Latest / Connectivity Without Chaos - The Podcast / How AI Is Quietly Solving the Real Customer Experience Crisis

Transcript

How AI Is Quietly Solving the Real Customer Experience Crisis · Connectivity Without Chaos - The Podcast

• 0:00
  SPEAKER_00: Welcome to Connectivity Without Chaos,
• 0:01
  SPEAKER_00: the podcast for IT leaders who are done managing networks by guesswork.
• 0:06
  SPEAKER_00: In each episode,
• 0:07
  SPEAKER_00: we break down the systems,
• 0:08
  SPEAKER_00: strategies,
• 0:09
  SPEAKER_00: and decisions that eliminate the hidden IT tax,
• 0:12
  SPEAKER_00: so your infrastructure becomes reliable, secure, and built to scale.
• 0:16
  SPEAKER_00: Today, we discuss ways AI is solving the customer experience crisis.
• 0:20
  SPEAKER_01: Yeah, and it is a massive crisis right now, honestly.
• 0:23
  SPEAKER_00: Oh, absolutely.
• 0:24
  SPEAKER_00: Yeah.
• 0:24
  SPEAKER_00: I mean, think about it.
• 0:25
  SPEAKER_00: Have you ever been stuck, you know, in a customer service queue for 45 minutes?
• 0:29
  SPEAKER_01: Ugh, far too many times.
• 0:31
  SPEAKER_00: Right.
• 0:31
  SPEAKER_00: Only to finally reach an agent who has to spend another 10 minutes just digging
• 0:35
  SPEAKER_00: through like three different systems to figure out who you are.
• 0:38
  SPEAKER_01: Yeah, contact centers are really buckling under these chronic agent shortages and,
• 0:42
  SPEAKER_01: well, skyrocketing call volumes.
• 0:44
  SPEAKER_00: Exactly.
• 0:45
  SPEAKER_00: So in today's deep dive,
• 0:47
  SPEAKER_00: we are examining Call Tower's blueprint for AI adoption to fix this specific mess.
• 0:53
  SPEAKER_00: Our mission here is to figure out how to deploy AI to solve actual operational bleed.
• 0:58
  SPEAKER_01: Rather than just, you know, buying the latest tech toy, which is the biggest trap right now.
• 1:02
  SPEAKER_01: IT leaders see a shiny new language model and they just ask, where can we plug this in
• 1:07
  SPEAKER_01: Instead of actually identifying the friction points in their customer experience.
• 1:11
  SPEAKER_00: Okay, let's unpack this.
• 1:13
  SPEAKER_00: Because if you deploy AI without tying it to a specific operational problem,
• 1:18
  SPEAKER_00: you're just inventing a new hidden IT tax.
• 1:21
  SPEAKER_01: Yeah, you really are.
• 1:22
  SPEAKER_00: It's like dropping a Ferrari engine into a bicycle.
• 1:25
  SPEAKER_01: Yeah.
• 1:25
  SPEAKER_00: You know, you aren't fixing the commute.
• 1:27
  SPEAKER_00: You're just going to crash.
• 1:29
  SPEAKER_01: Right.
• 1:29
  SPEAKER_01: You'll just arrive at the wrong answer at 200 miles an hour.
• 1:32
  SPEAKER_00: Exactly.
• 1:32
  SPEAKER_01: What's fascinating here is that the blueprint argues you have to anchor the tech
• 1:37
  SPEAKER_01: to executive KPIs immediately.
• 1:39
  SPEAKER_00: Like what kind of KPIs?
• 1:40
  SPEAKER_01: We're talking about things like average handle time and first contact resolution or FCR.
• 1:46
  SPEAKER_00: Got it.
• 1:47
  SPEAKER_01: Yeah.
• 1:47
  SPEAKER_01: If an AI tool can't mechanically shrink the time it takes to resolve a ticket or,
• 1:51
  SPEAKER_01: you know,
• 1:52
  SPEAKER_01: increase the percentage of issues solved on the very first call,
• 1:55
  SPEAKER_01: it just shouldn't get past the whiteboard stage.
• 1:57
  SPEAKER_00: Here's where it gets really interesting, though.
• 2:00
  SPEAKER_00: If the primary goal is shrinking handle times and maximizing containment.
• 2:05
  SPEAKER_01: Meaning keeping people out of the human queue entirely.
• 2:08
  SPEAKER_00: Right, exactly.
• 2:09
  SPEAKER_00: Doesn't that fundamentally mean you're trying to automate human agents out of a job?
• 2:13
  SPEAKER_00: I mean, what is the end game for the staff?
• 2:17
  SPEAKER_01: Well, on the surface, yeah, it looks like a replacement strategy.
• 2:20
  SPEAKER_01: But mechanically, AI in this space actually splits into two different functions.
• 2:24
  SPEAKER_00: Okay, so what's the first one?
• 2:26
  SPEAKER_01: First, you have autonomous AI, like agentic AI.
• 2:29
  SPEAKER_01: And that's not just a traditional chatbot matching keywords.
• 2:32
  SPEAKER_00: Right.
• 2:33
