AI Opportunity:
Influencing FCR remains the most critical contact center metric: for every 1% improvement in FCR, there is a 1% improvement in CSAT*. However, programmatically influencing FCR has always been a challenge due to technological limitations (in making agents smarter during conversations): legacy technology has been using pattern matching (based on keywords and phrases) to detect caller sentiment (which we are now realizing does not yield high accuracy levels). Additionally, historical data models built to analyze agent calls were not designed to capture voice attributes beyond call time, call duration, speaking time (and at most, the ‘inferred’ caller sentiment). Today. with Natural Language Processing (NLP) and Deep Learning speech-to-text algorithms approaching >95% accuracy, the ability to accurately transcribe and understand the different components associated with a call (humor, sarcasm, subs-conscious bias, etc) has never been this comprehensive in human history. We can now detect and store several dozen dimensions associated with agent calls (caller’s gender / emotions / demographics / prior interactions, agent’s level of empathy / courtesy / positive energy language / adherence in reading terms & conditions, etc.). If we leverage AI on this dataset along with relevant customer data from different customer transaction systems, we can develop a comprehensive view of the customer’s prior interactions, and proactively assess the customer’s issue. This enables us to intelligently segment callers and match them with the most appropriate agent in real-time, improving FCR rate. Agents can spend less time finding information about the historical conversations with the customer and more time presenting solutions. Very soon we will start seeing deep learning algorithms, applied on several hundred terabytes of voice data, get trained so well that agent scripts can adapt in real-time and improve not just agent efficiency in resolving a problem, but also pick up opportunities to guide an agent to effectively upsell/cross-sell. Supervisors can spend less time analyzing (call/customer interaction reports, figuring out agent scheduling, which agents are adopting best practices) and more time in strategic planning, in agent coaching, and potentially driving revenue.
There are 18M+ customer service agents worldwide. For every 6 of these customer service agents, there is generally 1 QA manager who maually audits calls. As a result, only 1-2% of agent calls are audited by QA managers. Making business decisions on such a statistically insignificant data set is dangerous — at these audit levels agents cannot be evaluated correctly (if they are, I don’t blame them for leaving on the basis of being mis-evaluated). In many industries, particularly in financial services, regulations require a significantly higher % audit rate. It’s even more important in financial services to ensure agents are adhering to best practices (reading disclosures, terms & conditions, etc.); supervisors need to be alerted whenever agents are not in compliance and that feedback needs to be given in real-time. Today, call audit coverage is increased by adding more QA managers, who listen to a random selection of calls and provide feedback several days/weeks later. With AI-based solutions, we can now get to 100% audit coverage at a fraction of the cost, enabling QA managers to spend less time listening to a sample set of calls and more time improving agent performance. Managers can provide focused feedback on specific skills/topics, highlighting where there is any deviation from best practices – in real-time. Agent coaching will soon reach a new level: tailored to complexity of problem, level of empathy exhibited, historical choice of words; surfacing the exact voice recording clip for each scenario.
Separately, organizations are facing challenges around keeping a tab on all the social conversations happening about their business/product set, particularly with the emergence of new communication tools & multiple social media channels. It’s increasingly complex and challenging for communications/social media managers/customer service agents to stay updated on discussions happening about their company on Twitter, LinkedIn, Facebook, Instagram, Reddit, WhatsApp, Telegram, etc. Instead of spending time reading and responding via each channel we are now starting to see AI-powered tools that unify all digital channels, have the optionality to generate trending themes/discussions/customer sentiment, and programmatically respond across channels. This enables customer experience teams to be more strategic and proactive in driving their company product narrative, more aware of customer sentiment versus being reactive in consuming social messaging data.
