New Delhi, September 4: Deloitte has launched a global Open Model Engineering practice aimed at helping enterprises develop, customize and deploy open artificial intelligence models for business applications, reflecting growing interest among organizations in having greater control over their AI systems.
The new practice will initially focus on markets across North America, Europe and Asia Pacific. Deloitte said its approach will cover open-model engineering, model fine-tuning, AI security, open-source frameworks and sovereign AI, as enterprises increasingly move AI applications from experimentation into operational environments.
The initiative comes as businesses evaluate alternatives to relying entirely on closed, proprietary AI models. Open models can be adapted and fine-tuned for specific requirements and can potentially be deployed across different infrastructure environments, giving organizations greater flexibility over how AI systems are developed and operated.
Enterprise AI workloads can have significantly different requirements. Customer-service applications may prioritize speed and efficiency, while financial or government systems can require stronger data controls and specific deployment arrangements. Manufacturing applications may also require AI systems to operate closer to industrial infrastructure. An open-model approach can allow organizations to select and customize models according to these requirements.
Deloitte’s initial approach includes NVIDIA’s Nemotron open models and NIM microservices. The company is also emphasizing sovereign AI capabilities, which are intended to give organizations greater control over where AI systems and associated data are processed. Such considerations can be particularly relevant to governments and highly regulated industries where data location, security and compliance requirements play an important role.
Cost is another consideration as enterprises expand AI deployments. AI systems that process large numbers of requests can generate significant inference and infrastructure costs. Organizations may therefore evaluate models not only on their capabilities but also on factors including performance, computing requirements and operating costs. Open models can provide additional options for balancing these factors through customization and deployment choices.
The practice also reflects a broader shift in enterprise AI requirements. Organizations increasingly need to integrate AI models with existing data, software applications, infrastructure, security systems and governance processes. Deloitte said it plans to hire, train and certify Forward Deployed Engineers as part of the initiative, highlighting the growing demand for professionals capable of implementing AI systems within business environments.
The company’s strategy also includes agentic AI, in which AI systems can perform multi-step tasks, use digital tools and interact with business systems with greater autonomy than conventional chatbot applications. Deloitte’s Zora AI platform incorporates agentic capabilities built around NVIDIA technologies.
The expansion of agent-based systems could further increase demand for flexible AI architectures. Different business agents may require different combinations of reasoning capability, speed, cost, security and specialization. Organizations could therefore use multiple models for different workloads rather than relying on a single model across all applications.
However, adopting open models also creates additional responsibilities for enterprises. Organizations must address issues including security, model evaluation, governance, reliability, intellectual property, bias, infrastructure requirements and ongoing monitoring. Greater control over model customization and deployment also requires technical expertise to manage the resulting systems.
Deloitte’s move comes amid a broader evolution in enterprise AI strategies. Businesses are increasingly assessing AI architectures based not only on model performance but also on cost, security, flexibility, data control and deployment requirements.
The development suggests that enterprise AI adoption is moving beyond simply accessing individual models through applications or APIs. Increasingly, organizations are focusing on how different models, data sources, infrastructure and AI agents can be integrated into reliable systems capable of operating at scale.
