The race to redefine enterprise operations with generative AI is nicely underway. Whereas many firms are experimenting with automation and clever instruments, Genpact is taking a extra structured and impressive path.
The corporate’s push into generative AI and automation is now converging into what it calls the agentic enterprise, the place AI doesn’t simply assist the method, it turns into the method.
Whereas talking at AIM’s DES 2025 occasion, Aniruddha Ray, senior vp of agentic merchandise and platform engineering at Genpact, described this as “essentially the most highly effective assemble in my lifetime”.
He outlined Genpact’s transformation beneath the ‘Genpact Subsequent Technique’. This technique focuses on three pillars—main with superior know-how, prioritising service traces and unlocking worth for patrons.
From Agent-Powered to Agentic Enterprises
Ray highlighted the distinction between agent-powered enterprises and agentic enterprises.
Relating to agent-powered enterprises, he defined, AI is embedded inside current processes. In agentic enterprises, nevertheless, AI turns into the method. “In the present day, if I run a course of and I embed AI or software program in it, what is going to occur is that the method shall be pushed by AI, and people will really allow a part of it,” he stated.
On this shift, information is not static and siloed. Genpact is transferring in the direction of unified, contextual, real-time information, supported by information graphs and semantic ontologies. Advanced software interfaces are being changed with agent-driven interactions.
From human within the loop to human on the loop, the connection between AI and other people is steadily evolving. It means that people are transferring from energetic contributors to strategic screens of AI methods.
“[While] brokers are doing quite a lot of work, people are monitoring, augmenting and governing,” Ray defined.
Nevertheless, to totally transition into an agentic state, one thing was nonetheless lacking. Genpact had experimented with methods of insights and methods of integration, however neither crammed the hole.
Ray stated that it seems the lacking hyperlink was “methods of innovation”—an orchestrated layer the place brokers work together amongst themselves, with methods, and with people. “That’s the agentic layer, and that’s the layer of innovation,” he added.
Empowering Information Employees
One of many first steps, in response to Ray, has been to “supercharge information employees”. He pointed to instruments like Dataverse, which may generate industry-specific fashions shortly with minimal information, and Cora Code GenY, a software program growth mannequin to generate code and take a look at instances.
To additional push this, Genpact launched a mannequin referred to as the AI GigaFactory. It permits for large-scale automation in opposition to particular use instances, bringing down undertaking timelines considerably. “A change that must be executed in six months…we will attempt to do it in two to 4 months,” Ray stated.
The third step is delivering these options as agentic choices. Moderately than measuring success by the share of automation, Genpact focuses on tangible enterprise worth, like decreasing bill processing value by 70% or bettering claims dealing with pace.
Service as Agentic Software program
Ray described two main working shifts inside Genpact. The primary is the transfer to service-as-agentic-solutions, what he termed “SaaS 2”—the place the service itself is pushed by the software program.
The second is the AI GigaFactory mannequin, which mixes agent-led automation with productised options. Ray introduced a product for accounts payable for instance.
The structure connects methods of report, methods of perception, and methods of engagement. On the core lies a layer of specialist and generalist brokers working collectively via orchestration. “That is the agentic AI engine,” he stated.
Nevertheless, success relies on extra than simply brokers. In keeping with Ray, there are three major challenges—information high quality, integration, and agent choice. “We’re realising we have to create an ensemble of agent…decided by the kind of information it’s making use of to,” he stated. Area experience and semantic understanding are essential.
Constructing In the direction of Agentic Merchandise
Genpact is making use of this considering throughout finance and danger, from accounts payable to invoice-to-cash and record-to-report. Some are productised options for a handful of consumers.
To scale, the corporate is investing in a strong functionality mannequin. This begins with the information engine room, the place anonymised or artificial information is ready and used to coach various kinds of brokers.
A stay semantic metadata layer sits above it. The AI agent foundry is the place a number of brokers are created, every fine-tuned with domain-specific LLMs and SLMs.
“You don’t do something with out working with companions,” Ray stated. Whether or not via partnerships, agent marketplaces, or proprietary instruments, the objective stays clear—to construct and scale agentic AI methods that drive measurable worth for purchasers.
He identified that the corporate has developed G Options, a market of Genpact’s AI property. It at the moment runs over 1,000 energetic AI fashions deployed throughout inside platforms, buyer merchandise and enterprise options.
Ray additional added that whereas not all fashions are agentic, Genpact has already developed 21 executable brokers and plans to scale this quantity to 50 by the tip of the 12 months.“Current AI merchandise have pushed as much as 40% productiveness features, with a goal of 60–70% over the subsequent 12–24 months,” he concluded.
The submit Genpact Strikes Past Automation to Construct Agentic Enterprises appeared first on Analytics India Journal.
