Silent Churn Is the Biggest Customer Support Risk

Most customer support organizations are built to serve the loudest 10% of users — the ones who open tickets. The greater risk lies with the silent majority who encounter friction, say nothing, and quietly churn.

As artificial intelligence moves from experimentation to deployment, a new class of agentic AI is emerging to address this hidden revenue leak — not by replacing humans, but by resolving problems before customers decide to leave.

At the forefront of this shift is Nadav Kemper, CEO and co-founder of Quack, a platform that does more than build chatbots — it trains agentic AI. His perspective reflects a growing augmentation-first approach that reframes AI not as a replacement for human agents, but as a tool for protecting customer relationships and revenue.

Kemper argues that most organizations are solving the wrong problem. Traditional support models focus on the 10% of customers who open tickets, while ignoring the "silent 90%" — users who encounter friction, never complain, and ultimately leave.

From Cost Center to Revenue Protection

Addressing this issue is not just about efficiency. It is about protecting revenue that quietly slips away unnoticed.

"Customer support has been stuck in reactive mode for too long," Kemper notes. "It’s costly to run, painful to scale, and a frustrating experience for customers."

He believes organizations can finally scale without losing the human touch by deploying AI agents with true agency — systems that monitor product signals and resolve issues before they escalate into churn.

For Kemper, agentic AI is what makes a white-glove support experience possible at scale, preserving the human touch while removing the friction that traditionally limits growth.

In his view, AI is not intended to replace human agents' jobs. Instead, the AI innovation lets workers clear their desks of repetitive, low-value noise, so they can focus on the complex, high-stakes interactions that actually build brand loyalty.

"Our platform proactively detects signals and patterns in customer behavior by continuously monitoring indicators of user frustration, such as prolonged hovering, abandoned carts, excessive clicking on certain buttons, and more – even before customers open support tickets themselves," he told CRM Buyer.

That shift — from reactive ticket handling to proactive resolution — is beginning to show up in how organizations deploy and staff around AI in customer support.

In December, Gartner reported that AI adoption in customer service is rarely about workforce reduction. Only 20% of service leaders indicated they had reduced headcount after deploying AI, while 42% reported creating new roles to support AI operations — reinforcing a move toward augmentation rather than replacement.

Why Silent Churn Is the Real Support Risk

I spoke with Nadav Kemper about how agentic AI is redefining support ROI, exposing silent churn, and changing how support teams are structured as AI becomes operational.

CRM Buyer: How does Quack's agentic AI platform listen for these silent signals before a customer decides to leave?

Nadav Kemper: Rather than optimizing solely for reactive metrics like First Time Resolution (FTR) or Time to Resolution (TTR), we focus on expedited, proactive outcomes. We measure this through a Proactive Resolution Rate, which reflects the percentage of relevant signals our AI agents can identify and address before issues escalate or lead to churn.

Traditionally, CS has been a cost center that responds to problems. How does the shift to proactive AI agents change the ROI conversation for a CFO?

Kemper: Most companies still invest in deflection without actually resolving their customers' issues. As enterprises scale, increased customer demand strains the human support teams that sustain that cycle.

By adopting a proactive AI agent solution, agents can manage high-volume, predictable interactions and resolve customer issues before a ticket is ever opened, all while empowering human CX teams to focus on higher-impact, more complex issues. This way, AI agents can reduce churn, helping companies retain and grow their customer base while scaling sustainably.

How does Quack determine the accurate metric?

Kemper: In this case, ROI should not be measured as cost per ticket, but rather by the measurable impact per interaction. We evaluate ROI using resolution accuracy, churn reduction, and automation efficiency, which we consider the most effective KPIs for scaling.

Gartner found that 42% of organizations create new roles due to AI. From what you see at Quack, what do these new AI-adjacent human roles actually look like on a day-to-day basis?

Kemper: We see support teams building a small “AI ops layer” around their organization. These typical roles we see are analytical, product-oriented, and focused on running the AI as part of daily operations:

• AI Trainer: Keeps briefs, topics, and fallbacks up to date. Think “editor-in-chief” for the AI.
• AI QA Lead: Uses Scorecards to catch hallucinations, tone issues, and gaps.
• AI Ops Analyst: Lives in Explore and Trackers to surface patterns, bugs, and missing knowledge.
• Workflow / Skill Designer: Turns real processes into agentic skills and escalation rules.

How should CS leaders retrain their staff to handle a 100% high-complexity and high-emotion workload?

Kemper: Primarily focus on developing agents’ emotional intelligence (EQ) and strong product knowledge, not simply speeding up resolution times. High-impact support requires agents to master communication skills and understand when to rely on AI and when to take matters into their own hands.

Metrics should shift from handle time to ownership, recovery, and quality, ensuring agents are accountable for customer outcomes. Additionally, agents should proactively focus on follow-ups, churn prevention, and feeding customer insights back into training to continually improve AI and support systems.

What advice can you offer about AI being a partner rather than a replacement?

Kemper: The future model of CS is better understood as collaboration rather than replacement. Agentic AI addresses predictable, repetitive issues by playing a preventive role – identifying friction points, monitoring signals, and triggering proactive outreach before problems escalate. This allows human agents to become product experts and focus on areas where nuance, judgment, empathy, and relationship-building matter most.

In practice, support agents don’t simply work alongside AI. They increasingly shape it by guiding its behavior, quality, tone, and consistency. Rather than removing roles, agentic AI shifts the support function upward, elevating agents into positions with greater responsibility, influence, and impact on the overall customer experience.

How is the distinction between chatbots and agentic AI in Quack critical to improving customer loyalty?

Kemper: Traditional chatbots are clunky, rigid, and limited in their ability to handle complex customer needs. They wait for questions or inputs and try to guess answers. Our agentic AI is built to act.

At Quack, agency means our AI can proactively resolve issues, adapt to interactions, understand customer intent, take autonomous action, and confirm end-to-end resolution. Customers don’t feel they are interacting with a “chatbot” because we have built a holistic operational support system. That is what drives customer loyalty.

How does a trainable AI platform allow a startup to maintain a white-glove experience rather than impair CX as they grow?

Kemper: Quack’s platform trains AI agents through a process similar to onboarding a human employee, but with greater precision, ease, and scalability. Our AI mirrors the exact tone, language, and quality of existing support teams, continuously improving past interactions with real customers through internal QA.

Companies can train a model and deploy it across unlimited agents, effectively scaling their best human support agent. This approach increases ticket resolution rates and frees human teams to handle escalated or complex issues while maintaining the highest quality standards.

Is there a risk that the AI will learn wrong behaviors, and how do we keep a "human-in-the-loop" for quality control?

Kemper: Most existing CS solutions either ignore QA altogether or only review a small portion of interactions, leaving blind spots that erode service quality and future performance.

We’ve seen the need for this firsthand, which is why our team developed AutoQA to work alongside human support teams, monitoring interactions, feeding insights back into the training model, and tailoring the customer experience to each organization's preferences.

Do you envision a world where Customer Support as a department no longer exists, replaced by a hybrid of Customer Success & AI Operations?

Kemper: I don’t see CS disappearing, but I do see it evolving. As AI scales, speeds up, and gains predictive capabilities, CX becomes less about handling reactive issues and more about creating value across the customer lifecycle.

What emerges is indeed a hybrid model that combines human teams focused on customer success, relationships, and judgment, supported by AI that proactively handles automated, signal-driven resolution at scale. In that sense, CS doesn’t vanish. It matures into a higher-tier function with greater strategic importance.