Organisations today are unknowingly carrying a potential burden of invisible risk. Most chats, calls, emails and social media interactions are not monitored or evaluated.
Erik van Eekelen, Founder and CEO of Icana.AI, advises, “Traditional Quality Assurance (QA) processes typically analyse only a very small fraction of customer interactions, often less than 2%, which means the vast majority of conversations go unseen. Within those unreviewed interactions are missed compliance risks, inconsistent customer experiences and lost opportunities to improve performance”.
While organisations have successfully deployed automated agents to handle surging ticket volumes, the infrastructure required to oversee these digital workers—and the humans who manage the complex cases left behind—has failed to keep pace. According to research from Solidroad, a staggering 81% of customer interactions go entirely unreviewed, creating a vast reservoir of ‘invisible risk’ that threatens brand reputation and regulatory compliance.
The barriers to Quality Assurance (QA) are no longer about a lack of data, but rather a fragmentation of insight. While the transition from manual to automated QA has solved the problem of scale, it has created new hurdles in depth and actionability. Traditional quality assurance tools were designed for a static world of manual call monitoring and small-scale human teams.
Mark Hughes, CEO and co-founder of Solidroad says, “The biggest issue right now is that QA tools simply haven’t evolved to match modern support workflows. Most existing systems were built for a world where human agents handled every interaction, and team leads could manually review a small sample of calls each week. But today, AI agents are managing thousands of conversations, while humans are handling the more complicated requests”.
“This combination means more conversations are being reviewed, but QA tools aren’t ready to handle this volume[RS1] . In turn, human teams are left trying to bridge the gap. They’re strapped for time and forced to rely on limited sampling and manual reviews, which makes it nearly impossible to get a clear read on conversation quality. Agents say that the majority of their conversations are never reviewed for quality, leaving a massive blind spot in QA”, adds Hughes.
Quality versus speed – a leadership pivot
The pressure to maintain performance and service level is often complicated by a historical obsession with speed. Metrics like Average Handle Time (AHT) have long been the industry standard because they are easy to track, yet they frequently incentivise the wrong behaviours. “A common challenge in contact centres and customer-facing teams is when performance metrics prioritise speed (e.g., average handling time, calls per hour) over quality. This results in agents feeling pressured to rush interactions and can compromise effective service”, says Van Eekelen.
Hughes elaborates, “Speed often becomes the default metric because it’s easy to track, but it’s a poor reflection of what good support actually looks like. When speed becomes the primary way agents are evaluated it results in agents prioritising closing tickets quickly, sometimes before fully resolving the issue. That’s where companies start to see confidence drop, as the focus shifts from doing the job well to doing it quickly”.
Leadership must pivot toward quality-centric metrics that value accuracy, policy alignment, and soft skills like empathy. By investing in comprehensive visibility, companies can actually improve efficiency, as Hughes points out, “The teams that successfully prioritise quality and wait times do two things. They measure quality across a broader set of interactions so speed isn’t the only signal, and they invest in coaching that reflects real scenarios, giving agents better preparation for live conversations. This approach keeps wait times under control because confident, well-trained agents handle issues more efficiently”.
The Customer Satisfaction (CSAT) Distortion
Even traditional satisfaction surveys are under fire for being unrepresentative. They tend to capture only the loudest voices, leaving a massive data gap in the middle. “CSAT scores are also overdue for a serious makeover. They are often treated as a proxy for customer experience, but in reality they mostly capture the views of customers who feel strongly enough to respond… Everyone else is largely invisible. That creates a distorted picture”, says Van Eekelen.
“The better alternative is to assess every conversation by analysing customer sentiment and emotional trajectory across the full interaction, which gives a more representative view of performance, including the neutral and slightly positive or negative experiences that CSAT tends to miss. Technology is now advanced enough to automate this process”.
