Ctrl+CX Issue 02
Hi, I'm Rob Dwyer 👋
I cut my CX teeth on the phones in the retention department for one of the Big 3 telecom companies. I know the feeling of "Of course they picked that call to listen to." I know how it feels to get an auto fail. I know we ask frontline supervisors to drive performance while rarely giving them the skills or the data to do it.
And I know that "This call may be used for quality and training purposes" was almost always false. The chances of anyone listening to that call sat somewhere between slim and none.
I've spent the past six months at Level AI supporting enterprise CX organizations. I've completed implementations for Fortune 500 customers, supported enterprise teams on enablement and adoption day to day, and met practitioners using our platform at our first annual Level UP conference.
Ctrl+CX is my shortcut to what's happening in customer experience vs. what this industry keeps repeating.
The Ctrl+CX View
“Customer-facing AI is all about killing jobs”
There are very real impacts on the workforce due to AI. Just two years ago, Klarna made headlines by eliminating hundreds of customer support jobs in favor of AI chatbots only to make headlines again a year later with a (partial) walk back. I know at least one person who was personally impacted by this move. As it turns out, people still want to talk to people sometimes. But in those two years, chatbots and voice bots have made incredible gains in their ability to support all kinds of customer needs. And voice is cannibalizing a lot of chat interactions.
Virtual Agents alleviate three universal constraints that contact centers have struggled with for decades.
Contact Center hiring managers and Workforce managers have nightmares fueled by the need for an agent that speaks fluent Danish, staffing overnights and weekends, and when the queue gets so deep the SIP trunk simply stops accepting calls.
What I’ve witnessed when we deployed Virtual Agents are real results that weren’t based on “let’s cut headcount” motives, but instead were based on “how do we support customers better?” motives. Those Virtual Agents always have a built-in handoff mechanism to a real human.
7 Critical Steps to Implementing Customer-Facing AI Agents
1. Start with Why
I’m no Simon Sinek and this isn’t a Ted Talk, but any decision to change how you approach interacting with customers should have a well-defined reason. That reason will guide what AI should help with and both when and how you should involve human support.
2. Specific is terrific!
A well-defined scope enables you to deploy an agent with appropriately configured guardrails. AI agents, just like people, can make assumptions. You need to be very explicit about what it is and is not allowed to do.
3. Test your knowledge!
More specifically, dig into all the relevant KB (Knowledge Base) resources and make sure they are up to date, in alignment, and match your brand’s tone. Your AI agent will only be as good as the knowledge it draws from.
Recommended by LinkedIn
4. Take the blue pill.
Simulations of real conversations are the best way to test the performance of your AI agent before it ever talks with a customer.
5. Hand the ball off.
Your AI agent needs to be prepared for a quick handoff to a person whether it’s an out of scope request or simply customer preference. That handoff ideally includes any relevant contextual information like the steps taken with the AI agent. This enables the human agent to seamlessly pick up the conversation and move to resolution faster and with less effort from the customer.
6. QA, QA, QA.
Evaluating performance isn’t just for people. AI Agents should get the same treatment. And don’t stop at what the agent said - make sure you analyze what it did and why. An AI agent who sounds great with customers but is making rogue API calls isn’t improving customer service.
7. Loop and Learn!
Ideally, your AI agent is learning from its mistakes just like people do. Build a learning loop because your business and your customers' needs are always evolving.
Where to find me
TODAY!! May 7 The 3-Step Guide to Configure, Control and Evaluate Your AI Support Agent Free webinar
May 21 XMGlobal360 Conference Free virtual conference
Jun 16–17 Forrester CX Forum East New York, NY
Jun 22–25 CCW Caesar's Forum, Booth #145, Las Vegas, NV
Listen
The Brief Case Podcast hosted by Vadi K. - AI Is Your Agent’s Secret Weapon
CX Innovators Podcast – Emmett "Wes" Dudley and I discuss a CX Transformation that eliminated 2.7 million calls almost overnight
Next in Queue – Jeff Toister and I discuss Human Service and when to Automate
Level AI pairs a proprietary AI stack with quality, coaching, automation, and analytics running on the same conversation data. Coaching plans reflect what QA scores surface. Product teams see the same friction patterns operations sees. Founded by former Amazon Alexa product leader Ashish Nagar and backed by Adams Street Partners, Battery Ventures, Eniac Ventures, and others. thelevel.ai
Thank you Rob Dwyer! Really enjoying this. Clear ideas, but very impactful.
Really appreciate how grounded this is - especially the emphasis on knowledge quality, handoffs, and QA instead of treating AI like a plug-and-play solution. A lot of the strongest AI conversations I’m seeing right now aren’t really about the model itself. They’re about the operational system around it: clear scope, good knowledge, defined escalation paths, and continuous learning loops. Otherwise teams just end up automating confusion faster. The point about AI agents needing QA too is such an important one.
Excited to hear you discuss how to QA these Virtual Agents
Great way to portray the CX great share #CFBR
This hits so much closer to reality than the “AI will replace everyone” headlines. When you start with scope, handoffs, and QA, virtual agents become force multipliers, not headcount hacks. Love the focus on why and on treating AI like part of the CX team, not a shortcut.