The hype surrounding AI agents is at full momentum, with the global market set to grow at a compound annual growth rate of 45% over the next five years, and UK businesses are rapidly looking at where the technology can help them gain competitive advantage.
With adoption of agents taking place across all industries, where businesses of all shapes and sizes are using them for different purposes, they are no longer just experiments — they are being put to work. As agents have the capacity to act autonomously as virtual employees, managing confidential data, and connecting with customers, the potential is huge.
Since July 2024, the UK AI industry has been attracting an eye watering average of £200 million in investment every day. When it comes to agent adoption, seeing return on investment from the likes of this cash will depend on organisations prioritising governance. The businesses trying to run before they can walk — i.e. those without governance in place at the time of agent deployment — are taking a significant risk. With the EU AI Act, upcoming UK legislation, and sector-specific rules tightening the net, governance is also no longer just a responsible option, it’s the foundation of trust and how organisations deploying AI agents can gain true competitive advantage.
No guardrails, high stakes: Why governance can’t wait
Regulatory scrutiny is only increasing. AI agent deployments must adhere to stricter safety, transparency, and accountability standards in alignment with the EU AI Act, upcoming UK legislation, and sector-specific regulations.
Yet too many organisations are still operating as they go. Measuring the quality of agent behaviour is often ad hoc, based on gut feel rather than consistent benchmarks, which undermines trust and makes it incredibly difficult to prove value.
Data is another obstacle. AI agents depend on proprietary, well-governed datasets, yet in practice, many organisations lack the volume, accessibility, or quality to train them effectively. Add to this the relentless rate at which AI models and tools evolve, and it’s no wonder that some projects are stalling before they can deliver meaningful results.
Governance drives trust
With the right governance, the source of all actions and outputs of agents can be traced back via the data lineage, from the raw data used for training to the logic executed in real time. Strong access restrictions and security measures are applied via a unified governance model, which handles agents with the same oversight and discipline as human employees.
Furthermore, it creates a single, consistent view across data and AI assets, removing siloes and enabling safe discovery and re-use. Governing the business semantics that underpin decision making is equally crucial, so both people and agents work to the same definitions of metrics and KPIs. It’s also imperative to monitor agents after deployment to detect any drift, bias, or harmful behaviour before they cause any real damage.
In the era of AI agents, fragmented governance models will fail to scale. These systems act autonomously to complete tasks, taking actions that can affect customers, finances, and brand reputation, and this isn’t something to be taken lightly. They must be governed with the same principles that apply to human workers: security, transparency, accountability, quality, and compliance. As an organisation’s technology stack evolves, governance needs to be both unified across all data and AI assets and open to any tool or platform. Otherwise, innovation risks being slowed by integration barriers.
Safe and effective scaling: from pilots to production
When properly implemented, lineage and governance fuel the rapid development of agents, without causing problems – transforming promising experiments into systems fit for production, and delivering results. This process is helping the most cutting-edge companies reduce the time between idea and implementation. They are able to adjust performance to strike the fine balance between cost and quality by automating the assessment and optimisation of their agents, creating synthetic data that fills gaps in proprietary sources, and developing domain-specific benchmarks.
Automated evaluation is especially important. Businesses that lack or overlook it are often forced to rely on inaccurate “gut checks” to determine whether an agent is performing well, which leads to inconsistent quality and costly trial-and-error. By contrast, those that generate task-specific evaluation, use synthetic data to enhance training, and optimise across the latest models and systems, can scale agents with justified confidence, knowing they meet quality standards while controlling costs.
Flo Health, the world’s leading health app for women, offers a great example. With an AI agent system, it doubled medical accuracy over standard commercial LLMs, while meeting strict internal standards for safety, privacy and clinical validity. This converted an experimental tool into a trusted production system in a highly regulated sector.
AI agents: The catalyst for competitive advantage
UK businesses have a tight time frame to take the lead in AI agents before global competitors overtake them. This leadership will come from deploying the correct agents, those that are safe, transparent, and based on controlled, high-quality data, as opposed to deploying a high volume of agents at lightning speed.
In order to achieve this, businesses must ensure that every system is based on a consistent business context, integrate assessment and optimisation into the agent lifecycle, and crucially, view governance as the backbone of their data and AI strategy.
Unregulated innovation is a no-go zone for any business. By designing AI agents with governance and clear lineage from the outset, UK organisations can build trust, inspire market confidence, and move beyond the hype to deliver real, measurable AI impact.
About the Author
Dael Williamson is EMEA CTO at Databricks. Databricks is the Data and AI company. More than 15,000 organizations worldwide — including Block, Comcast, Condé Nast, Rivian, Shell and over 60% of the Fortune 500 — rely on the Databricks Data Intelligence Platform to take control of their data and put it to work with AI. Databricks is headquartered in San Francisco, with offices around the globe, and was founded by the original creators of Lakehouse, Apache Spark™, Delta Lake and MLflow.
