Jito Chadha, Founder and CEO of Nventr, is an experienced technology executive and investor whose career spans artificial intelligence, data science, venture capital, and business development. Since founding Nventr in 2019, he has led the company’s development of enterprise AI technologies, while also serving as CEO of predictive analytics company Rule14. Chadha has been part of HandsOn Global Management since 2007 and became a partner in 2013, contributing to the sourcing, structuring, and value creation of numerous transactions while leading investments across venture capital, data science, and natural resources.
Nventr is a Los Angeles-based enterprise AI company that develops technology for building, deploying, and managing intelligent applications and automated workflows. Its platform brings together AI agents, machine learning models, neural networks, data management, intelligent document processing, visualization, and development tools. Products include Agent IO, an operating system for orchestrating autonomous agents across business functions, and nQube, a workflow builder that supports no-code, low-code, and full-code development, along with deployment across cloud, on-premises, and edge environments.
Before leading Nventr and Rule14, you spent years sourcing investments, creating value across portfolio companies, and working with data science initiatives inside complex organizations. How did that experience shape your decision to build an enterprise artificial intelligence platform focused on workflows rather than standalone tools?
Working across portfolio companies exposed the same pattern in different businesses. Every organization had information moving through several people, systems and approvals before a process could be completed.
That experience shaped Nventr around workflows rather than standalone tools. Improving one task has limited value if the result cannot move efficiently through the rest of the organization. We focused on structuring the entire process and running it at scale.
Nventr combines agentic actions, machine learning models, and neural networks within scalable enterprise workflows. What limitations did you observe in conventional automation platforms that led you to develop this broader architecture?
For years, companies operated through collections of cloud and locally hosted SaaS applications that could be clunky, rigid and expensive. The workflow system, models, agents and company data were usually treated as separate layers, which made broader automation more difficult.
We developed an architecture where those components can operate together and scale independently as demand changes. The process does not have to be forced into one fixed application or one type of computing environment.
Many organizations begin their AI strategies by selecting a model or launching a chatbot. In your experience, what foundational work involving data, permissions, processes, and system integration needs to happen before an AI deployment can deliver meaningful results?
A company first has to define the process, identify what information the agent needs, and determine where that information lives. That may involve databases, file repositories, PDFs, articles, email systems, or other company data sources that have to be made available to the agent in a usable form.
The company then has to define what happens after the agent receives that information. A conversation might produce a response, a summary, a follow-up queue, or another agentic workflow. Approved tools, data permissions, and user permissions have to be established by role so each action occurs inside a defined process.
nQube is designed to help both data scientists and business analysts build, test, deploy, and monitor AI-powered workflows. How do you make these capabilities accessible to non-specialists while retaining the technical controls required for production environments?
People think about AI and software products through the scope of their own role and how the product can make their job easier or allow them to be more productive. The interface should let them describe what they need, ask questions and adjust an agent or workflow through normal language and voice.
The workflows, computing processes and scaling remain underneath that interaction. The organization still controls the tools, data sources and access assigned to each role, so a business analyst can use the system without having to manage the infrastructure directly.
Nventr’s agent operating system uses a coordinating agent to direct specialized agents through sequential and iterative workflows. How does the platform determine which agent should handle each task, preserve context between agents, and handle an unreliable result?
The workflow determines how a request moves through defined steps, branches and specialized agents. Information from one stage can be summarized and passed forward so the next agent understands what happened, what was requested and what follow-up should occur.
Handling an unreliable result depends on the process and the consequences of the action. The workflow controls have to reflect where the system can continue and where the result requires review.
Enterprises increasingly want the flexibility to use different large language models, specialized machine learning systems, and third-party agents. How should organizations design their AI infrastructure to avoid becoming dependent on one model provider or technology stack?
The infrastructure should support a multitude of models and allow them to operate in parallel within the same workflow. One provider might handle the speech-to-speech component, while another model performs reasoning or processing behind it.
The toolkit should be chosen according to the customer’s cost, performance, security, latency, and openness to third-party service providers. Some enterprises are comfortable combining providers, while others want the agent on their own cloud or dedicated infrastructure. The architecture has to support both approaches.
Once an AI agent can access company data and take actions across business systems, security becomes more complex than protecting a conversational interface. What controls should govern an agent’s permissions, autonomy, and ability to initiate consequential actions?
Access should be controlled by role, project and workflow. An agent should only be able to use the information, tools and actions approved for that context, with requests outside that scope sent for authorization.
The amount of autonomy should reflect the sensitivity of the data and the consequences of the action. For transactions involving money, private information or another significant outcome, the company should be able to see what the agent is doing and require approval before completion.
Nventr presents applications across areas including document processing, customer service, reporting, healthcare communication, and fraud detection. What technical or operational characteristics make a workflow suitable for agentic automation, and which processes should remain primarily human-led?
We balance the number of people or cost associated with a process against its complexity and viability for automation. A process can be expensive or labor intensive and still be the wrong place to begin if it is too complex or unreliable to automate effectively.
The general approach is to go for the lowest-hanging fruit with the biggest impact. That formula is fractal because it can be applied to one person’s role, several departments or an entire organization. Processes involving uncertainty or consequential judgment should continue to include people.
You have discussed the idea of employees eventually having AI counterparts that handle repetitive elements of their work. How might this change organizational structures, and what should leaders do to ensure these systems strengthen human judgment rather than gradually removing it from important decisions?
The idea is to give each person on the payroll an agent corollary that can learn parts of the job over time. Across a larger percentage of the organization, that creates centralized intelligence and optimization over the org chart rather than limiting the benefit to one or two employees.
The agent can take on repetitive work such as collecting information, moving data or building routine presentations. That gives people more time for important discussions and decisions, while the organization keeps authority over sensitive information and consequential actions.
As AI agents evolve from assistants into an operating layer for complex organizations, which capabilities must improve most before companies can safely automate workflows that span entire departments or business units?
Agents need lower latency and stronger integrations with company databases, files and larger systems. Information has to move across roles and departments without requiring people to transfer it manually or creating new bottlenecks.
The interaction layer also has to improve. Voice agents should feel like a normal conversation where a person can interrupt and change direction, and businesses will eventually need direct ways for authorized agents to interact with their systems. Browser automation can bridge the gap, but the industry will move toward a more efficient structure because compute and infrastructure have a cost.
Thank you for the great interview, readers who wish to learn more should visit Nventr.
