> Markdown version of [/videos/100130-ai-in-production-applied-ai-enterprise-use-cases](https://www.wearedevelopers.com/videos/100130-ai-in-production-applied-ai-enterprise-use-cases). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# AI in Production: applied AI & enterprise use cases

Scaling enterprise AI requires a shift from mere prompts to strict harness engineering. Learn to transform isolated models into secure, autonomous agents using deep observability.

- **Speakers:** [Mohak Chadha](https://www.wearedevelopers.com/@mohak-chadha)
- **Event:** World Congress 2026 Europe
- **Published:** July 9, 2026
- **Duration:** 29:42
- **URL:** https://www.wearedevelopers.com/videos/100130-ai-in-production-applied-ai-enterprise-use-cases

## Summary

Moving from isolated AI models to fully functioning enterprise AI systems requires a fundamental shift in engineering focus. Prompt and context engineering have evolved into "harness engineering," where the surround-system—responsible for executing tools, updating context, and managing state—is just as critical as the reasoning model itself. An agent is defined not as a separate model, but as a model repeatedly invoked inside a control loop to reason, act, and observe within strictly bounded limits. Model output degrades if the temporary working memory is populated with irrelevant data, underscoring why structural context management is vital. Moving these agents to production necessitates robust lifecycle coverage, from development tracing to active production telemetry. Even if an agent reaches the correct answer, the path it took could be expensive or unsafe, making deep observability an absolute requirement. The open-source NeMo Agent Toolkit serves as this operational layer, integrating seamlessly with orchestration frameworks like LangChain to govern guardrails, evaluations, and logging without replacing the core code. This foundational infrastructure enables advanced enterprise blueprints, such as NVIDIA IQ for deep research. Instead of basic one-to-one retrieval, IQ breaks down complex queries into delegated tasks across specialized sub-agents, a strategy utilized by ServiceNow and RBC to achieve highly autonomous ticket resolution and accelerated financial report generation. Because unstructured enterprise data frequently exists as video rather than text, extending these architectures requires specific multimodal capabilities. The Video Search and Summarization (VSS) blueprint demonstrates how integrating the Cosmos 3 vision-language model with Nemotron powers agentic video search, real-time safety alerts, and automated compliance reporting. Deployed on Kubernetes platforms with intent-based routing, these video-native agents successfully transform passive factory floor footage into a secure, directly actionable AI ecosystem.

**Keywords:** enterprise AI systems, harness engineering, context engineering, agentic control loops, nemo agent toolkit, production AI telemetry, AI observability, NVIDIA IQ blueprint, enterprise deep research, video search and summarization, multimodal video agents, cosmos 3 VLM, nemotron 3 nano, autonomous ticket resolution, intent-based routing

## Chapters

1. **Shifting focus from isolated models to enterprise AI systems** (00:03) — Building real-world AI systems requires moving beyond single prompt models to continuous agent loops.
1. **Managing state and dynamic context in agent workflows** (02:05) — Agents function as reasoning engines requiring dynamic assembly of working context at each step.
1. **Controlling model execution with dedicated runtime harnesses** (04:22) — A dedicated harness manages tool invocation, context updates, and failure recovery securely.
1. **Infrastructure requirements for deploying production enterprise agents** (05:50) — Moving to production requires connecting agents to secure infrastructure, observability tools, and continuous evaluation frameworks.
1. **Engineering lifecycle coverage with the NeMo agent toolkit** (07:55) — The open-source toolkit provides guardrails, telemetry, and debugging capabilities across multiple orchestration frameworks.
1. **Evaluating agent task execution and workflow performance** (09:47) — Comparing models and retrieval strategies requires automated metrics capturing latency, token usage, and accuracy.
1. **Royal Bank of Canada document processing capabilities deployment** (11:05) — Scaling financial data processing pipelines relies on orchestration engines to manage structured and unstructured data.
1. **Decomposing enterprise research tasks into multi-agent workflows** (12:44) — Complex research questions demand planner capabilities that delegate tasks to specialized sub-agents and preserve evidence.
1. **Blueprint architectures for scalable enterprise intent routing** (14:03) — Intent-based routers combined with retrieval models and secure execution sandboxes automate comprehensive research investigations.
1. **Autonomous agent resolution for enterprise service operations** (16:44) — Specialized AI agents triage tickets, parse logs, and implement solutions to drastically reduce engineering backlogs.
1. **Transitioning an agentic workflow to enterprise video intelligence** (18:52) — Integrating vision language models and embedding search enables continuous monitoring and alert generation for video streams.
1. **System design and orchestration for video AI agents** (21:52) — Deploying Cosmos 3 models on Kubernetes with elastic search infrastructure handles concurrent real-time intent routing.
1. **Logistics factory search via video analytics dashboard** (24:27) — Querying massive video data isolates exact timestamps of unrecorded events such as aisle obstructions using precise event chunking.
1. **Generating automated reports and reasoning about long videos** (26:16) — Asking direct questions to the video intelligence model generates multi-page safety reports from visual observations.
1. **Configuring specific real-time alerts for compliance monitoring** (28:26) — The intelligence layer identifies safety protocol violations instantly as they occur on a live application camera feed.

