Series
Blogs
Custobots and consumer law: Navigating the risks and opportunities of agentic AI
Custobots and consumer law: Navigating the risks and opportunities of agentic AI
15 June 2026
Series
Blogs
15 June 2026
Authors: Rachel Malloch, Kim Rust
The emergence of agentic AI signals a fundamental shift from reactive, generative tools to proactive, autonomous systems. AI agents move beyond the capabilities of chatbots. They sense, decide and act to accomplish a specific goal with limited human oversight. AI agents may soon be embedded into how we buy, sell and produce goods and services. While there are considerable variables (and inherent uncertainty) around how agentic AI will be deployed across sectors, it is entirely possible that in the near future AI agents will act as personal assistants or “custobots” intermediating between customers and merchants, reshaping markets and market actors.
AI agents offer huge potential in terms of consumer welfare gains by reducing the “friction” of everyday admin. But they also give rise to significant consumer risk, including around transparency and fairness, as well as the concern that agents may be vulnerable to manipulation. In an earlier post, we explored the key antitrust risks surrounding agentic AI. In this post, we zoom in on the consumer opportunities and risks, how to comply today plus some crystal ball gazing on whether consumer law will remain fit for purpose in an age of “custobots”.
The promise of personal agents is that the user can delegate everyday tasks (within clearly defined preferences and constraints), freeing up time for other things. AI agents, by automating the customer journey, can generate significant time savings and reduce cognitive load, secure better deals through augmented “shopping around” and ongoing monitoring, guide or even execute switching between providers, and offer a hyper-personalised experience by, for example, proactively flagging unused subscriptions or alerting customers before price rises.
At the same time, AI agents give rise to considerable consumer risk. The CMA in its recent research and analysis points to various risks including:
All of the above raise issues around fairness, transparency and accountability, and with that the potential for breaches of consumer law, whether related to specific statutory rights (e.g. cancellation rights and mandatory disclosures) or broader rules on misleading consumers.
Given its relative nascency and the uncertain trajectory of development, the CMA has focused its efforts on research to understand how technology is developing and providing guidance to help businesses comply with consumer law as agentic AI develops.
The CMA’s starting point is (unsurprisingly) that consumer law applies, regardless of whether a human takes a decision directly, or delegates decision-making to a bot. It takes a principles-based, compliance-by-design approach across four key stages of AI development and deployment:
The adoption of agentic AI is exciting, fast-paced and unpredictable. Questions of liability and compliance today extend to broader questions around the sufficiency of consumer regulation in a brave new world of AI bots. Could we see a migration towards markets where both customer and merchant deploy agents which transact freely with one another? A world in which “custobots” allow consumers to anonymise themselves (and in the process overcome several limbs of the “vulnerable consumer” test), while merchant AI agents deploy personalised pricing – based on willingness to pay, inflationary increases, competitor benchmarking, optimised bundling/discounting etc. – and AI-driven inventories? Would the “average consumer” and “transactional decision” tests, designed with human behaviour in mind, remain appropriate?
For now, the CMA’s approach appears to be curiously enthusiastic: encouraging growth, cautious to ensure compliance, and keen to come alongside businesses as technology develops.