation, where, based on
environment and facts, what action needs to be taken and which policies and rules to apply.
By the way these are the parameters that you need for an agent in agentic AI. So for example,
when disk space is low, then the machine serving the, and where the machine serving the disk
space fails, then remove the log files to solve the issue and take a backup before you delete the
files. That's an example of an agentic operation. So, there's a vast potential to automate various
aspects of running your IT operations in autonomous mode using agents.
Just like we had a vision for finance, what about a vision for IT operations? The vision for IT
operations is autonomous IT operations.
We strongly believe that AI tools, agentic AI will allow increased levels of autonomy, eventually
leading to what we call zero touch. IT operations management, automated product backlogs,
automated product portfolio optimization, dynamic grooming of applications, autonomous
infrastructure that operates on its own and heals itself, self-corrective support.
So IT operations, again, given the scale, given the complexity and given the repetitive nature of
the task to be performed and the magnitude of the data that needs to be analyzed are a very
target-rich environment for automation using a genetic systems and autonomous AI
technologies.
So that's how we see AI, generative AI, and agentic AI transforming the process of managing
their ideas.
© Kellogg School of Management | Northwestern University. All rights reserved Page 2 of 2