Three years ago, a keynote on AI might have revolved around possibility, what AI could become, what might be achieved. But as Sanjeev Vohra, Chief Technology and Innovation Officer at Genpact, stated at MachineCon 2025, the focus has now shifted. “I am not going to talk about the potential of AI,” he noted. “We have done that already. Today is about the realism of AI, what it actually does in the enterprise, right now.”
Agentic AI is now recognized not just as a concept, but as an operational system. These systems are defined by autonomy, where decisions and actions are taken in alignment with business goals, with minimal human involvement. Such systems are being built to perceive, reason, and act within real-world enterprise environments. Rather than simply assisting, they are increasingly being designed to execute entire workflows independently.
The path to this point has been long. AI in the 1990s relied on expert systems and rule-based logic that delivered fixed outcomes. This was followed by the predictive era, where statistical algorithms and machine learning were deployed to analyze patterns and forecast future trends. As neural networks and deep learning matured, the field expanded into image recognition, language understanding, and generative content. The turning point came in 2022, when generative AI models demonstrated the ability to create text, images, and code at scale. That phase, defined by creativity, is now giving way to one focused on autonomy and action.
Agentic AI marks this next phase. Enterprises are no longer evaluating what AI might be able to do in the future. Real deployments are underway, and use cases are being scaled. Refunds are being processed, invoices are being verified, and supplier interactions are being handled without direct human oversight. These actions are being performed by systems that reason through business logic, apply thresholds, and act within defined parameters.
Within Genpact, this transition has already been operationalized. Its Client Zero initiative challenges Genpact to adopt the same cutting-edge solutions it brings to its clients. So far, over one hundred AI agents have been internally deployed across business functions such as finance, HR, legal, marketing, and sales. As a result, productivity improvements of more than ten percent have already been achieved so far.
One agent, Amber, functions as a Chief Listening Officer. Through more than 500 million conversational touchpoints, employee sentiment is continuously assessed and translated into personalized HR actions. This level of engagement has contributed to an 85 percent positive sentiment score across the organization. What was once managed through annual surveys is now being maintained in real time.
In finance operations, a family of agents has been structured to handle the full scope of accounts payable. Genpact AP Capture extracts structured data from transactional records. Genpact AP Advance processes invoices and applies matching logic. Anomaly detection and leakage identification are managed by Genpact AP Trace, while supplier communication is facilitated by Genpact AP Assist. These agents operate collectively, enabling activities such as three-way invoice matching to be completed in seconds. Manual delays are being slashed. And suppliers have reported improved satisfaction due to faster payment cycles.
All of these systems operate under a unified enterprise architecture. A connector hub gathers data from various systems of record. A data engine standardizes and structures the information, making it accessible for downstream AI agents. The AI agent foundry functions as the design, deployment, and governance layer. This architecture is being built and managed within a responsible AI framework, ensuring that transparency, safety, and observability are maintained.
Execution at scale has required four foundational shifts at Genpact. First, the initiative is being led from the top. CEO sponsorship has ensured that AI is positioned as a business transformation lever rather than a back-office experiment. Second, the selection and prioritization of use cases is driven by business teams, supported by value architects who size outcomes and map implementation roadmaps. This alignment has been critical in ensuring that technology delivery remains tied to measurable impact.
Third, a dedicated AI factory has been established to build, train, and refine agents. It has been recognized that AI systems cannot be developed like traditional software. Unlike static programs, AI systems must be trained continuously and adjusted to evolving business requirements. This has introduced the need for multidisciplinary teams like domain experts, technology experts, and integration architects who help embed AI agents within existing enterprise systems and workflows.
Finally, enterprise-wide adoption has required a coordinated approach to workforce enablement. Over 90,000 employees have already been upskilled on AI. Training programs are being aligned to specific roles, ensuring that individuals are equipped to both use and co-create with AI systems. Adoption is being driven at every level, with leaders and operational teams integrating AI into their daily activities.
This transformation has been enabled by a set of converging forces. AI is now a priority in boardrooms, with leadership discussions focused on deployment strategy and return on investment. Technological maturity has accelerated. Foundation models are advancing rapidly, infrastructure costs have decreased, and on-device AI has reached enterprise-grade capability.
Data strategies have evolved. Organizations are moving from federated lakes to integrated platforms where data is treated as a product available for both human and machine consumption. Interoperability across the ecosystem is being fostered. Platform providers, startups, academic institutions, and consulting firms are working collaboratively to create domain-specific, composable solutions. Investment has surged. Over 60 percent of global venture funding in the first half of 2025 has flowed into AI. Regulatory frameworks are also being established. National governments are investing in sovereign AI strategies that promote innovation while introducing safeguards for ethical deployment.
Together, these trends are creating a unique window of opportunity. However, enterprise readiness remains uneven. Fewer than five percent of organizations have developed roadmaps for full-scale AI deployment. Many remain limited to pilot programs. The risks of waiting are growing. In recent forums, it has been observed that planning cycles must give way to execution. Feedback loops must be established, and experimentation must be treated as a continuous operating model. The contextual nature of AI demands that systems be iterated in real business environments.
Agentic AI is no longer a matter of what is possible. It is a question of who is moving fast enough to capitalize on what is already real. Enterprises that act now, learn continuously, and scale responsibly will be positioned to lead in a future that is fast-approaching.
Sanjeev believes that the time for hesitation has passed. As he puts it, “We cannot sit on the sidelines. We cannot waste time planning. We have to execute. We have to act immediately, and as we learn, we need a continuous feedback loop for taking actions and learning from those actions.”
