Genpact launches Onyx platform: Process intelligence is last mile for AI agents

Genpact CEO Balkrishan "BK" Kalra argued that enterprises need to rethink their operating models to account for agentic AI where "agents run the work," "experts own the outcome," and process intelligence drives returns. Genpact also launched Genpact Onyx, an agentic operations platform.

Kalra laid out the company's focus on agentic operations and its own transformation at Genpact's Agentic Operations Day in New York. The company's vision revolves around delivering outcomes and technology that differentiates enterprises. To deliver those outcomes, process intelligence has to come first.

"Genpact is an agentic operations company. We are not the company you once knew. This is a new Genpact where frontier AI meets context-rich process intelligence. We are model agnostic, responsibly governed at enterprise scale," said Kalra.

Genpact's vision melds AI as well as process intelligence. Kalra said Genpact can provide custom AI agents and melding humans and AI agents. "The most differentiated part of this platform is process intelligence and decades of know-how," said Kalra. "There is no artificial intelligence without process intelligence.

Kalra also addressed the idea that customers can do everything on their own.

"Does this workload create competitive advantage in your end markets? There are many workloads. You must deploy your scarce builder talent to it, and we are enabling our clients to help them accelerate their journey through our custom agents and enabling them," said Kalra.

Genpact is positioning itself as a services provider that can deliver agents that address process inefficiencies and drive returns.

The company also released a report on agentic operations and autonomy. Findings from the report include:

• 79% of enterprises have AI agents in production, but 20% use them broadly and consistently across functions.
• 15% of enterprises have tied AI investment to profit and loss outcomes.
• 49% of AI initiatives funded in the past 24 months have delivered the value expected.
• AI leads or fully runs 40% of decisions.
• 9% of enterprises have completed significant operating model transformation.

"While many AI agents are helping individuals and teams to become more efficient and more effective in their work, the enterprises are not able to deploy these additional time or capacity towards their growth," said Genpact CTO Sanjeev Vohra. "The leading companies are reinventing the end-to-end process and deploying consequential agents across the entire process across functions."

Vohra said enterprises need to ask key questions about agentic operations. Questions include:

• Do you need a single agent or multiple agents?
• How do you build a flexible and secure AI architecture?
• What data systems and permissions should agents get?
• Who's accountable when things go wrong?
• What should you build or buy?

Genpact Onyx

To bring AI and process intelligence together, Genpact launched Genpact Onyx, the company's agentic operations platform.

Priya Vijayarajendran, Genpact's chief product and platform officer, outlined Onyx, which is designed to manage the process complexity with AI and combines human experts with automation.

"This is not a bunch of prompts and a static process map, which we hear a lot. It's built from actually running the work. That's the principle behind everything we do here," said Vijayarajendran.

She outlined the following capabilities of Onyx:

Process intelligence is built in.

• Intelligence compounds with every run, every exception and every expert decision and feeds back into a recursive learning loop.
• Onyx has security and governance with a guardrail harness.
• "AI controls are baked into the platform and integrated with data and AI foundations," said Vijayarajendran.
• Onyx powers Genpact's agentic operations offerings covering finance, supply chain, banking and more industries.

A demo of Onyx featured multiple use cases and disputes where an agent was able to recommend actions and deliver intelligence for future cases.

"Genpact Onyx delivers process intelligence with a learning loop, built-in governance, leveraging extensive partner technology. We are building our agentic solutions on it. We are building custom solutions with you and for you, and we are building with our partner ecosystem," said Vijayarajendran.

Underwriting outcomes

Vijay Vijayasankar, Global Agentic AI Officer at Genpact, said the company has moved from selling software and services to underwriting outcomes.

Vijayasankar said Genpact's agentic operations have focused on finance and accounting, supply chain management, banking and capital markets and insurance, but ultimately sees the CFO role evolving from historian to navigator.

"CFOs have been the best historians of the company. They are the stewards of the business. Everything that happened had to be recorded, explained, and kept track of, and that requires a lot of analysis, lot of back and forth with auditors, regulators, and so on," said Vijayasankar. "What we would like to do is to move the CFO from being the best historian to the navigator of the business, to look forward and drive the business."

• Genpact's AI chief Vijayasankar: Don't forget process, talent debt in AI projects

In this example from a CFO perspective, the outcomes in the middle are Genpact's responsibility so the CFO can be more strategic. Finance is the heart of the business.

Vijayasankar said Onyx is at the heart of the purpose-built offerings for finance and the processes as AI enters the picture. When the human and AI processes are honed, some "crazy good things" can be accomplished. "We are living in a time where the market is changing extremely fast, and looking at the rear view is not good enough to run the business," he said.

The customers

Genpact's conference featured multiple customers. Here's a look at the common themes and lessons learned.

• If you don't have your processes right, AI agents are just automating confusion. Multiple customers said process work was a prerequisite for scaling AI agents.
• One customer said the core outcome from using Genpact-built AI agents was converting more revenue to cash more quickly.
• Agentic AI operations are being used for everything from consolidating ERP systems to automating manual processes such as accounts payable.
• Build the data and process foundation before deploying AI. Otherwise, you're just deploying another technology.
• A customer noted that the accounts payable process went from no automation to 67% touchless with AI agents.
• Operating models need to change to account for agentic operations.
• Customers were balancing partnering vs. building their own AI agents.
• Data architecture is what defines AI success.
• Visibility into what AI agents and models are doing is critical. Humans are accountable for outcomes in the end.
• A human signs off on AI agent decisions with a full audit trail.