This is part of The Journal, a series highlighting the perspectives of high performers in CX, customer service, and technology. Our goal is to capture authentic insights from leaders shaping the future of the industry—in their own words.
I've been in the contact center industry long enough to say something that might sound controversial: quality assurance, as we know it, isn't even needed anymore.
Let me be clear—quality still matters. But the way we've been measuring it—with forms, checkboxes, and sampling a tiny fraction of calls—is fundamentally broken. In a world where AI can analyze 100% of conversations and surface insights beyond what any human could program into a scorecard, hanging onto traditional QA isn't just inefficient. It's limiting your potential.
In this piece, I'm going to break down the biggest misconceptions contact center leaders have about QA, explain why auto-QA is just a bridge to something better, and share why the future belongs to organizations that stop measuring conversations and start understanding them.
The Core Problem: QA Isn't Even Needed Anymore
Here's my reasoning: traditional QA was built around limitations. You couldn't listen to every call, so you had to sample. You couldn't analyze nuance at scale, so you created forms. Those constraints made sense when they were the only option. But they're not anymore.
Today, AI can analyze all conversations. It can pick up reasons for conversations, analyze what went well and what went bad, and recommend what needs to be done. Auto QA and AI-based QA that just automates your old forms? Honestly, that's already becoming an old concept too.
The real shift happening here is from measurement to understanding. Instead of asking, "Did the agent check five boxes?" the question becomes, "What actually happened in this conversation, and what should we do about it?"
"When you force yourself to do QA with preset forms, you're limiting yourself to what you're looking for. What about everything else that happened in the conversation?"
Traditional QA forces you to decide upfront what matters. But real conversations are messy. Customers raise unexpected issues. Agents handle problems in creative ways. Policy gaps surface. If your QA form only asks five questions, you're only capturing five data points—and missing the full story.
You're essentially saying, "I want these five questions answered. What about everything else that happened?" You're not even capturing it, not understanding it, and certainly not doing anything with it.
That's the fundamental flaw: traditional QA optimizes for consistency at the expense of insight.
Reimagining What QA Could Be
If traditional QA is dead, what replaces it?
The future isn't about abandoning structure entirely—it's about rethinking what you measure and why. I still think there's value in focusing on the very specific things you need for your business. But for things like customer experience, AI is significantly better at understanding what's good, what went well, what didn't go well, and what could have been done better—without needing five checkbox questions like "Did they show empathy? Did they show rapport?"
Two Types of Analysis
I draw a clear distinction between two use cases:
1. Compliance and Metrics: These are the specific, recurring things you must track—regulatory disclosures, required behaviors, or key performance indicators that need to be measured week over week. The use case for auto QA on a continuing basis is to look for these very specific things so you can measure them consistently and turn them into actionable metrics. This is really more about compliance than quality assurance.
2. Experience and Understanding: This is where AI truly shines. Instead of scoring agents on predefined criteria, AI can evaluate the entire conversation—tone, resolution quality, customer effort, policy friction—and surface insights you never thought to look for. Customer experience is a free-flowing conversation, and understanding that holistically is a lot easier with AI than trying to capture it on a form.
Stop trying to make AI fit into your old QA forms. Instead, train AI on what great conversations look like in your business—and what bad ones look like—then let it identify patterns you never programmed it to find.
Find the positive conversations that you really love—the ones that represent your ideal customer interaction. Let AI understand those and identify similar patterns across all your conversations. Find the negative conversations where things didn't go well, so AI can learn what failure looks like in your context. Train the AI on your existing experiences to understand what to look for, rather than being overly tactical with a fixed set of 10 questions. There are way more things happening in every conversation than any form can capture.
Auto QA: The Necessary Middle Step
If AI is so powerful, why does auto QA still exist?
Because it's hard for people to go from checking five specific things to starting with a blank slate and letting AI tell them what's happening. That's a big leap.
Auto QA serves as a bridge. It takes your existing QA forms and automates the scoring process—removing the manual bottleneck, increasing coverage, and building trust in AI's ability to interpret conversations accurately. It can review way more calls for those 10 questions than your human analysts ever could. That's a very basic phase one of what AI can actually do in the long term.
But it's just phase one.
