Scaling AI agents: Why trust and data will make or break enterprise value 

Irish companies are betting on AI agents to transform work

Irish organisations are moving beyond AI agent pilots, but trust and data foundations remain the biggest barriers to scale. PwC’s AI Agent Survey 2025 finds that 70% of organisations plan to increase AI-related budgets off the back of interest in AI agents. Yet only 9% report broad adoption, with most still exploring or progressing limited deployments.

Early benefits are visible. While 53% of Irish firms report productivity gains from AI agents, only 38% see cost savings. This highlights the gap between task efficiency and enterprise impact. Trust is particularly thin: just 7% express high trust in agents across multiple functions.

Why trust is the bottleneck

The impact of low trust is not abstract; especially for specific high stakes activities. No Irish respondents report high trust in agents to conduct financial transactions, only 4% for autonomous customer interactions, and 9% for data analysis and insights.

Companies hesitate to hand over critical decisions to systems they can’t fully audit, govern, or explain. This explains why measurable value is lagging despite investment intent.

In parallel, the 2026 Global Digital Trust Insights Survey shows a world grappling with other data-related tensions: 41% of Irish organisations are adopting AI and machine learning to strengthen their cyber defence, while globally managed services are being used to support AI (38%), threat management (28%), data protection (27%) and third-party risk (17%).

Yet, 39% of respondents globally indicate that a data breach has cost their organisation over $500,000 in the last three years, and most companies (83%) split spend evenly between proactive and reactive measures — a sign that resilience and trust mechanisms are still maturing.

Data foundations still decide outcomes

Ireland’s trust gap is anchored in data readiness. Four in ten organisations cite data issues as the top barrier to realising value from agents. Integration with legacy systems is also more acute locally than in the US.

The Digital Trust Insights Survey aligns with this finding: Irish firms are increasing cyber risk investment (57%), but knowledge gaps and unclear risk appetite remain prominent barriers to using AI for cyber defence.

Without high quality, well governed, securely accessible data, agent performance, trust and return on investment will stall.

From back office to the front line

Adoption patterns are shifting. Customer service is now the leading use case in Ireland, ahead of operations and finance. But adoption remains patchy in sales and marketing, and only a minority are using agents to redesign processes or create new products.

That hesitancy mirrors cyber trends: while 78% of organisations expect cyber budgets to rise and 41% are adopting AI and machine learning to strengthen their cyber defence, many teams admit to skills and knowledge gaps in applying AI responsibly.

Until governance, assurance and risk ownership are crystal clear, leaders will keep agents near the “human in the loop” perimeter and away from autonomous decisions in revenue critical journeys.

What Irish leaders can do now

Build transparent guardrails: Treat trust as a design requirement, not an afterthought. Define “high stakes” thresholds (such as financial postings, contractual commitments or regulated customer interactions, for example) and require explainability, human validation and auditable logs for those flows. The Digital Trust Insights Survey data shows boards are backing cyber and data risk investment; use that momentum to embed Responsible AI policies and model risk controls alongside your existing cyber frameworks.

Fix data at the source: Prioritise the datasets that feed your first scaled agents and apply basic hygiene: lineage, quality rules, access controls and retention. Integrate with legacy systems via standardised APIs to limit brittle point-to-point builds. Our AI Agent Survey indicates that data issues and integration are the biggest Irish barriers; solving them will lift both performance and confidence.

Make cyber and AI one conversation: Security leaders are already prioritising AI capabilities; align this with your agent roadmap so threat detection, data loss prevention and identity controls are tuned to agent behaviours (for example, elevated scope service accounts or automated actions). Given that only a small fraction of organisations report full capability across data risk measures, closing this gap raises assurance and accelerates adoption.

Upskill for oversight, not just usage: The surveys point to knowledge and skills gaps as top obstacles. Train for oversight, not just usage. Governance, failure modes, red-teaming and incident playbooks matter as much as daily operations. This supports safe autonomy where it matters and reduces reliance on ad hoc controls.

Measure enterprise value, not task wins: Link agent outcomes to cost to serve, cycle time and revenue metrics, not just task time saved. With only 38% reporting cost savings despite widespread productivity gains, shifting measurement will force process redesign and accountability at the business unit level. Start with two or three journeys where you control both the data and the decision rights.

The destination is trusted autonomy

Irish leaders agree: AI agents will reshape work. But without trusted autonomy built on secure data and transparent governance, the promise will stall.

Start earning trust now, and enterprise-scale value will follow.   

This article was first published on www.businesspost.ie