Why Orchestration Matters More Than Numbers

Don Schuerman, CTO and Vice President Marketing and Technology Strategy, Pegasystems.

The tech industry loves a good arms race, and agentic AI is the latest battlefield. Software vendors are sprinting to deliver armies of AI agents for their clients. If you believe the hype, every business will soon be managing thousands of agents, specialized to automate every conceivable business task—and they all come with breathless promises of unparalleled efficiency and productivity that your company must deploy or risk helplessly falling behind in the market.

It all sounds impressive in a marketing ad, but as enterprises pivot to actually implementing these AI solutions, they’re discovering a critical truth: the sheer quantity of agents matters far less than the quality of outcomes they deliver.

Beyond The Numbers Game

AI agents—those autonomous digital workers that can perceive, decide and act—are on track to fundamentally transform how businesses operate. However, we’ve reached a moment of reckoning as organizations evolve from experimentation to implementation.

Are all these new agents delivering the tangible business results most organizations envisioned? Not always, considering that Gartner predicts at least 30% of generative AI projects will be abandoned by the end of 2025 due to issues such as reliability concerns and governance gaps.

Think of it as if you were building a well-oiled team: Would you rather have a thousand unpredictable workers or a smaller group of trusted professionals, trained in the very best practices of your industry, who consistently deliver great results?

What Makes An AI Agent Truly Valuable

Everyone has their own definition of what an AI agent should do. In my opinion, a properly designed agent should accomplish three fundamental goals:

1. Automating Tasks

Agents should streamline tasks through effective automation—cutting down on manual work without sacrificing precision. That’s table stakes. More critically, when a proven workflow or best practice is already in place, the agent should adhere to it consistently rather than defaulting to unpredictable large language models.

2. Designing Workflows

To support this, strong workflows must be defined. AI agents should be capable of designing these workflows—mapping out the most efficient path to achieve a desired outcome. This ability is often overlooked, but it’s critical for ensuring agents stay focused and deliver consistent results.

3. Optimizing Workflows

Finally, it must optimize those workflows as it moves through its processes. This means it should be able to detect when a better approach is possible on the fly and then implement that change—if and when a human approves.

These last two capabilities are where most current implementations often fall short. Without the underlying workflows guiding their actions and their continuous optimization, AI agents can become too unpredictable and unreliable for an organization to consider deploying them in any serious business scenario.

The Predictability Problem

If you’ve ever played around with any LLM like ChatGPT, try to ask it the same question twice and see what happens. More likely than not, you’ll get two very different answers. While this creative flexibility is wonderfully useful in tasks like content generation, it’s wholly unacceptable in any mission-critical business process.

A bank could never tolerate an AI agent approving a loan for one person while declining the same loan for another person with largely the same application credentials. Similarly, a healthcare provider managing patient care can’t risk inconsistent responses to questions of critical care.

Regulatory compliance simply doesn’t have room for creative interpretation. This is precisely the risk holding back many AI agent deployments today.

The Orchestration Imperative

Imagine the digital chaos of thousands of AI agents running amok across your enterprise. What’s missing is an intelligent orchestration layer to ensure all agents are working together, acting in the organization’s best interests instead of freelancing as they see fit. This enables them to understand the user’s request, connect with the right agent at the right time, coordinate how they operate together and guide them through trusted workflows that keep them on track.

This means that enterprises need to rethink their existing workflows to leverage agentic automation and execute across channels. This requires a deliberate shift from ad hoc automation to a holistic strategy—one where teams redesign workflows with agent collaboration in mind, embedding modular structures, measurable outcomes and governance checkpoints directly into existing operational practices.

With effective orchestration, organizations can take the guesswork out of AI agents by:

• Following predefined best practices, which are critical in regulated scenarios.

• Calling out to the specialized agent best equipped for the task at hand.

• Dynamically creating new workflows on the fly when faced with unique scenarios.

• Seamlessly collaborating with human workers in the loop when needed.

• Continuously improving through learning from outcomes.

The Role Of Governance

Agent orchestration is just the start. Underpinning any proper agent architecture is a comprehensive governance layer that must ensure all AI agents are closely tracked, fully auditable and completely secure.

This means implementing frameworks that monitor agent behavior, explain their decisions and maintain compliance with regulatory requirements. At any moment, this system should be able to produce reports that provide total transparency as to what their agents have done and why they took those actions.

Without this governance layer, organizations expose themselves to significant risks—from regulatory violations to customer experience failures—regardless of how advanced their agent technology might be.

Another Big Step Toward The Autonomous Enterprise

The true winners in the agentic AI revolution won’t be the companies with the most agents—they’ll be the ones whose agents are most effectively orchestrated to deliver consistent, reliable outcomes.

Just as a symphony isn’t improved by simply adding more musicians, an agent ecosystem needs strong orchestration to transform its individual capabilities into harmonious business value.

After all, the ultimate goal is to deliver on the vision of the autonomous enterprise—one that brings people, processes and technology together to propel your business forward. Like a successful orchestra, this can’t happen without everyone playing from the same sheet music, providing the equivalent of well-defined guardrails that ensure reliability, accountability and compliance for every note played.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?