There needs to be a middle step. You can't expect organizations to jump from "I must track these five specific things" to "I'm trusting AI to figure it out completely." Auto QA helps you move from limited human sampling to comprehensive AI coverage, which builds confidence. Once you see that AI understands what's going on, you can graduate to the final step—AI-driven conversation intelligence and customer experience management, where the system discovers insights you never thought to look for.
"Phase one is moving from human-based QA to AI-based QA. But the final transformation is moving beyond traditional QA entirely—to true conversation analytics."
The AI to Rule All AIs
One of the biggest challenges in modern contact centers isn't a lack of AI—it's too much AI.
Every tool today claims to be "AI-powered." But in most organizations, AI lives in silos. You've got AI in your survey platform, AI in your CRM, AI in your QA tool—and none of them talk to each other.
The result? Lots of data, but no coherent story.
The good and the bad of AI are the same thing. The good is it has a lot of information. The bad is it has a lot of information. What you do with all that information becomes the real challenge.
My vision for the future is what I call "an AI to rule all AIs"—a unified intelligence layer that connects data from quality, CRM, surveys, workforce management, and more, then surfaces insights that no single system could see on its own.
You need an AI that sits on top of all of them to connect all the dots together. All these other tools have AI built into them, but they're limited to their own data sets. You need an AI that can gather data from all the different sources and then route it to the proper teams and organizations within the company who can actually act on it.
This isn't just about aggregating reports. It's about understanding why things are happening and who needs to act on them.
AI can analyze what went well and what went bad. It can recommend what needs to be done. But then you need another layer of AI overseeing all that data to route it properly—give the right teams the right reporting, insights, and analytics they need, and then track their actions to see if they actually improve anything.
How to Evaluate AI QA Vendors
If you're shopping for an AI-powered QA or conversation intelligence platform, here's my advice: don't evaluate vendors on the easy stuff.
If you really look at LLMs and technologies you probably already have access to inside your company, you can generate this information yourself today. You could download your calls, send them to ChatGPT, ask it questions, and boom—you have your answers.
The hard part isn't generating insights. It's turning them into improvement.
Running an LLM against a set of questions and a transcript—that's really not rocket science anymore. You could hire your kids to do it. The evaluation of vendors shouldn't be based on "Can they actually do that?" because honestly, anybody can do that at this point.
"The real evaluation should be: Can they help you implement it? Do they have experience turning data into actions that actually improve business results?"
Generating data isn't a task in itself. You're ultimately doing all this to improve something—reduce costs, improve customer experience, increase sales, reduce churn—whatever your goal is.
All of it requires analytics. All of it requires action. The data is really just the beginning of the journey, not the journey itself.
When you're evaluating vendors, don't ask, "Can they analyze my calls?" Ask, "Can they help me improve what happens after?"
It's Not Just Customer Calls—It's Every Conversation
Here's something people don't think about enough: this technology isn't just for customer interactions.
There's no reason these types of analysis can't be applied to other conversations. Almost every organization is already using it for meetings—summarization of meeting notes, action items, all that stuff. But there's no reason that kind of technology can't be used for everything. Every conversation you have. Group meetings, one-on-one meetings, customer conversations, coaching sessions—no matter what conversation you're having, there's valuable data that can be generated from it.
Think about coaching conversations between managers and agents. Peer collaboration sessions. Internal project meetings. All of those contain insights that, if analyzed properly, could improve how teams communicate and perform.
I joke that most corporate meetings are a waste of time—but AI could at least prove it. More importantly, it can show you why some conversations drive clarity and action while others don't.
This broader vision—using conversation intelligence across the entire organization—hints at where the technology is really headed: not just better contact centers, but better organizations.
Final Thoughts: From Measurement to Understanding
Here's my central argument, and it's simple but profound: stop thinking about quality assurance as a measurement exercise and start thinking about it as an understanding exercise.
Traditional QA was built for a world of constraints—limited capacity, manual processes, narrow definitions of quality. AI removes those constraints. It can analyze 100% of conversations. It can spot patterns you didn't program it to find. It can connect insights across systems and recommend actions that actually drive improvement.
The question isn't whether AI will replace QA. It's whether your organization is ready to move beyond forms and checkboxes and start asking better questions: What's really happening in our conversations? Why? And what should we do about it?
The companies that win won't be the ones doing more QA. They'll be the ones using AI to learn from every conversation—to connect dots across systems, coach people faster, and continuously improve. Not just to measure quality—but to actually make it better.