A stressed contact center agent at a bank takes a call from someone who says their credit card was stolen. The caller sounds distressed and wants the agent to help them log into their mobile banking app so they can see if any charges have been made.
The agent begins walking through the verification process, but the caller says they can’t remember the answers to their security questions and stresses that they need to get into their account right away. The agent, sensing that the caller is becoming frustrated and worrying about their handle time, bypasses the security questions and helps the caller reset their account password.
You can probably see where this hypothetical scenario is going. A bad actor gets access to a customer’s bank account by manipulating a contact center agent, ultimately costing the bank and damaging the real customer’s trust.
Voice fraud is becoming increasingly common in contact centers. According to a TransUnion survey, 6% of all calls into US contact centers last year were fraud attempts, up 33% from the previous year.
Customer impersonation is the most common type of voice fraud, typically involving a bad actor pretending to be a customer to gain account access or collect more personal information to use in future attacks. However, voice cloning fraud—which involves using AI deepfake technology to replicate a person’s voice—is on the rise. There are now more than 350 voice cloning tools available, and deepfake calls rose 1337% between 2023 and 2024.
As fraud attempts become increasingly sophisticated, they’re also becoming increasingly challenging for humans to detect—creating an urgent need for fraud prevention technology that can identify suspicious patterns at scale and empower agents to respond effectively in real time.
Below, we’ll cover:
Why are contact centers vulnerable to voice fraud?
Contact centers face a perfect storm of conditions that make them attractive fraud targets. The high-volume, fast-paced environment creates multiple vulnerabilities for bad actors to exploit.
Pressure to meet performance targets
Contact centers often hold their agents accountable for meeting average handle time (AHT) and first-call resolution (FCR) targets. This may cause agents to rush through calls, potentially missing verification steps or failing to spot red flags in their effort to get to a resolution as quickly as possible.
Emotional calls that take a toll on agents
It’s no secret that being a frontline agent can be emotionally taxing—87% of contact center workers report experiencing high stress. Taking an already-stressed agent and adding in emotional or urgent pleas from a caller (common social engineering fraud tactics) is a recipe for successful voice fraud attacks.
Many agents=many attack opportunities
The numbers game works in fraudsters' favor. If a fraudster senses that one agent is becoming suspicious, they can simply hang up and call back. With large contact centers receiving thousands of calls daily, they'll likely reach a different agent who may not have their guard up. Bad actors only need to find one vulnerable entry point to succeed.
What is voice fraud costing contact centers?
Voice fraud comes with both direct and indirect costs:
Direct financial losses
The direct financial impact of voice fraud is the most obvious one, with costs stemming from successful account takeovers, fraudulent refunds, and unauthorized transactions. A study from technology company Regula found that surveyed businesses across industries lost an average of $450,000 in the past year to deepfake fraud alone. Financial service businesses (a sector at a particularly high risk for fraud attempts) reported losing more than $600,000 on average.
Operational costs
One of the ways contact centers combat fraud is by adding more verification steps (e.g., requiring a caller to provide their account number and answer a series of security questions). More verification steps mean longer call times, driving up the average cost per call.
Reputational damage
Reputational damage can be the most costly consequence of fraud. When customers learn their accounts were compromised through a contact center breach, trust erodes quickly. With 73% of customers saying they would abandon a brand after just one negative experience, the long-term impact of fraud-related security incidents can be devastating.
The indirect costs of voice fraud compound over time. Contact centers may need to invest in additional training, implement more complex security protocols, or face increased regulatory scrutiny—all of which divert resources from customer experience improvements and growth initiatives.
What are contact centers doing to combat voice fraud?
Most contact centers are taking a multi-pronged approach to combating voice fraud. Common tactics include:
- Voice biometrics. This involves using a customer’s voice to validate their identity. However, sophisticated voice cloning technology can now get past this.
- Multifactor authentication. This requires customers to provide two or more forms of identity verification, such as a personal identification number and answers to security questions. This creates more friction for fraudsters, but knowledge-based authentication (KBA) is becoming less effective. 58% of contact center leaders surveyed by TransUnion said the use of stolen customer information to pass KBA has become more common in the past year.
- Agent training. Agents may receive ongoing training on signs of fraud, common scams, and steps to take if they suspect they’re speaking to a bad actor.
- Call reviews/quality assurance. Contact centers may use manual call reviews and quality assurance to look for fraud attempts and missed verification steps. However, contact centers can typically review less than 5% of their interactions through manual QA, meaning they’re likely missing the vast majority of fraud attempts.
The fundamental challenge is that fraud is evolving too fast for traditional defenses to keep up with. Fraudsters continuously refine their scripts, test new social engineering tactics, and leverage AI tools to create more convincing personas. Contact centers need defenses that can evolve at the same pace—which is where conversation intelligence and agent assist technology come into play.
