Virtual agents and hand offs to human agents

In the UK, the use of bots (such as chatbots or AI-driven systems) for customer service, debt resolution or other forms of interaction with retail customers (i.e. consumers) is becoming increasingly common. However, there are several best practices and regulations to ensure that bots are implemented ethically, fairly, and legally, especially when enabling access to a human agent. This is particularly true when engaging with consumers with characteristics of vulnerability.

1. Transparency and Disclosure

Clear Identification of Bots:
It is essential that users are made aware when they are interacting with a bot. The Mega AI bot will clearly identify itself and explain its role in the interaction.

Availability of a Human Agent:
Users should be informed when they can switch to a human agent or escalate the conversation to one.

Out-of-Hours Scenarios:
Mega AI will consider overflow and extended hours service capacity planning.

2. User Consent

Informed Consent:
Users must be informed that a Mega AI bot is collecting or processing their data.

Opt-Out Options:
Users can opt out and request a human agent at any time.

Inbound and Outbound Considerations:
Especially relevant for PSR or VRS-registered consumers.

3. Data Privacy and Protection

GDPR Compliance:
Mega AI complies with UK GDPR regulations, including rights to access, correct, or delete personal data.

Data Minimisation:
Only necessary data is collected and securely stored with clear retention policies.

4. Accuracy and Escalation to Human Agents

Escalation Procedures:
Designed to transfer seamlessly to humans when needed.

Training & Monitoring:
Trained to handle a wide range of queries, with access to action need codes.

Record of Human Interaction:
Humans receive conversation history for context.

Service Level Considerations:
Availability of human agents during demand spikes is essential.

5. Ethical Considerations

Bias and Fairness:
Bots must be tested and monitored for bias.

Avoiding Deception:
Users should never be misled into thinking they are speaking to a human.

6. Consumer Protection

Advertising Standards:
Bots must adhere to ASA and potentially FCA promotion requirements.

Complaint Mechanisms:
Users must be able to lodge complaints and access human support.

7. Regulatory Compliance

• UK Communications Act 2003
• Electronic Commerce Regulations
• ICO and CDEI Guidance
• AI in regulated sectors

8. Accessibility

Accessibility Standards:
Bots should support screen readers, simple interfaces, and clear language.

9. Security

Data Protection:
Encryption, authentication, and other measures must be in place.

Cybersecurity Regulations:
Compliance with UK NIS standards.

Initial Conclusions

The UK's best practices for bot usage revolve around transparency, user rights, and data protection, while regulations ensure that bots operate within a framework of ethics and accessibility.

Escalation to Human Agents

Key Elements of an Effective Escalation Process

1. Clear Escalation Triggers

• Automatic Triggers
• Manual Escalation

Example Triggers:

• Bot fails multiple times
• User requests a human
• Bot detects frustration

2. Seamless Transition

• Context Transfer
• User Expectations

Example:
"I'm transferring you to a human agent now. It may take a few moments."

3. Smooth Handover Process

• Human Agent Availability
• User-Friendly Messaging

Example:
"One of our agents will now assist you..."

4. Escalation to the Right Human Agent

• Skill-Based Routing
• Priority Handling

Example:
Billing issues routed directly to billing department.

5. Transparency and Communication with the User

• Clear Updates
• Proactive Communication

Example:
“Thank you for your patience... In the meantime, would you like to check out our help articles?”

6. Human Agent Training and Readiness

• Effective Training
• Agent Awareness

7. Post-Escalation Feedback

• User Satisfaction Surveys
• Continuous Improvement

Example:
"How satisfied are you with how we handled your issue today?"

8. Compliance with Consumer Protection Laws

• Access to Human Support
• GDPR Compliance

Key Takeaways

• Seamless transitions with conversation history
• Easy and transparent escalation
• Trained human agents with full context
• Continuous improvement from user feedback

The Regulatory Outlook

AI voice agents in debt resolution are raising regulatory considerations, including:

• FCA Oversight if designated a Critical Third Party (CTP)
• FCA Authorisation if performing regulated debt collection activities

References

• Ofcom’s AI Strategy 2024/25
• FCA Debt Collection Authorisation
• FCA Handbook: CONC 7
• AI in Debt Collection Trends
• FCA AI Update 2024
• PS24/16: Operational Resilience for CTPs
• HM Treasury CTP Designation
• ICO AI Toolkit

ICO AI Governance and Accountability
Money Advice Trust – Data Sharing Principles

Kevin Still MCICM
Chair of the Advisory Board – Mega AI

‍