The DRCF’s latest consumer research highlights why successful AI adoption depends on more than technical performance. AI’s value is not just what it can do for a business. It is what it delivers for the person using it. For businesses, the opportunity is to build trust through context-aware safeguards, meaningful transparency and clear accountability.
Published on 12 November 2026, Consumer Interest and AI, a report by pattrn.ai for the Digital Regulation Cooperation Forum (DRCF), explores consumers’ understanding and experience of generative and agentic AI, and the factors shaping their tolerance of AI-related risks. It recognises significant potential benefits, including greater accessibility, tailored advice, and time and cost savings. Yet benefits and harms can arise from the same features: personalisation can improve relevance but also enable manipulation, while human-like or artificially empathetic design can create misplaced trust.
The central message is that AI can deliver significant consumer benefits, but consumers cannot reasonably be expected to understand and manage all its risks. For organisations developing, procuring or deploying AI, efficiency and consumer protection should be designed together, not traded off against each other.
What is the report, and why does it matter?
This report by researchers at pattrn.ai draws on findings from the DRCF’s recent call for input and relevant literature, including AI research from its members: the Competition and Markets Authority, Financial Conduct Authority, Information Commissioner’s Office and Ofcom. The report draws on 30 responses to Phase I of the DRCF’s call for input and a targeted literature review. It should be noted that it is an evidence review, not a statement of regulatory positions, advice or guidance. Its findings should also be read in context: more than half of the submissions came from industry, and direct consumer evidence was comparatively limited.
Even with those limitations, however, the report provides a useful lens for reviewing AI governance as its scope extends to internal tools that affect customer experience. Three themes are particularly relevant.
Understand the context, not just the customer category
Consumer vulnerability is not simply a permanent characteristic or demographic label. Older people, children and digitally excluded users may face particular risks, but financial pressure, emotional distress or the circumstances of an interaction can make even technically confident users vulnerable. AI can help identify those moments and connect people with support. It can also exploit them through persuasive recommendations, personalised pricing or interactions that encourage misplaced trust.
For businesses, this means testing the customer journey, not just the model. A recommendation that is helpful in one context may be harmful in another. Product and risk teams should consider how a system behaves when a user is distressed, uncertain or under pressure, and whether its commercial objectives could work against that user’s interests.
Further, the report finds that greater use of AI does not necessarily equate to greater trust. Continued use should not be treated as evidence of informed trust or acceptance of risk. AI literacy remains valuable, but it is not a substitute for safe design. Consumers’ willingness to use and trust AI depends on context, the stakes of the decision, reversibility, the extent of AI autonomy and meaningful human oversight. Consumers may continue using AI because they lack alternatives or are unaware of the risks.
Make transparency useful at the point of decision
Providing information is not the same as creating understanding. The report records a preference among respondents for layered, contextual and just-in-time disclosures rather than lengthy explanations that users are unlikely to read.
The practical question is whether consumers can understand what matters when it matters: (a) whether AI is involved, (b) what it can do, (c) how their data is used, (d) what controls they have and (e) where to get help.
For agentic AI, which moves from providing information to taking action, that distinction becomes critical. A user should be able to distinguish between asking an agent to compare products and authorising it to make a purchase. Consider clear permission boundaries, spending limits, confirmation before consequential actions and accessible ways to stop the agent.
The aim should be informed control, not simply an accepted notice. Disclosures should explain the safeguards built into the service, not transfer responsibility for managing complex risks to the consumer.
Share accountability without complicating redress
The evidence reviewed in the report suggests that clear protections, accountability and access to redress can strengthen consumers’ trust in, and comfort with, AI. Consumers value knowing who is responsible, what protections apply if something goes wrong and how to appeal a decision. Those protections must be explained accurately, without implying that safeguards or remedies exist where they do not.
The report identifies broad support among respondents for proportionate accountability across the AI value chain. It does not establish a new joint legal duty for developers and deployers. How responsibility should translate into legal obligations remains an open question, particularly where several agents or organisations contribute to an outcome.
For businesses, the immediate priority is to make accountability work in practice. Supplier arrangements and internal governance should clearly allocate responsibility for controlling and reducing risks, investigating incidents, retrieving records, intervening and putting things right. Consumers should not have to navigate those arrangements to get help.
Action logs, permission records and traceable hand-offs between agents can help reconstruct what happened. A clearly signposted complaints route should connect consumers with people who have the authority and technical capability to resolve the problem. Human oversight must be capable of changing the outcome. As techUK emphasises, but need not be continuous. For example, and as techUK emphasises, requiring approval at every step may be impractical for agentic systems.
Industry response: clarity over a new regime
techUK welcomes the report’s emphasis on contextual vulnerability, meaningful oversight and consumer outcomes. Its policy position is that existing UK laws provide a strong foundation, including for agentic AI. It calls for clarity and coherence across the four DRCF regulators, with responsibility allocated to those who control risks and can reduce them, and less placed on consumers as AI autonomy increases.
Its response also shifts the focus from pre-launch paperwork towards monitoring and rapid remediation after deployment. techUK argues that tolerable risk should be assessed against the process AI replaces, recognising that no process operating at scale is error-free. For smaller firms, it seeks worked examples, reusable tools and a single clear entry point, such as a permanent DRCF AI and Digital Hub.
These are industry proposals, not conclusions adopted by the DRCF. The report records disagreement on the adequacy of existing regulation and strict liability. For businesses, the useful takeaway is to make safeguards demonstrably effective throughout deployment, rather than assume that either more documentation or new legislation will resolve every risk.
What can businesses do now?
The following six checks are practical takeaways from the report, rather than new regulatory requirements:
- Map consumer impact. Assess risk by use case, degree of autonomy and potential consequences. Include back-office AI and third-party tools, not just customer-facing applications.
- Test changing vulnerability. Assess foreseeable moments of distress, financial pressure or confusion, alongside the needs of groups at greater risk.
- Review disclosures in the journey. Test whether users understand the system’s role, data use and permissions before consequential steps.
- Set boundaries for agents. Make permissions understandable and controls accessible. Define what requires approval, when permissions expire and how users or staff can stop an action. Ensure human escalation can lead to intervention and correction.
- Rehearse an incident. Clarify responsibility across suppliers and internal teams, and retain records that allow them to reconstruct events. Test whether consumers can access a straightforward complaints process and an effective remedy.
- Measure outcomes, not just adoption. Monitor outcomes after launch, including complaint rates, resolution times and reversals or overrides. Compare performance with the process AI replaces, while keeping fairness, effective human escalation and consumers’ ability to challenge decisions in view.
Our view
Strong AI governance should enable innovation, not simply constrain it. Read together, the report and techUK’s response make a compelling practical case for proportionate safeguards that work in real customer journeys, supported by monitoring and effective remedies. AI quality should be judged by consumer outcomes, not technical performance alone: accuracy matters, but so do fairness, privacy, control, accessibility and the ability to challenge decisions or obtain a remedy.
The report offers a useful organising framework: performance, agency and recourse. Does the system work well? Can the consumer exercise meaningful control? Can something be done when it goes wrong?
Businesses that can answer all three will be better placed to turn AI’s efficiency gains into lasting consumer value.





.jpg?crop=300,495&format=webply&auto=webp)
.jpg?crop=300,495&format=webply&auto=webp)




.jpg?crop=300,495&format=webply&auto=webp)
.jpg?crop=300,495&format=webply&auto=webp)




_11zon.jpg?crop=300,495&format=webply&auto=webp)

.jpg?crop=300,495&format=webply&auto=webp)