GDPR, HIPAA, PCI-DSS… meeting such compliance requirements remains paramount when introducing new contact center innovations.
So, contact center and IT leaders do their due diligence, and when they implement new AI solutions are typically confident in meeting those standards and securing their deployments.
However, the world moves on, and these AI solutions evolve.
Consider an AI agent. As it grows, it leverages more data to generate better responses. Without proper oversight, the development can drift away from what’s acceptable for the business.
The Nightmare Scenarios
Over the past 18 months, many customer-facing AI solutions have gone rogue.
Some have been comical incidents, such as the Virgin Money chatbot taking offence to the word “virgin”. However, others have been more serious.
A classic example is the New York City virtual agent telling small business owners to break the law. Another is a recent attack on Lenovo that had its chatbot spewing past customer conversations.
These risks are often discussed. Yet, there are rising dangers in agent-facing AI use cases, too.
For instance, in banking, if an AI guides agents to sell more credit products to customers already struggling to pay, the contact center risks violating fair credit regulations. That’s dangerous territory.
That incident beckons the first of three new security and compliance considerations for contact center leaders.
1. Don’t Over-Rely on Human Oversight
Both technology and human oversight are critical to successful AI deployments. AI must be constrained with boundaries and agents educated.
While “human-in-the-loop” is a term often banded about, just relying on human feedback is risky, since metrics may push them to follow AI guidance even when it’s problematic (per the banking example).
The key is to build a strategy around AI reporting, with alerts feeding into the contact center system. After all, real-time system monitoring is far more effective and future-proof than relying on humans.
2. Shift from Reactive to Proactive Quality Assurance (QA)
Continuously review outputs through QA, not just of call recordings, but also of the guidance AI provides. That’s critical. Yet, as Crispen Maung, SVP of Compliance and Privacy at Five9, stressed: