Deloitte has launched a new Open Model Engineering practice in India and other global markets, aimed at helping enterprises build, deploy and scale Agentic AI using a combination of open and proprietary AI models.
The initiative reflects a growing shift in enterprise AI, where organizations are looking beyond model performance alone and increasingly focusing on flexibility, cost, data sovereignty, governance and control as agentic AI deployments expand.
Open Models Take a Bigger Role in Enterprise AI
Deloitte’s new practice is designed to help organizations develop AI applications using open models and full-stack open-source technologies, tailored to their business requirements, regulatory environment and technology infrastructure.
The practice will initially support clients across North America, Europe and Asia Pacific, including India.
Deloitte says enterprises are increasingly evaluating open models because they can provide greater flexibility in how AI systems are deployed and managed.
Four Priorities for Scaling Agentic AI
According to Deloitte, enterprises are dealing with four major considerations as Agentic AI usage grows:
- Model flexibility: Selecting the right model and architecture for different workloads
- Cost predictability: Managing token and infrastructure costs as agent usage scales
- AI sovereignty: Greater control over where and how AI models are deployed
- Data and IP control: Protecting enterprise data, intellectual property and model behaviour
The Open Model Engineering practice is intended to help organizations address these requirements while choosing between proprietary models, open models, agentic platforms, cloud services and on-premises infrastructure.
NVIDIA Nemotron at the Core
Deloitte’s initial focus will be on enterprise AI applications powered by NVIDIA Nemotron open models and NIM microservices.
The practice will also support organizations in building Sovereign AI stacks, fine-tuning open models and using open technologies to strengthen cybersecurity capabilities.
Deloitte will also leverage its Zora AI digital workforce platform, which includes agentic harness capabilities built around NVIDIA Nemotron models and NIM. This is intended to give enterprises a way to deploy, continuously optimize and govern AI agents within their own environments.
Deloitte Plans to Invest in AI Engineering Talent
The company is also expanding its investment in engineering talent.
Deloitte plans to hire, train and certify forward-deployed engineers through fiscal year 2027. These engineers will work directly with clients to implement open-model AI solutions and support the growing demand for enterprise Agentic AI.
Why This Matters for Agentic AI
The announcement highlights an important evolution in enterprise AI.
As companies move from GenAI experiments toward autonomous agents operating across business processes, they need more than powerful models. They need architectures that provide control, flexibility, predictable economics and governance.
Deloitte’s approach suggests that the future enterprise AI stack may increasingly combine open and proprietary models, with organizations selecting models according to workload, cost, regulatory requirements and business context rather than relying on a single model provider.
The Bigger Picture
The rise of Open Model Engineering could accelerate the transition of Agentic AI from experimentation to production. For enterprises, the competitive advantage may increasingly come not just from which AI model they use, but from how effectively they engineer, integrate, govern and scale those models across their business.
Deloitte’s move also signals that Agentic AI is becoming an engineering and architecture challenge—not simply an AI model selection exercise.
