By Chris Vessey, VP of Innovation & Customer Value
We have a global customer who deployed a single agentic workflow that alone processes 15,000 transactions per month, drives a 60% reduction in human effort, and now achieves a 93% first-time compliance rate. Jaw-dropping, and just the start of what’s possible.
If the “there’s an app for that” era marked a shift in convenience, the agentic shift is nothing short of utterly transformative and on par with the birth of the internet. People at every level of the organization are racing to understand what an agent is, how to deploy one, and how to make the business case for funding.
And boy, is there an opportunity. At ORO, we now have 48 agents live or in customer UAT environments, spanning Contract Creation, Third-Party Risk Analysis, AP Fraud Detection, and even Autonomous Negotiation. We have definitively entered the agentic era.
But while this all feels like progress - more intelligent systems, more automation, less manual effort - there is something we need to be honest about. Something a number of tech companies seem to be willfully ignoring.
If you’ve spent any time inside real enterprise environments - dealing with scale, risk, and regulatory pressure - you see the flaw quickly:
Trying to build a world inhabited only by agents is very much the current "oasis." It’s a quick fix and looks plausible, but the nearer you get to reality, the more you realize the oasis isn’t real…
The appeal is obvious: why design processes, define ownership, or build structured systems when an agent can just… handle it? The allure of speed is blinding people to the difference between short-term wins and long-term value.
Need sourcing support? There’s an agent. Need contract review? There’s an agent. Need supplier onboarding? There’s an agent.
But this thinking confuses execution with architecture. Beneath that apparent simplicity lies something far more complex: an expanding network of agents, each operating with partial context, each requiring configuration, monitoring, and governance - and each taking no responsibility for their actions, their interactions, or their mistakes.
The uncomfortable truth: most organizations do not have the capability or the operating model to manage that ecosystem effectively and safely. Many of the companies offering this agentic army are in the same boat - not acknowledging that the agent builder and interface technology is actually the easiest part of the problem to solve. Where it gets hard is providing the right data, in the right structure at the right time to the agent. Rich bi-directional process and data integrations, as well as access to and the accuracy of data sources for the agent is critical. Delivering robust agent hardening, resilience, context shaping, and developmental experience required is not something you can just “flip a switch” with.
In the early days, agentic solutions seem efficient and accurate. They reduce effort. They accelerate tasks. They demonstrate quick wins. As part of an organization powered by agentic technology, I am a genuine and vocal proponent of agents and the value they bring. We need more agents, and more aspirational agents solving bigger problems.
I would go as far as to say that the only thing that can limit us in solving problems with agentic technology is our imagination - and governance.
In a world that acts as though every problem has an agentic solution, two things begin to happen over time: the number of agents grows, and the cost and risk of managing them grows faster exponentially.
When you deploy an agent, the costs aren’t just the agentic execution or the token charges. You pay for oversight, coordination, exception handling, and failure recovery. You pay for the framework and scaffolding to keep agents safe and controlled.
That framework also includes oversight. Yes, you can have agents that oversee other agents, and we have seen good success with that - again when deployed in the right framework. But ultimately, you need human oversight… and I don’t think you can delegate that to a vendor offering you an army of agents without establishing a clear chain of command and a strategic general.
There is value in deploying agents that are designed to live in a specific context. Taking an "off the shelf" agent and expecting it to thrive in your ecosystem is like taking an employee who only speaks French and asking them to run a business in Japan. Similarly, agents born to live and operate within their orchestration platform of origin will always be more effective than a generic agent parachuted into your world.
The real test of any system isn’t when it works. It’s what happens when it doesn’t.
What happens when an agent fails? Not in a demo environment, but in production. Under pressure. At scale. With real financial or regulatory impact. Who diagnoses the issue? Who understands the decision path the agent took? Who can confidently intervene?
This is where the “agent for everything” model begins to unravel. When agents are building, modifying, and maintaining increasingly complex logic - often abstracted away from human understanding within your organization - you create a gap. A gap between what the system is doing and what people can explain or control.
In regulated environments, that gap is not just inconvenient. It’s unacceptable.
This is where the conversation needs to mature: from “agents for everything” to agents as part of a governance roadmap and within an orchestration framework.
The organizations seeing real success are not the ones deploying the most agents. They are the ones designing and using the best platforms - ones that incorporate agentic technology in a way that delivers value now and stability in the future.
Orchestration isn’t just workflow tooling. It’s a platform that enables deep integrations, a single intuitive user experience, and effortless transparency. Policy and control are built in, so the right process is always the easy process. ORO’s “LEGO-block” architectural approach makes it easy to integrate all parts of the Source-to-Pay workflow into a dynamic, context-sensitive flow - and to insert agents cleanly, with full context across every data element the platform has access to.
Deploying agents outside of an orchestration platform is like releasing balloons without strings, or explorers without a map. Initially things look fine. Very quickly, they drift.
At ORO, we’re seeing a consistent pattern: while many capabilities begin as standalone agentic solutions, over time we are able to graft the agents into the platform itself, delivering the same outcome but anchored in a control framework. Structured. Repeatable. Natively governed. They stop looking like agents and become part of the platform. They move from experimental intelligence to operational capability.
The result: big problems get solved in an ever-more controlled and transparent way. Although new agents are always being introduced to tackle the most complex global challenges, at the other end we embed optimised agents directly into the platform - keeping the overall agent count, and therefore potential risk, at a manageable level. Stronger processes. Better visibility. Enterprise-safe governance. And critically: retained human oversight.
There’s a powerful place for agents. But treating them as the answer to everything is no different from the mistakes we’ve made before with apps, platforms, and previous automation waves.
Technology does not remove the need for design. It increases it. We are at a point in human history where - perhaps for the first time - tooling isn’t what will limit us, and resources aren’t what will limit us. Creativity, governance, and attention to long-term strategy are.
The leaders who succeed in this next phase won’t be the ones asking: “How many agents can we deploy?”
They’ll be asking:
“What should be agentic?”
“What should be native?”
“What must remain human-led?”
And how do we bring all of that together into a system that actually works - at scale, under pressure, over time?
The future isn’t a leaderless agent-army.
Rather, it’s an orchestrated battle plan, utilizing ground, air and sea technologies! The future will be one where:
Agents execute.
Platforms govern.
And people lead.