Human agents value feedback
While 79% of support professionals actively value feedback, according to Solidroad’s research, the constant pressure of the clock often robs them of the guidance they need to grow. When agents feel their work is judged by ticket counts rather than the value they provide, burnout becomes inevitable. Van Eekelen asserts, “When agents are measured primarily on speed but know they’re being judged on quality, it creates tension. They don’t have the time or support to improve, yet they’re still expected to perform at a high level. Without clear, consistent and timely feedback and support agents can become frustrated and disengaged, which contributes to higher turnover.
Hughes adds, “When agents feel stuck, disengaged, or unsupported, turnover naturally increases. The solution is better visibility into every interaction. When feedback is consistent, actionable, and tied to what actually drives quality, agents can improve, gain confidence, and see that their work matters. That’s how companies can reduce burnout and keep their teams engaged as support scales”.
Coaching and onboarding
Manual QA is fundamentally limited by time, as it’s impossible to analyse every interaction by hand. Unfortunately, technology has also lagged behind. Van Eekelen comments, “Many solutions focus on transcription or basic sentiment analysis, which only captures a small part of what’s happening in a conversation. They often miss the nuances of delivery, the emotions on both sides of the conversation, and don’t provide coaching. As a result, even when organisations invest in automated QA, they’re not getting the depth of insight needed to drive targeted upskilling and meaningful improvement”.
Hughes believes, “Coaching is the most effective form of feedback, but it’s also the hardest to scale. On large teams, managers can’t consistently provide that level of guidance, leaving agents without the support they need to improve.
“The solution is objective, actionable feedback tied to real conversations. Each conversation can be evaluated against a consistent rubric, and feedback can include examples from the agent’s own work along with exercises or simulations to practice. By linking coaching directly to real experiences, the human touch is preserved even in automated or asynchronous formats, helping agents apply lessons immediately and integrate learning into daily workflows”.
Accelerating proficiency through continuous learning
Building a successful sales or support team requires embedding learning into the daily workflow rather than confining it to an initial training period. Van Eekelen says, “Instead of treating onboarding as a one-off event, organisations can create continuous feedback loops where agents receive regular insights based on real interactions. Combining this with scenario-based practice allows agents to develop skills in context. This approach not only accelerates onboarding but also ensures that learning continues as customer needs and expectations evolve”.
The impact of modernising this process is significant. Van Eekelen highlights how a large education organisation recently accelerated staff proficiency by more than 30% simply by leveraging AI tooling that provided high-quality feedback on every single call. This approach proves that when learning is constant and data-driven, teams not only onboard faster but maintain a higher standard of performance long-term.
How AI can help
AI simulations can act as the flight simulator for customer service, providing a safe but realistic environment to bridge the gap between theory and live interactions. Rather than throwing new hires into high-stakes calls, these technologies allow agents to fail, learn, and iterate in a controlled setting.
Hughes comments, “You can teach policies and product knowledge in a classroom, but agents aren’t fully prepared until they’ve handled a frustrated customer in real time. AI-powered simulations let agents practice realistic scenarios that adapt dynamically to their responses, helping them build the judgment and confidence that old-school training just can’t provide”.
Simulations allow agents to handle complex, emotionally charged, or rare scenarios without the risk of losing a real customer or damaging the brand. This builds ‘muscle memory’ for compliance and soft skills. Hughes affirms, “The feedback loop is key: agents try a scenario, get immediate guidance, and repeat. Mistakes happen in a safe environment, so learning is accelerated without risking real customers. This closes the gap between training and live support, ensuring agents are confident and capable when it counts”.
Rather than generic training, automated, AI enabled coaching can use an agent’s own data to build personalised practice. The AI uses natural language to roleplay, allowing the agent to practice their own voice and style in a low-stakes environment before the next live call. “AI and simulations can play a critical role by creating realistic, safe environments where agents can practise and refine their skills. When those simulations are combined with the same analysis used for real interactions, it creates a strong link between training and performance. This helps bridge the gap between theory and practice and accelerates skill development”, says Van Eekelen.
The future of customer support lies in moving beyond the invisible risk of unreviewed interactions and the narrow constraints of speed-based metrics. By transitioning from a historical obsession with efficiency to a focus on comprehensive visibility and emotional intelligence, leadership can eliminate the blind spots that threaten brand reputation.