Agentic AI and consumer law: Why trust will determine adoption at scale
Guest Blog: Jocelyn Paulley, Partner at Gowling WLG
Agentic AI - systems that can act autonomously to make decisions, complete tasks and transact on behalf of users - is moving rapidly from concept to real-world deployment. Across UK sectors, the opportunities are clear: faster decision-making, more personalised services and reduced friction for consumers.
But the real question is not simply where agentic AI can be used. It is whether it will be trusted enough to be used at scale.
For both consumers and businesses, adoption will depend on confidence that these systems are reliable, accurate, predictable and compliant with existing regulation. Without that trust, the expected gains in efficiency, competition and customer experience will be difficult to realise.
The opportunity: Reducing friction across UK consumer markets
Agentic AI has strong potential in sectors where consumers face complex choices or repetitive decisions.
In retail and e-commerce, agents could compare products, select options and complete purchases. In financial services, they could identify suitable products or support switching. In utilities and telecoms, they could manage contracts and optimise tariffs. In travel, they could plan and adjust journeys dynamically.
Across these use cases, the value is consistent: saving time, reducing search costs and improving outcomes. For businesses, it offers more responsive and personalised services at scale.
However, these benefits rely on a key assumption: that the AI is acting in the consumer’s interests, within clear and understood boundaries.
Trust issue one: Can an AI agent really act in the best interests of a consumer?
As agentic AI becomes more autonomous, a central question emerges: what does it mean for an AI system to act as an agent for a consumer? The legal concept of agency requires the agent to act in the best interests of its principal. However, an AI agent will do what it is instructed to do, but cannot understand broader context, ethics or benefits without being instructed. It is not clear that the current legal concepts of agency can be easily applied to an AI agent.
If an AI is selecting products or entering transactions, key issues include:
Authority: what permissions does the agent have, and how are these controlled?
Accountability: how can a consumer be confident the AI stays within its instructions?
Explainability: can it justify its decisions?
Error handling: what happens if it gets something wrong?
These questions have practical consequences. If a contract is entered into by an AI on behalf of a user, is it binding? Does a counterparty need to know it is dealing with an AI rather than a person?
Many consumers will not fully understand how an agent operates or its limitations. That raises the question of whether AI agents are fit for purpose where decisions have financial or legal consequences for the consumer. What if an AI agent makes a bad decision which the consumer acts on? Terms and conditions for software products, and particularly AI, typically shift risk and responsibility for use to the user. But contracts with consumers must not contain unreasonable terms and must be easy for consumers to understand. That seems like a difficult square to circle with complex technology.
Ultimately, trust depends on whether an AI agent can be relied on to act in the best interests of its user - and in a way that is understandable and controllable.
Trust issue two: when businesses deploy agentic AI
Trust is equally important where agentic AI is used by businesses.
If an AI system is recommending products or guiding a customer journey, the business remains responsible for its behaviour. At scale, this can amplify risks quickly.
For example, an AI agent could:
present incomplete or misleading information
steer consumers towards favourable outcomes for the business, not necessarily the consumer
create artificial urgency or pressure
fail to clearly communicate key terms
These risks are not new, but automation increases both their speed and reach, and raises the possibility that they happen without the knowledge of the business.
There are also data protection challenges. Agentic systems may process large volumes of personal data, generate inferences and act in ways that are harder to monitor or explain. This increases the importance of accuracy, transparency and accountability.
Even basic design choices matter. Making terms and conditions technically available to an AI agent may not satisfy the legal obligation to make information available to the consumer. Similarly, if an AI provides inaccurate information to the consumer, this could amount to misrepresentation by the business.
In this context, explainability and transparency are essential to both compliance and trust.
Trust issue three: when both sides use AI agents
The complexity increases further when both consumers and businesses rely on AI agents.
If AI is effectively interacting with AI, traditional concept of 'agreement' become harder to apply. There may be no clear moment of human review or explicit assent by the parties themselves, raising questions about whether there was a genuine intention to be legally bound or simply automated execution.
If something goes wrong, allocating responsibility also becomes more complex. For example, what losses were reasonably foreseeable where outcomes were determined by automated systems? How should indirect or unintended consequences be treated?
These are developing issues, but they underline the need for clear safeguards.
What needs to change to enable responsible scale?
To unlock the full potential of agentic AI across UK sectors, three sets of levers need to be addressed.
Technical levers include improving reliability and accuracy, building explainable systems, and implementing clear permission controls so agents act within defined limits. Strong identity and authentication are also critical to ensure agents are properly authorised, and systems must be auditable and interoperable to avoid consumer lock-in.
Organisational levers include clear accountability for AI behaviour, ongoing monitoring and testing against real customer journeys, effective complaint handling, and robust oversight of third-party providers. Agentic systems should be treated as operational actors that require supervision.
Policy levers include clearer regulatory guidance, development of trusted digital identity infrastructure, and standards that support transparency, portability and fair competition.
Trust is the condition for scale
Agentic AI presents a significant opportunity to reshape UK consumer markets, making services more efficient, personalised and easier to use. But adoption will not be driven by capability alone. It will be driven by confidence.
Consumers need to trust that AI agents act within clear limits and in their interests. Businesses need confidence that these systems are compliant, controllable and commercially defensible.
In that sense, trust is not a secondary concern. It is the foundation for scaling agentic AI responsibly across the UK economy.
