As AI agents move from experimentation to wider deployment across enterprise workflows, companies are increasingly focusing on whether the technology can deliver measurable returns, rather than simply demonstrating its ability to automate tasks.
The shift comes as enterprises that initially ran isolated AI pilots and experiments seek to embed AI more deeply into core business processes such as customer relationship management (CRM), enterprise resource planning (ERP) and other workflows. While pilots have delivered incremental benefits, their impact on overall business performance has often been limited, Shishank Gupta, SVP & head of the digital workplace ecosystem and Microsoft practice at Infosys, told FE.
“The current phase of AI adoption is focused on embedding intelligence directly into CRM and ERP processes to drive end-to-end outcomes,” Gupta said. The move is aimed at bringing AI into the natural flow of work, reducing context switching and improving adoption, rather than requiring employees to use AI as a separate tool.
However, the wider deployment of AI agents is creating a new set of challenges for enterprises, particularly around cost, governance and measuring value. With agents increasingly being used to perform tasks across business processes, organisations are beginning to treat them more like digital workers than standalone software tools, Gupta said.
This is prompting companies to establish controls covering agent onboarding, lifecycle management, security, compliance and monitoring, while also dealing with agent sprawl and rising AI token consumption. “Value realisation has become a critical part of the puzzle now,” Gupta said.
Enterprises are therefore looking to identify high-impact use cases before scaling them, while tracking metrics such as productivity gains, reduction in cycle times, higher throughput, automation rates and improvements in customer experience. The objective is to establish whether the investment in AI is translating into measurable business outcomes.
The Microsoft ecosystem, which includes Dynamics 365, Microsoft 365 Copilot and Power Platform, provides enterprises with tools to embed AI into existing workflows and build and orchestrate agents. Dynamics 365 Copilot, for instance, brings AI capabilities into CRM and ERP processes, while Power Platform tools such as Copilot Studio and Power Automate enable organisations to build and automate agents across business processes.
But as the number of agents deployed across an organisation increases, controlling what these digital workers can access and do is becoming equally important. Infosys combines Microsoft Purview with its own responsible AI and governance systems to manage data access, classification, data loss prevention, connector governance, oversharing and auditability, Gupta said.
The company is also tracking its own AI-led productivity through metrics including effort reduction, person-days released, cycle-time compression, throughput gains, quality improvements, automation rates and financial performance. These measures are linked to enterprise value levers such as productivity, efficiency, growth and cost optimisation, he said.
The extent of productivity gains, however, varies across functions and enterprises depending on factors such as industry and organisational maturity. This suggests that the next phase of enterprise AI adoption will be less about the number of agents deployed and more about how effectively companies integrate them into workflows while keeping their costs and business outcomes under control.