Agentic AI: Bridging the Widening Gap Between Ambition and Execution

AWS recently partnered with Harvard Business Review Analytic Services to understand the current state of agentic AI in organisations.¹ The results were exciting and informative: While expectations are high, the path to value at scale has yet to be discovered.

Outlined below is what we found creates the gap between appreciating AI’s importance and using it effectively.

The market for AI is exploding, with investments forecast to reach over $190 billion by 2034, up from $5.2 billion in 2024. The ambition is considerable. Of the 623 business decision-makers surveyed, 84% believe it will transform their own business. These predictions are backed by investments, with 79% of respondents stating that their organisation plans to increase financial investment in agentic AI over the next year. And 17% of organisations are so excited about the potential that they predict more than half of their organisation’s business processes will be fully automated by agentic AI in the next two years.

This excitement is already translating into real business results for those who use the technology effectively. 36% of those using agentic AI today report achieving greater organisational productivity; 35% cite better data-driven decision-making; and 33% point to cost savings.

However a closer look reveals the baseline challenge: While 74% of leaders agree that AI use is very important, only 26% report that their organisation is currently “very effective” at leveraging any type of AI for positive business outcomes. Organisations recognise transformational potential; leaders express enthusiasm; investments flow—and yet, when it comes to capturing value, many struggle to bridge the execution gap at scale.

The Root Causes: Foundational Readiness and Trust

Our survey highlights three foundational areas where organisations are severely underprepared, hampering execution at scale:

• Data: Only 13% of respondents believe their data architecture is “well-equipped for agentic AI use,” while another 64% rate it as “somewhat equipped.”
• Governance: Just 11% report being “very well-prepared” with adequate structures, while 55% are “somewhat prepared.”
• Workforce: This is most concerning, with a mere 5% feeling “very well-prepared” to take advantage of the technology and 48% citing “lack of skills” as a top barrier.

In addition to these structural gaps, trust remains a huge barrier—trust that agentic AI will work in a way that does not harm the organisations and trust that the value from agentic AI will be provable.

Trust isn’t a technical problem to be solved through better algorithms or more robust testing. It’s a human and organisational challenge that requires transparency, explainability, and demonstrated reliability of the technology. While technology is ready, humans are not.

Employees might resist agents due to a “black box” effect: If they can’t see the agent’s chain of thought, they won’t delegate critical tasks. This lack of trust partly stems from the extent of autonomy organisations are willing to give their agents. Our survey finds that nearly half of organisations are hesitant to cede operational decisions, preferring a level of human intervention that could ultimately defeat the desired benefits of speed and efficiency.

Then there is the issue of trust in results. While executives’ guts and peers tell them of this technology’s potential impact, they fear that the expected value won’t materialise. Fear of missing out is driving a rush to invest. In this rush, nearly three-quarters of organisations lack a clear measure of value, making it difficult to prove business value.

Closing the Gap

The organisations that are successfully bridging the execution gap understand that foundations translate into value. These leading organisations are more likely to see results in innovation (42%) and customer experience (39%) compared to laggards.

To join them, we recommend that executives focus on four priorities:

1. Invest in Foundations

In addition to experimenting with and building agentic solutions, invest in addressing broken infrastructure, such as data architecture and governance structures.^2,3,4

2. Invest in Talent Now

Be transparent about upskilling and concerns about job displacement. To fully integrate agentic AI into your workflows, you need to invest in both technical skill training and organisational change. For example, coach people to calibrate trust—when should they let the agent run and when should they intervene? They may have an identity crisis as they shift from executing tasks to judging and stewarding agents and need your help to work through it.^9,10,11

3. Build Trust Systematically

Start with lower-stakes applications to demonstrate reliability and learn which guardrails are required to find the right balance of autonomy and oversight for your business. Experiment with solutions like observability toolkits. Recording an agent’s reasoning as a traceable log allows humans to audit why an agent made a particular decision. Neither complete human control nor full agent autonomy is optimal. Instead design systems with appropriate guardrails and oversight for the decisions being made.^6,7,8

4. Define Success Early

The study shows that 95% of organisations lack clear success metrics, leading to executive blind spots. Establish concrete, measurable objectives before you begin. This discipline keeps the focus on outcomes, not technology.

Use tooling to track progress against these metrics. Think big; don’t focus on short-term results—this is a long game. The most sophisticated enterprises are not just using AI to optimize existing (expensive) processes; they’re using it to design new products and services for their customers. Real value comes from reimagining end-to-end processes and cross-functional value chains. These organisations rethink how their technology organisation integrates with the business. They melt away functional boundaries and measure business value rather than SLAs and other internal metrics. They use metrics to learn, not to meet absolutes.¹¹

A Bright, Agentic Future?

Agentic AI offers a world of opportunity to reimagine cross-functional value chains, while innovating faster for customers. But as the data suggests, ambition is outpacing readiness. The widening gap between the 84% who see AI’s potential and the minority who are prepared to execute on it shows us that adopting agentic AI is not a plug-and-play upgrade. It requires a deliberate strategy to modernize data estates, recalibrate risk frameworks, and upskill talent. The future of AI might be bright, but it belongs to the leaders who treat this as a systemic organisational evolution.

I hope you find this report as insightful as I did, and that it gives you the impetus, focus, and courage to capitalise on this moment.

—Jana

1. Agentic AI: Expectations, Readiness, and Results, Harvard Business Review Analytic Services.
2.  From Automation to Agency: Leading in the Era of Agentic AI,  Ishit Vachhrajani
3.  Data and Generative AI: A Window into Your Organisation’s Soul?, Phil Le-Brun
4. Data Governance in the Age of Generative AI, Tom Godden
5. Your AI is Only as Good as Your Data, Tom Godden
6. Responsible AI: From Principles to Production, Helena Yin Koeppl
7. Overseeing AI Risk in a Rapidly Changing Landscape, Mark Schwartz
8. Responsible AI Best Practices: Promoting Responsible and Trustworthy AI Systems, Tom Godden
9. How Technology Leaders Can Prepare for Generative AI, Phil Le-Brun
10. Learners lead, leaders learn: The case for technology fluency in the C-Suite, Phil Le-Brun
11.  The Octopus Organization: A Guide to Thriving in a World of Continuous Transformation, Phil Le-Brun and Jana Werner, Harvard Business Review Press 2025