From AI Projects to Profits: How Agentic AI can Sustain Financial Returns

A Report by CYS Global Remit FinTech Development Unit

The AI ROI Reset 🎯 

The initial excitement surrounding generative AI has given way to a more grounded understanding of its potential and its limits. This marks an important step in how businesses are implementing AI and assessing its impact. Instead of chasing short-term, project-based returns, organisations are now focusing on outcomes that truly influence the bottom line. 

In the early days, generative AI pilots promised extraordinary returns—sometimes as high as 31%. But as these projects scaled across enterprises, the figures settled at a more modest 7%—slightly below the average cost of capital at around 10%. Even so, operating profit improvements tied to AI have continued to grow steadily since 2022. This shows that AI is no longer an experiment but a genuine driver of financial performance, delivering real value to shareholders. 

Companies are also becoming more strategic in their approach: 

• 64% of budgets are now channelled into core business functions that underpin competitive advantage. 

• Only 36% goes to non-core activities. 

While this strategy is more complex than focusing on “quick wins” at the margins, it lays the foundation for greater scale and sustainable long-term returns.

Agentic AI: A Paradigm Shift ⚙️ 

A new chapter is opening with agentic AI—autonomous systems designed to manage and optimise complex workflows. Unlike robotic process automation (RPA), which automates static, repetitive tasks, agentic AI enables dynamic workflows that move businesses from simply “saying” to actually “doing”. 

This shift is expected to be profound: 

• AI-enabled workflows are forecast to grow eightfold, from just 3% in 2024 to 25% by 2026. 

• This represents nothing less than a fundamental rethinking of business processes across industries. 

  
  

Executives clearly see the potential: 

• 70% believe agentic AI is both market-ready and vital to their organisation’s future. 

• 76% are actively encouraging experimentation to complement theory with hands-on practice. 

  
  

Key expectations for agentic AI include: 

• 83% expect a dramatic uplift in efficiency and output. 

• 83% anticipate greater success in process re-engineering and workflow reinvention. 

• 71% foresee AI agents adapting autonomously to evolving workflows. 

• 69% highlight better decision-making thanks to improved data access and insights. 

• 67% expect significant cost savings through automation. 

The Crossroads: AI-First vs. Fragmented Approaches 📈 

As global adoption accelerates, AI is creating a clear performance divide. Around one in four organisations are considered “AI-first”—and they are pulling ahead of competitors with more fragmented, tactical approaches. 

AI-first businesses report that over half of their revenue growth and operating margin improvements stem directly from AI. These gains go beyond operational efficiency: such organisations are using AI to: 

• Enter new markets. 

• Reach untapped customer bases. 

• Rethink business models rather than simply competing within existing boundaries. 

A critical differentiator is data maturity. While 68% of AI-first organisations report strong data management and governance, only 32% of others can say the same. These leaders understand that no algorithm, however advanced, can overcome poor data quality—and they invest heavily in ensuring their foundations are sound. 

Taking Action: A Guide for Leaders 🗺️ 

To thrive in an AI-first world, leaders must move beyond pilots and start deploying AI agents capable of managing complex workflows and delivering measurable outcomes. The following actions are essential: 

• Secure consistent executive sponsorship – adopt a “do as I do” mindset, openly acknowledging that AI will disrupt all roles, including leadership. 

• Align initiatives with strategic priorities – focus efforts on business problems where AI, especially agentic AI, can deliver maximum value. 

• Shift accountability to business leadership – business units, not technical teams, should own AI outcomes with clear, measurable targets. 

• Embed governance from the outset – lead with trust, ensuring monitoring, guardrails, and compliance are in place from day one. 

• Industrialise data into products – replace ad-hoc data pipelines with reusable, well-governed datasets to fuel AI agents at scale. 

References

https://www.ibm.com/thought-leadership/institute-business-value/report/agentic-ai-profits