For the past couple of years, most enterprise AI conversations have orbited the same question: how do we build AI agents? Organizations have been experimenting with copilots, assistants, and task-specific automation across finance, procurement, operations, and customer service. Progress has been real, but a more important question is starting to surface.
How do you manage an ecosystem of AI agents at scale?
Databricks addressed this head-on at Data + AI Summit 2026 with the announcement of Omnigent, an open-source “meta-harness” released under Apache 2.0. It is a meaningful signal about where enterprise AI is heading, and it is worth unpacking what it does and why it matters.
What Omnigent Is (and What It Is Not)
Omnigent is not a new AI model or another agent framework to add to the pile. It sits above the agents and harnesses your teams are already using, whether that is Claude Code, OpenAI Codex, Inflection Pi, or something you built in-house, and provides a single coordination layer on top of all of them.
Databricks built Omnigent because they ran into this problem themselves. With more than 5,000 engineers running multiple agents simultaneously, they found their teams spending significant time copying outputs between tools, switching between environments, and losing track of what each agent was doing or spending. Omnigent is their answer to that friction. It focuses on three things: composition (combining multiple agents without rewriting code), control (enforcing cost budgets, access permissions, and security policies at the orchestration layer, not inside a prompt), and collaboration (letting teammates share live agent sessions and review work together in real time).
The governance piece is architecturally significant. When policies live inside a prompt, a model can reason around them. When they are enforced at the meta-harness layer, the model never sees them. That distinction becomes important when you are operating agents across sensitive business data.
Omnigent is currently in early alpha. It is worth watching and evaluating, but it is not a production deployment today.
The Problem We Have Seen Coming
Omnigent matters to us at Syngentic because it points to a challenge we have been talking with customers about for a while.
The technology industry tends to focus heavily on creation and underestimate operationalization. We have watched this play out before, when organizations invested in analytics before establishing data governance, the result was fragmented solutions, limited adoption, and business impact that never matched the initial promise.
AI agents are following the same trajectory. Building one agent is not particularly difficult. Managing dozens of them across finance, supply chain, compliance, operations, and customer service is an entirely different challenge. Which agents have access to which data? How are decisions traced and audited? How do you prevent runaway costs? How do you ensure agents are working from consistent, accurate business context rather than each developing their own interpretation of what the data means?
These are not model problems. They are architecture and governance problems, and they do not get solved by adding more agents.
Why the Data Foundation Determines Everything
Here is what tends to get overlooked in conversations about agentic AI. The quality of what an agent can do is almost entirely determined by the quality of the information it operates on.
Most organizations already hold an enormous amount of valuable business knowledge. It lives in SAP systems, OpenText repositories, engineering documents, contracts, maintenance records, project files, and operational data of every kind. The challenge has never been a lack of information. The challenge is making that information trustworthy and consistent enough for AI systems to use it reliably.
This is exactly why the rise of data products across SAP Business Data Cloud and Databricks is so important right now. A data product is not just data. It comes with business definitions, clear ownership, governance controls, quality standards, and documented context that makes it reusable across systems and teams. For an AI agent, that distinction is everything. Without governed data products, every agent risks operating on a different version of the truth. With them, you get a shared foundation that multiple agents can consume consistently, and you get alignment between your analytics, your operations, and your AI decision-making.
A tool like Omnigent can govern how agents interact, but it can only govern the quality of decisions agents make if the underlying data is trustworthy in the first place. These two layers reinforce each other, and you need both.
What the Future Enterprise Looks Like
The next generation of enterprise AI will not be a single assistant doing everything. It will be a coordinated network of specialized agents, each performing a distinct role.
Think about what that looks like in practice for a manufacturing or process-intensive business. A maintenance agent monitors asset performance and flags anomalies. A procurement agent identifies sourcing risks and surfaces alternatives. A compliance agent tracks regulatory requirements across jurisdictions. A finance agent analyzes cost variances and models scenarios. A knowledge agent surfaces relevant engineering documentation or institutional expertise at the moment someone needs it.
Each of those agents is doing meaningful work. The challenge is making them work together without creating new chaos. They need shared context. They need consistent access to the same trusted data. And they need governance that prevents any single agent from going rogue on cost, security, or data access.
What This Means If You Are Investing in SAP and Databricks
For organizations that are building on SAP Business Data Cloud, Databricks, and SAP Joule, the direction here is clear. Competitive advantage will not come from having more agents. It will come from having well-governed agents operating on trusted, consistent business context.
That requires modern data architecture; It requires governed data products; It requires strong integration between structured and unstructured data across your enterprise, and it requires thinking about orchestration and governance as foundational capabilities, not afterthoughts you address once the agents are already running.
Organizations that establish these foundations now will be able to scale AI far more effectively than those that prioritize agent deployment ahead of data readiness. We have seen this pattern play out in analytics and in earlier AI investments. The organizations that got the foundation right early captured the advantage. The others spent years retrofitting governance onto systems that were never designed for it.
The Signal Worth Paying Attention To
Omnigent is one announcement, and it is early. The direction it signals reflects something we are seeing across the broader market. Agentic AI is moving from capability demonstration into a governance maturation phase. The early conversations were about what agents can do. The conversation that matters now is about how to deploy them accountably, sustainably, and at enterprise scale.
At Syngentic, this is exactly where we believe organizations should focus their attention. The future of enterprise AI will not be defined by which company has the most agents or the most sophisticated models. It will be defined by the quality of the information, architecture, and governance that those agents operate on.
The question is no longer whether your organization will deploy AI agents. The question is whether those agents will be working from a foundation you can trust.
