In this exclusive interaction with CIO&Leader, Krishna Prasad, Chief Information Officer at UST, shares how the company is advancing its enterprise AI journey from experimentation to measurable value. He outlines the key pillars of UST’s AI strategy, the infrastructure adjustments needed for scale, and the cultural shifts required to prepare the workforce.
CIO&Leader: How do you define success when transitioning an AI initiative from pilot to production?
Prasad: Any initiative that is considered successful is based on actual usage and delivery of business value in a production environment. AI pilots are no different. While there might be a lot of experimentation during the pilot phase, we do define expected business outcomes before we deploy systems into production. Successful AI pilots are those that result in delivering the anticipated business value in the production environment. In addition, successful deployments have reduced risk, enhanced user experience or improved organizational speed while delivering improved productivity at scale.
CIO&Leader: What are the core pillars of your current enterprise AI strategy?
Prasad: Our enterprise AI strategy is designed to risk reduction and innovation. We look at AI being deployed in a variety of different ways including AI enabled customer facing products and services, internal productivity as well as opportunities for more extensive transformation of the way we do business. Our AI strategy is firmly rooted in identifying and deploying use cases that deliver business value, while protecting our data and enabling our talent. At this stage, we use AI to enable our employees rather than using AI to autonomously run our products, services or processes.
CIO&Leader: What key AI use cases have successfully moved into production, and what measurable impact have they delivered?
Prasad: A good number of our AI initiatives in production have been individual productivity focused. For example, we have deployed the Microsoft copilot and are seeing better and faster responses to customer queries and proposals. We have also deployed an AI enabled tool to help our consultants better prepare for client interviews and discussions. In those cases, we have found almost a 50% improvement in the ability for our consultants to be selected by clients for project work. We have also started to use AI for employee engagement, although it is too early to have quantifiable results.
CIO&Leader: What infrastructure or architectural changes were necessary to scale AI effectively within your organization?
Prasad: AI deployments that were delivered through the SaaS applications that we use have not required any specific infrastructure changes. For many of the internally developed models we have procured and deployed GPUs within our environment, in addition to using cloud-based capabilities. One of the key areas where we have had to make changes is around the classification and protection of our data, which has become a much higher priority than it was in the past. In addition, we have had to review and update our policies to reflect the ethical and responsible use of AI within our environment.
CIO&Leader: What are the biggest challenges you’ve faced in operationalizing AI, and how have you addressed them?
Prasad: Some of the biggest challenges associated with AI deployment have been around managing cost and protecting data. We have found that while pilots give you some indication of how to implement a solution it is hard to get a clear handle on the cost of running the solutions in production and the associated ROI justification. We have overcome this by monitoring the spend and alerting teams before costs get out of hand. Also, since data is not always correctly classified and secured, some of the confidentiality gets compromised when the data is ingested into the AI models. We continue to focus on validating the data that we use to train our models. It has also been quite challenging to find the right talent who bring the appropriate balance of technical and domain skills to find creative ways to solve for business problems.
CIO&Leader: How are you preparing your workforce for scaled AI adoption, and what organizational shifts have been required?
Prasad: Preparing our workforce for the world of AI will likely be our biggest challenge. We are in the expertise business and AI is rapidly disrupting the value of that expertise by making it cheap and more easily accessible to all. As a result, we have started to emphasize the importance of curiosity, creativity and critical thinking amongst our workforce rather than purely being focused on technical or coding skills. We have also started to emphasize the importance of individuals taking responsibility for their growth rather than waiting for the company to tell them what to do.
CIO&Leader: Looking ahead, what does your AI roadmap over the next 12–18 months look like — especially in terms of GenAI or foundation model deployments?
Prasad: Our AI roadmap is focused on enabling many more targeted use cases within our environment over the next 12 to 18 months. As agentic AI starts to gain traction, we anticipate that it will be increasingly possible to transform our business operations as well as enable our talent with subject matter expertise. However, we anticipate that we will need to continue to focus on delivering business value as a core pillar of how we justify the use of AI within our environment as well as for our clients.