From hype to high-impact: How business leaders can realize ROI with AI agents

AI agents are no longer experimental—they’re operational. But for many executive leaders, the results have been underwhelming. According to the 2025 IBM Institute for Business Values C-suite Study, only 25% of AI initiatives have delivered the expected return on investment (ROI), and just 16% have scaled enterprise-wide. These numbers reveal a critical gap between ambition and execution.

The problem isn’t AI itself—it’s how it’s being deployed. Success with AI agents requires more than enthusiasm. It demands a structured, transparent and business-aligned approach that balances experimentation with governance, and cost savings with long-term growth.

So how can executives shift AI agents from pilot projects to real business value? By starting with the right mindset, grounding their strategies in cost-saving use cases, and architecting for scale and flexibility.

One of the most common missteps executives make is starting with the wrong ROI lens. Many leaders aim for transformative, top-line growth from day one, but the most successful AI implementations often begin with cost savings.

Why? Because cost benefits are easier to measure, faster to realize and provide the foundational business case needed to scale.

For example, consider an industrial company managing millions of unstructured documents. By deploying AI agents to analyze this federated and dispersed data, they could reduce the manual workload of expensive analysts and improve decision-making.

This process has the potential to lower costs, enhance revenue and allow analysts to focus on higher value work. Or take a company that uses AI agents to automate its claims processing workflow. This company could achieve faster payouts, happier customers and the ability to scale without increasing headcount.

These examples aren’t moon shots. They are practical, ROI-driven use cases that build momentum for broader enterprise-wide transformation.

To help guide your AI strategy and ensure that you achieve the ROI you expect, here are four actionable steps to consider.

Avoid the urge to “spray and pray” AI across the enterprise. Instead, start with clear, high-impact use cases. Look for these signals of AI readiness:

• Well-defined repetitive or menial tasks (for example, data entry or claims processing)
• Manual handoffs between systems and personas (for example, order-to-cash workflows)
• Quality data availability to ground AI agents and measure against
• Complex policy interpretation (for example, HR compliance or insurance eligibility)
• High cost or inefficiency (for example, legacy code maintenance)

Start small, prove value and scale from there.

If you don’t know your starting point, you can’t measure ROI. Yet, many organizations lack clear baselines for time, cost or quality. Before deploying AI agents, conduct a process decomposition exercise:

• How long does this process currently take?
• How much does it cost?
• What are the pain points?

Use internal data or industry benchmarks (like those from IBM’s Institute for Business Value) to establish a clear “before” picture. This information is going to be critical when demonstrating ROI to stakeholders.

One often-overlooked decision in AI strategy is architectural. Many vendors are pushing proprietary AI agents tied to their platforms. But most enterprises run heterogeneous environments. Locking into a single vendor’s AI agent can limit flexibility and innovation.

Instead, consider an open orchestration layer that can sit above your existing systems and integrate with multiple AI agents. This approach:

• Avoids vendor lock-in
• Enables faster pivots as technology evolves
• Supports enterprise-wide transparency and governance

In a fast-moving space, architectural agility is a competitive advantage.

ROI from AI agents can be measured in three primary ways:

1. Speed to outcome: How much faster can you complete a process?
2. Cost to serve: How much cheaper is it to deliver the same outcome?
3. New capabilities: What can you do now that you couldn’t do before?

The first two are easy to measure. The third—new capabilities—is harder to quantify but often the most transformative. For example, obtaining insights from decades-old documents or refactoring legacy code that no one dared to touch.

Don’t ignore these “net new” opportunities. Just be clear that they might require a different ROI lens—one focused on strategic value rather than immediate savings.

The biggest misconception about agentic AI? That it’s either too hard or too easy to implement with great results.

The truth lies in the middle. AI agents are not a silver bullet nor a science project. They are a powerful tool—when used with purpose, structure and vision.

For executive leaders, the path forward is clear:

• Start with cost savings
• Build a strong foundation
• Scale with confidence
• And align AI with your business strategy every time

To learn more about how IBM can help you orchestrate and govern AI and AI agents across your business, visit IBM® watsonx Orchestrate®.