With AI rapidly reshaping the workplace, adaptation has become crucial, according to Jishnu Chatterji, Head of Data, Analytics & AI (UKI) at Cognizant.
Speaking to an audience of senior civil servants and public sector leaders at a workshop entitled “Safety and Scalability: Building an Agentic AI Future in Government”, Chatterji outlined how advances in large language models and agentic AI are allowing for the automation of increasingly complex workflows.
According to research carried out by Cognizant in collaboration with Oxford Economics, the extent to which the tasks involved in jobs in OECD economies could theoretically be performed by AI has increased significantly over the last three years. As changes continue to occur at pace, impacts on job markets which were not expected before 2030 are already being seen.
In this rapidly developing context, Chatterji outlined how he believed AI could help bodies, such as public sector organisations, streamline their own workflows and improve productivity.
Firstly, he emphasised AI’s ability to serve as a “hyper-productive agent”, helping to simplify tasks such as contracting, invoices, and admin on an individual level. Beyond this, Chatterji highlighted the possibility of automating entire business processes through networks of AI agents. He cited work to agentify the Joiner, Mover, Leaver (JML) process - a procedure which could previously take an average of 60 to 75 days.
More ambitiously, he envisages AI fundamentally reshaping enterprise software, potentially replacing some traditional platforms with AI-driven alternatives that could reduce operating costs. Chatterji suggested that, with appropriate governance and controls in place, agentic AI could take on increasingly complex business processes. He pointed to early examples of this approach being explored in areas such as human resources, marketing operations and regulatory submission processes.
Despite the rapid pace of technical progress, Chatterji warned that public trust is a significant challenge to attempts to fully agentify certain services, with the feedback from those running customer service functions being that “customers don’t want to talk to chatbots”, especially on emotionally charged subjects.
He also cautioned against overestimating current AI capabilities, particularly through claims that software engineering skills will become obsolete. While AI can accelerate the development of prototypes, he stressed that significant work is still required to build scalable, production-ready systems.
Chatterji noted that human oversight remains an important part of the software development process, with engineering teams continuing to review and validate AI-generated outputs before deployment. While coding assistants can improve productivity, he said human expertise remains essential to ensure systems are reliable and suitable for use in critical public services.
Turning to work underway within Cognizant’s AI Lab, Chatterji highlighted the Neuro AI Multi Agent Orchestrator and the Agentic AI Harness Platform. He described these as tools designed to help organisations develop and manage multi-agent AI systems, while maintaining appropriate levels of accuracy, governance and trust.