Artificial intelligence has moved beyond the experimental phase. CIOs and technology leaders are now making strategic decisions about deploying AI in real-world, production environments where operational constraints and real consequences are the norm. As organizations shift from AI pilots to full-scale adoption, the spotlight is turning from the models themselves to the infrastructure that supports them.
In a recent episode of the Analytics Insight Podcast, host Priya Diyalani spoke with Shrish Anand Lal, Executive Director and Chief Business Officer at NewStreetTech, about why enterprises need to focus on AI architecture and infrastructure rather than relying solely on AI models. The conversation covered the limitations of model-first strategies, core building blocks of enterprise AI infrastructure, cloud and hybrid deployment evolution, AI ROI, autonomous agents, platform strategy, and the importance of deterministic execution and governance. Below are the key takeaways.
Why AI Models Alone Are Not Enough for Enterprise Adoption
According to Lal, the AI model represents only about 20% of the challenge when it comes to enterprise adoption. The remaining 80% involves everything else — what NewStreetTech calls the FITS framework: Fear, Inertia, Trust, and Surprise. None of these issues can be solved by simply adopting a better model.
Referencing research from NASSCOM, Lal noted that 65% of companies in India have launched AI pilots, but only 15% have reached production. He emphasized that models must be applied at the design phase, while execution must remain deterministic. “The AI model is becoming a commodity very quickly, but the AI architecture will be a competitive advantage,” he said.
Core Building Blocks of Enterprise AI Infrastructure
Lal outlined four essential building blocks:
- Separation — AI operates at the design and configuration stage, while execution remains deterministic. The two layers must not interact.
- Governance by design — Maker-checker workflows, audit trails, version control, and role-based access should be foundational, not afterthoughts.
- Orchestration — Using a single AI platform for everything is the wrong approach. NewStreetTech’s MyFix.ai, for instance, uses over 36 configuration agents and 14 execution engines, all deterministic.
- Deployment flexibility — Organizations must have the option to deploy on-premises, in the cloud, or via a hybrid model, especially for regulated industries.
Cloud and Modern Architectures Evolving for Enterprise AI
AI computing costs are dropping rapidly, which is good for adoption. But the real revolution, Lal argued, is not just about computing — it’s about architecture. The shift is from “AI as a service” to “AI as an infrastructure.” Organizations need resilience through multiple providers to avoid lock-in. Some workloads must remain on-premise; others can go to the cloud. This architecture transformation is critical for regulated industries, where leaving everything to AI creates a black box with unanswered questions.
How the Conversation Around AI ROI Has Changed
Two years ago, boardrooms asked whether they should invest in AI. Today, the question is why AI hasn’t yet gone into production. The ROI debate has shifted from the cost of implementing AI to the cost of not implementing AI. Enterprises now worry about falling behind competitors. Companies that are seeing real ROI started small.
AI Agents and Autonomous Workflows: Reshaping Enterprise Infrastructure
Lal clarified that AI agents are not chatbots. They are specialized workers handling specific tasks. At MyFix.ai, there are more than 37 specialized configuration agents and 14 deterministic execution engines. Agents build the system, humans approve it, and engines run it. This is the fundamental difference between building enterprise-grade AI infrastructure and simply adding more computing power.
“The future will not be one AI system doing everything. It will be specialized agents working through orchestration, trust levels, cross-validation, and human oversight,” Lal concluded.
Listen to the full discussion on the Analytics Insight Podcast.
Leave a Reply