Get It Right for Your Agents: 5 Recommended Steps to Success Regardless of How Quickly the GenAI Market Evolves and Just Where It Goes

The Vital Evolution of Customer Service Agents in a GenerativeAI Automated Age’ - Chapter 3

Chapter 2 of this article series captured four agent-specific risks and touched on others outside that scope.  With diametric pros and cons, companies need a strategically sound yet risk-averse action plan.

Recommended Next Steps: Be Smart and Start Moving Forward at the Right Pace

Integrating new GenAI systems or agent tools in a sector strapped for time and resources seems like an impossible challenge to some that can simply wait a week.  It cannot, so the question is how to proceed at the right timing and level that fits your model and can be justified to management, who will ask for verification.  Quoting McKinsey, "Major disruptions are always painful, and the transition from a care paradigm dominated by human agents to one steered by AI technologies may be the biggest disruption in the history of customer service." 

So, take a deep breath!  Then what?

1. Proceed Forward (Get Going!): Initiate GenAI integration at the right level, as this is an irreversible trend that just cannot be ignored.  Implementation steps can be initiated and corrected, but a bigger mistake is ignoring the over-arching market trend that will continue to reshape customer service for years—simply waiting to see how the market morphs brings back memories of Blockbuster and Radio Shack.  Strategies and plans will vary dramatically for very valid reasons, but doing nothing is like endlessly circling the tower in a holding pattern.  The outcome is just not good.
2. Analyze Risk/Reward Scenarios First:  Before diving into GenAI customer service options, run a detailed analysis that examines the impact on agents, customers, resources, finances, ROI… everything.  Determine which are the least risky and present the best possible ROI, then consider regional testing first before adopting nationwide or globally.  Finally, data analytics, agent feedback, and customer feedback should be studied while in progress.  All are critical in determining what worked well, what needs improvement, or what options are not worth full-scale adoption.  This requires a system setup for a feedback input loop before starting trials.
3. Lather, Rinse, and Repeat: In other words, test, measure, and adjust to focus on what works, then continue with your next step.  And so on, and so on.  The chart below by McKinsey & Co. (March 2024) shows that many are still in the test mode, which is not necessarily bad.  I would say 'smart' if they are moving.  As the economy is in question, customer retention is simply critical. 

4. Anticipate Possible LLM Issues: Seek solutions that mitigate large language model (LLM) issues with Retrieval Augmented Generation (RAG). General LLMs are helpful but can lack specific domain knowledge, up-to-date information, or even occasional hallucinations.  General aging and outdating over time are also issues.  Rather than just using general LLMs, well-executed RAG, which takes time and effort, leverages specific vertical or business knowledge key to your customers and/or solution and integrates it.  Building that custom knowledge into the output increases accuracy and FCR, resolving issues before they occur.  Yes, it is easier said than done and takes time. However, well-tested (several times as general LLMs extrapolate when the facts aren't there!) and well-executed RAG using knowledge bases and APIs boost your chances to eventually shine.

5. Respect the Value of Your Agents: They are not a commodity for multiple reasons.

• The additional training time and GenAI usage are a challenge if your retention rate is already low.  Internally analyze all that's needed, from agent training to the reality of agent usage of GenAI in whatever form you adopt—map possible outcomes and implications just like customer journeys.
• Stepping back, if you transition to fewer agents, proceed in small increments.  If the system self-destructs, the time and money to acquire and train new ones or recall recent ones will hurt. 
• Plan ahead by leveraging the input of some agents you trust on what is needed initially to maximize a positive outcome.  Start with smaller trials to fix any issues first before scaling.
• Some functions (e.g., outbound calls) can adapt to GenAI integration more smoothly than inbound customer contacts.  

Several companies and service providers have taken that 'no agent' or 'virtually no agent' path.  In my opinion, and the industry is reinforcing more and more, it's risky.  In June 2024, Forrester VP and Research Director Rick Parrish said, "US consumers are having, on average, the worst experiences in a decade…".  For an example of 'smart balancing,' a Forbes May 2024 article focusing on successful automation still used 4-5 points of 15 on just how and why to manage agents to maximize customer retention.  One was "swiftly connect user chats to live agents when FAQs fail."  Another was "use dynamic learning to gauge user sentiment," quickly injecting a live agent when the' anger measure' runs too high.  There were more because the damage is irreversible once customers cross a particular line, and I'm also speaking first-hand. 

Certain mainstream businesses with limited service needs can consider this option, but many cannot afford the downside.  Note that SF-based CustServ growth firm UJET, whose CMO I quoted in my last article, just received $76M from multiple investors and uses a 'blended' solution to lower cost-per-contact while maintaining CX.  Generative AI and vital agent service can co-exist, creating synergies.

That's a Wrap: During the GenAI Morph, Value Your Agents Like They Are Your Customers.

We'll see a GenAI and Conversational AI maturation progress, albeit likely two steps forward, one back, and so on.  Intelligently map out the gradual progression that fits your solution, business, management expectations, and agents!  Mistakes and learning will occur.  A system designed in advance for absorbing input quickly and shifting direction as needed can excel regardless of the hurdles.  You will reach that pinnacle of automation, cost savings, NPS & CX, customer loyalty, and happier agents, all of which will reverberate in many ways.