As AI investments accelerate, trust and risk weigh heavily

The KPMG Q1 2025 AI Pulse Survey paints a clear picture: enterprises are doubling down on GenAI and AI agents, but confidence in managing risk and AI trust still trails behind. Investment is surging. Leaders now plan to invest nearly $114 million in GenAI over the next year, which is a sharp jump from $89 million just last quarter.

Organizations are also maturing rapidly:

• Piloting of AI agents has risen from 37% to 65% in just one quarter.

• Yet deployment of AI agents remains flat at 11%, which points to a gap between testing and full-scale implementation.

Despite this progress, major challenges remain top of mind for executives:

• 82% cite risk management (such as data privacy and security) as the biggest hurdle.

• 64% are concerned about the quality of organizational data. 

• 35% worry about personal trust in GenAI.

As usage becomes more embedded into daily operations, we’re seeing impressive adoption gains:

• Productivity tool usage on a daily basis surged to 58%, up from 22%.

• Knowledge assistants (like RAG models) are now used weekly by 61% of organizations, up from 48%.

• GenAI embedded into workflows grew from 24% to 35%.

All this data signals that businesses are moving beyond experimentation and starting to scale GenAI adoption—making it even more important to lay a strong foundation of reliability, transparency, and control.

Address trust and risk

Organizations can and should proactively manage the twin issues of risk and trust. Here are three ways to get started.

Fine-tune data carefully

GenAI and AI agents are only as good as the data they are trained on. Fine-tuning with high-quality, bias-free, domain-specific data is critical to building reliable and trustworthy AI.

• Focus on industry-specific and contextually relevant datasets.

• Use diversity audits to detect and correct bias during training.

• Continuously refresh datasets to keep agents aligned with real-world dynamics.

Fine-tuning data is about embedding accuracy, fairness, and relevance into the foundation of every AI-driven decision.

Manage agent-to-agent communication

As AI ecosystems grow more complex, managing how agents communicate with each other becomes essential.

• Set clear governance protocols for agent handoffs and messaging.

• Monitor and log communications to prevent misunderstanding or hallucinations.

• Implement standards like the Google A2A Protocol to promote clarity and stability.

Managing AI ecosystems properly will result in agents that are coherent, resilient, and capable of scaling responsibly.

Keep humans in the loop

AI agents are designed to be autonomous, but human oversight remains crucial for accuracy, safety, and brand protection.

• Build escalation protocols for when agents encounter uncertainty.

• Regularly review agent outputs and decision pathways.

• Learn from real-world cases, like when a customer support AI agent went rogue because of a lack of human oversight. Put safeguards in place to prevent similar breakdowns.

By taking these steps early, you can make your AI initiatives more resilient, accountable, and aligned with their strategic goals.

Centific can help you

Centific’s frontier AI data foundry platform gives you the tools you need to manage GenAI and AI agents effectively and responsibly:

• Fine-tune and validate data to train trusted, high-performing agents.

• Implement governance frameworks for agent collaboration.

• Embed human oversight into every layer of agent operations.

Working with Centific, you can maximize the value of AI while building the trust you need to thrive with customers, regulators, and employees alike.

Learn more about Centific’s frontier AI data foundry platform.