## Related Moments

- [Introduction to building real-world AI agent solutions](https://www.wearedevelopers.com/videos/1538-composable-intelligence-how-henkel-and-microsoft-are-shaping-the-agent-ecosystem) (from "Composable Intelligence: How Henkel and Microsoft Are Shaping the Agent Ecosystem")
- [Designing agentic AI solutions for the enterprise](https://www.wearedevelopers.com/videos/1831-ai-for-enterprise-developers-dr-damir-dobric) (from "AI for Enterprise Developers - Dr. Damir Dobric")
- [Navigating the components of the modern generative AI stack](https://www.wearedevelopers.com/videos/1512-building-ai-applications-with-langchain-and-node-js) (from "Building AI Applications with LangChain and Node.js")
- [Deploying AI agents for enterprise legacy code modernization](https://www.wearedevelopers.com/videos/100256-can-this-elephant-dance-ibm-bob-and-the-future-of-ai-first-software-development) (from "Can This Elephant Dance? IBM Bob and the Future of AI-First Software Development")
- [Introduction to building reliable AI agents in production](https://www.wearedevelopers.com/videos/1523-the-ai-agent-path-to-prod-building-for-reliability) (from "The AI Agent Path to Prod: Building for Reliability")
- [Scaling generative AI use cases across large enterprises](https://www.wearedevelopers.com/videos/916-beyond-the-hype-real-world-ai-strategies-panel) (from "Beyond the Hype: Real-World AI Strategies Panel")

## Related Articles

- [Stephan Gillich - Bringing AI Everywhere](https://www.wearedevelopers.com/magazine/489-stephan-gillich-bringing-ai-everywhere)
- [Graph and AI Trends 2026: Why Is AI Running but Not Yet Delivering?](https://www.wearedevelopers.com/magazine/680-graph-and-ai-trends-2026-why-is-ai-running-but-not-yet-delivering)
- [MLOps – What’s the deal behind it?](https://www.wearedevelopers.com/magazine/125-mlops-what-s-the-deal-behind-it)
- [From Prototype to Production: Build AI Agents with This Free 4-Course Learning Path](https://www.wearedevelopers.com/magazine/655-from-prototype-to-production-build-ai-agents-with-this-free-4-course-learning-path)

## Related Jobs

- [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat**
- [Senior AI Agent Software Engineer (Go, Python) (m/f/x)](https://www.wearedevelopers.com/jobs/48277-senior-ai-agent-software-engineer-go-python-m-f-x) at **Dynatrace**
- [Staff Software Engineer, Copilot Experiences](https://www.wearedevelopers.com/jobs/ext/164361-staff-software-engineer-copilot-experiences) at **GitHub**
- [AI Full Stack Engineer](https://www.wearedevelopers.com/jobs/ext/1354435-ai-full-stack-engineer) at **Almedia**
- [AI Operations Manager (all genders)](https://www.wearedevelopers.com/jobs/48263-ai-operations-manager-all-genders) at **envelio**
- [Principal Field Architect - AI Agents](https://www.wearedevelopers.com/jobs/ext/1442858-principal-field-architect-ai-agents) at **Twilio**