How to use Creovai Conversation Intelligence + Agent Assist to prevent fraud
Conversation intelligence technology uses machine learning and generative AI to analyze the content of contact center calls. It can analyze large data sets (i.e., all of a contact center’s call transcripts) to identify suspicious language patterns at scale. This might include behaviors such as:
- Asking the agent to bypass verification methods
- Struggling to answer knowledge-based authentication questions
- Using overly formal phrasing (something that can be a sign of voice AI)
- Using pressure tactics to try to get customer information from the agent
- Using the same scam scripts on multiple calls
Creovai Conversation Intelligence lets you build these suspicious language patterns into insight categories. For example, if your contact center has been getting hit by a recent scam in which a fraudster uses threats of taking legal action to pressure an agent into issuing a refund, you could create a “Known Scam: Legal Threat” category containing collections of phrases commonly used in the fraud attempt.
.png){kind=spacer}
These insight categories can help you detect fraud patterns in post-call analysis, but you can also use them to inform live prompts in Agent Assist, Creovai’s real-time agent guidance product.
When Agent Assist detects a fraud category in a live conversation, it can trigger a pop-up alert on the agent’s screen, letting them know they’re dealing with potential fraud. The alert can include coaching guidance, such as reminders to take additional verification steps, scripts for neutral language to end suspicious calls, or instructions to escalate to specialized fraud teams.
By combining Conversation Intelligence and Agent Assist, you can ensure your fraud categories evolve as new threats emerge. Conversation Intelligence continues monitoring your conversations and identifying new patterns, which can then be built into categories that are used to trigger Agent Assist prompts.
Combining Conversation Intelligence with Agent Assist provides an additional layer of security and support. Even if your agents receive regular training on detecting and managing fraud, it’s possible for someone to have a bad day or get flustered by an emotional plea and fail to identify fraud. Providing your agents with real-time guidance—informed by past interactions—removes human fallibility from the equation and ensures every agent has the tools to prevent fraud.
This integrated approach also improves operational efficiency. Agents get the information they need to complete additional verification steps or escalate the call quickly, meaning they don’t lose time wondering what to do next. Additionally, when agents are consistently able to prevent fraud in the moment, they discourage bad actors from repeatedly targeting your contact center. One of our customers, Thrasio, noticed a change in fraud volume after implementing Creovai. Miranda Grigar, Customer Obsession Manager, told us:
“We believe that bad actors previously targeting us have greatly reduced their efforts after seeing their attempts have been unsuccessful.”
Staying ahead of voice fraud threats
Voice fraud in the contact center isn’t new, but it’s becoming an increasingly daunting threat as technology and tactics evolve. The good news is that the technology and tactics to prevent fraud continue to evolve as well.
AI-powered conversation intelligence can identify subtle fraud indicators that human reviewers would miss in high-volume environments. Real-time agent guidance ensures that insights translate into immediate protection during live interactions. And bringing the two technologies together enables your contact center to adapt to new fraud tactics without making manual updates to static rules.
The contact centers that will best protect themselves—and their customers—are those that embrace technology capable of large-scale monitoring and real-time response. As fraudsters become more sophisticated, your defenses must become more intelligent.
FAQs
How do we balance fraud prevention with maintaining acceptable average handle times?
Implement real-time agent assistance technology that provides instant fraud alerts and pre-scripted verification steps. This eliminates the time agents spend deciding whether a call is suspicious and provides clear next steps. Additionally, consider adjusting performance metrics to account for necessary security protocols—a slightly longer handle time is far less costly than successful fraud attempts.
What's the most cost-effective way to scale fraud detection beyond our current 2-5% manual call review capacity?
Deploy conversation intelligence technology that can analyze 100% of your call transcripts for suspicious patterns automatically. This AI-driven approach identifies fraud indicators across your entire call volume without requiring you to increase your QA headcount. The technology pays for itself by catching fraud attempts that would otherwise slip through limited manual reviews.
What should we do when traditional verification methods like voice biometrics and knowledge-based authentication are becoming less reliable?
Implement layered detection that goes beyond traditional authentication to analyze conversation patterns, language use, and behavioral indicators. Conversation intelligence can identify social engineering tactics, pressure language, and script patterns that remain consistent even when fraudsters bypass biometric or KBA systems. Combine this with dynamic verification protocols that adapt based on detected risk levels.
What ROI should we expect from investing in AI-powered fraud prevention technology?
Calculate ROI based on prevented losses rather than just cost savings. With deepfake fraud averaging $450,000-$600,000 in losses per incident, preventing just 1-2 successful fraud attempts per year typically justifies the technology investment. Additional benefits include reduced operational costs from fewer lengthy investigations, improved customer trust retention, and decreased regulatory scrutiny—though these are harder to quantify, they significantly impact long-term profitability.
.png)
