To build skyscrapers, highways, and bridges, first they have to be imagined. After that, they have to be designed, and Autodesk, the producer of AutoCAD, Fusion, and 3ds Max, has been supporting designers and developers for decades.
Prakash Kota, who started as a quality analysis consultant for the company in 2005, and is now in his sixth year as Chief Information Officer, saw the evolution of Autodesk, both internally, and externally. He sat down with Frontier Enterprise to briefly discuss his multifaceted role as CIO, the company’s AI strategy, and how things are looking for the business, in terms of technology.
What’s your typical day-to-day as CIO of Autodesk?
I have been with Autodesk for 19 years, and I had 13 different roles over that time. During the last six-plus years as a CIO, the role has been evolving. I hold the responsibility for all internal staff, including those responsible for all our applications, infrastructure, cloud collaboration, services, engineering, and development-related affairs.
In addition, I also manage global customer operations. When our customers call Autodesk, the worldwide team which is the first line of support also answers to me. Also, our resellers and partner operations, when they have a problem and they call us, my team is responsible for meeting their requirements. And then product network operation centres including product availability, resiliency, reliability, also report to me. In that sense my role is quite broad.
Could you share a bit about Autodesk’s AI strategy?
Autodesk AI is something that we are focusing on for our customers, and our customers’ focus is led by our research and development team. AI is not new for Autodesk. We had integrated AI early on. However, generative AI is now going through a hype cycle.
Autodesk has been leveraging AI for more than 10 years especially with regard to our designing models. When cloud computing became so accessible we provided multiple options to our users to choose from in a sustainable way. In my role, I am heavily focusing on enterprise AI.
When I say enterprise AI, it’s more about looking at all different employee personas, and really seeing how we can make them productive in whatever they do. The timing cannot be better for us, because as a company we are growing and scaling. We have been seeing double-digit growth continuously over several years. We will continue to be doing that.
We have an audacious goal of getting to $10 billion annual revenue eventually, in the years to come. Therefore, generative AI has arrived at the right time for us when we are seeking scale and growth. It’s not just incremental growth; when we become a US$10 billion company, how do we operate? Do we have all the efficiencies that we need? We are looking at every single employee persona when we talk about: sales, marketing, customer success, and finance.
So in the last 18 months, we have been doing several different use cases, experimenting, leaning in, piloting, evaluating, with some of those use cases actually helping us achieve global availability while ensuring scale.
Do you depend on your SaaS provider to infuse AI into your workflows, or do you do it yourself?
It’s a combination, I would say. We started off with something called Autodesk GPT, where we took Microsoft’s private instance of OpenAI. Then a lot of product teams started creating APIs for that private instance, to explore and experiment different use cases, and to examine what will work and what will not.
And then the other next big use case that we went after was our intranet search which is like enterprise search. If you look at any enterprise company with a history of 40-plus years, findability is a tough thing because there is so much data.
Given the pace we are hiring and growing, if you are a new employee who is onboarded remotely, and if you are expected to be productive at work immediately, it might take weeks for you to find all that you need to be productive. Added to that cultural inhibitions can delay the time to productivity.
For this we have leveraged not only the RAG (retrieval augmented generation) architecture, but also OpenAI, behind the scenes,and connected it to all our intranet resources such as our internal knowledge sharing resources.
Today when you search for something internally the tool is more cognitive contextually. For example if an employee in Singapore searches for benefits that are relevant to employees there, it now provides more precise and contextual replies. The fact is that the tool can give an actual answer rather than merely pointing to a link.
The findability of our tool, which used to be one of the areas of concern for employees, whether it is existing employees or new employees, now has led to an 86% favorability rating for it. We conduct regular employee surveys to see if they are getting what they need to be productive. After the integration of generative AI, our productivity tool or what we call enterprise search, has an 86% favorability rating
We are also using the same search feature for our internal chatbot that we have for all employees. We call it Lumi. Irrespective of where you ping and chat from, you get the same answers, because the search and the backend models that you are researching use the same knowledge base.
Chatbots have come a very, very long way in the past five years, especially context-based ones which draw information from all your integrated software suite. Is this also something that you’re planning for your customers as well?
Yes, indeed. We already have something similar, called Autodesk Assistant for which we are leveraging our own LLMs. We are creating our own LLMs and training them on customer data in open public forums where customers are contributing to it. But it is something dedicated for customers, while the other one I was talking about is more internal, for employees. So that’s one use case.
The other interesting AI use case that went live is around marketing. For marketing we generate campaign and product videos that we send to prospective customers to generate leads – typically these videos are in English. Earlier, to localise them we needed to ship it manually to our vendors and they gave us a finished video in three weeks or so. If you are running a promotion video the delay that this process causes and the costs associated with it are quite high.
Leveraging Microsoft translation services, based on generative AI, the translation is automated and it is ready to use for marketers in hours. We created a platform with what you can call virtually a dictionary and it can translate videos into any language. In case you want to update the product name later, you don’t need to even record it again. You go to the dictionary and change Fusion 360 to whatever new name, and it updates the video in English any other localised language instantaneously.
A marketer then needs to only edit it quickly, see whether it looks good, then push it to whatever regions that they need to push. By instantaneously translating all the videos you have the potential to generate more global leads.
How has Autodesk’s transformation been, from physically installing software to end users using a disk drive, up to now, where everything is done in the cloud?
It has been a transformational journey considering when I joined the company 19 years back, we used to ship software in DVDs. And one of the things that I did in the first couple of weeks was to go to a warehouse because every evening, a truck had to leave once all of the DVDs were loaded. Later you have to invoice it back in the system.
Now, you actually swipe your card in our Autodesk Store and you get instantaneous access to our software. The earlier process involved distributors and resellers all which took so much time to get the software to the end user. Then the installation also took more time. To give instantaneous access today we have come a long way. And since we are subscription-based now, there is mutually assured and continuous value exchange between us and the customer. That’s why they expect instant and regular updates.
Customers value recurring, yearly subscriptions. We have 5000-plus engineers in the company. We used to have our own private cloud but now we have ported out those workloads to public cloud providers. For development and testing, we use Azure, and for all of our production workloads, we use AWS.
How do you see the evolution of computer hardware, and what are the possibilities for Autodesk?
Leveraging cloud and all the compute capabilities that we have available now, I think sky’s the limit for what we and our customers can actually achieve. However, I believe with all the different providers competing to introduce more value, the competition is actually advantageous to our customers. New standards are being set in how people operate in the industry. Some of these innovations will eventually become commoditised like how cloud computing has.