  SPEAKER_00: Those are usually terrible anyway.
• 2:34
  SPEAKER_01: Exactly.
• 2:35
  SPEAKER_01: This is a system with permissions to actually execute tasks across your databases.
• 2:41
  SPEAKER_01: It can reset a password, process a return, or track an order without human intervention.
• 2:47
  SPEAKER_00: Wow, okay, so that absorbs all the massive repetitive volume.
• 2:51
  SPEAKER_00: You got it.
• 2:52
  SPEAKER_00: Meaning the human agent is only dealing with the highly complex,
• 2:55
  SPEAKER_00: emotionally charged escalations.
• 2:57
  SPEAKER_01: Right.
• 2:57
  SPEAKER_00: But honestly,
• 2:58
  SPEAKER_00: taking calls from exclusively angry,
• 3:00
  SPEAKER_00: confused people all day, that sounds like a recipe for massive burnout.
• 3:03
  SPEAKER_01: Oh, it absolutely would be, which is exactly why the other half of the strategy is agent assist.
• 3:08
  SPEAKER_00: Agent Assist.
• 3:09
  SPEAKER_00: How does that work?
• 3:09
  SPEAKER_01: So this model doesn't talk to the customer at all.
• 3:13
  SPEAKER_01: It listens to the live call.
• 3:14
  SPEAKER_00: Wait, really?
• 3:15
  SPEAKER_00: It just listens?
• 3:16
  SPEAKER_01: Yeah, it just listens.
• 3:17
  SPEAKER_01: If a customer is furious about, say,
• 3:20
  SPEAKER_01: a complex billing error,
• 3:22
  SPEAKER_01: the Agent Assist model is instantly pulling up the customer's payment history.
• 3:25
  SPEAKER_00: Oh, wow.
• 3:26
  SPEAKER_01: It's checking the current FLA and even suggesting refund policies right on
• 3:31
  SPEAKER_01: the agent's screen in real time.
• 3:33
  SPEAKER_00: That is wild.
• 3:34
  SPEAKER_00: So it takes over the cognitive load
• 3:36
  SPEAKER_00: of navigating 10 different legacy software systems so the human can focus purely
• 3:41
  SPEAKER_00: on empathy and de-escalation.
• 3:43
  SPEAKER_01: Exactly.
• 3:44
  SPEAKER_00: That makes total sense in theory.
• 3:46
  SPEAKER_00: But, you know,
• 3:46
  SPEAKER_00: connecting a large language model to legacy billing systems sounds like a nightmare
• 3:50
  SPEAKER_00: for compliance.
• 3:51
  SPEAKER_01: Oh, it definitely can be.
• 3:52
  SPEAKER_00: I mean, how do you deploy eugenic AI without it hallucinating
• 3:56
  SPEAKER_00: and giving away sensitive customer data?
• 3:58
  SPEAKER_00: Yeah.
• 3:58
  SPEAKER_01: This is where the deployment roadmap gets incredibly strict,
• 4:01
  SPEAKER_01: specifically around data readiness.
• 4:03
  SPEAKER_01: You can't just unleash an LLM on unstructured, messy internal data.
• 4:08
  SPEAKER_00: Right, because if your security protocols aren't mapped out...
• 4:11
  SPEAKER_01: The AI might pull proprietary data it shouldn't have access to.
• 4:15
  SPEAKER_01: You have to clean and structure the data first.
• 4:17
  SPEAKER_01: You make sure role-based access controls apply to the AI just like they would to
• 4:22
  SPEAKER_01: a human employee.
• 4:23
  SPEAKER_00: Because if you have garbage data with no guardrails,
• 4:26
  SPEAKER_00: the AI will just confidently synthesize someone else's private data
• 4:29
  SPEAKER_00: and hand it directly to a customer.
• 4:31
  SPEAKER_01: Exactly.
• 4:32
  SPEAKER_01: And that creates a massive liability.
• 4:34
  SPEAKER_00: Definitely.
• 4:34
  SPEAKER_00: So what does this all mean for someone listening?
• 4:37
  SPEAKER_00: How do you validate this tech before a massive risky rollout?
• 4:42
  SPEAKER_01: Well, if we connect this to the bigger picture, it breaks down into a five-step roadmap.
• 4:46
  SPEAKER_00: Okay.
• 4:47
  SPEAKER_00: Break that down for me.
• 4:48
  SPEAKER_01: First, you confirm priority use cases.
• 4:51
  SPEAKER_01: Then you establish baseline metrics.
• 4:53
  SPEAKER_00: Like the handle time we talked about.
• 4:55
  SPEAKER_01: Right.
• 4:55
  SPEAKER_01: Third, you assess data readiness early.
• 4:58
  SPEAKER_01: Then you select a low-risk adoption model, like a proof of value or a pilot.
• 5:02
  SPEAKER_00: And the fifth step?
• 5:03
  SPEAKER_01: Finally, you scale.
• 5:04
  SPEAKER_01: You measure the pilot against those baselines to see if handle time actually dropped
• 5:08
  SPEAKER_01: before you ever expose it to your broader architecture.
• 5:11
  SPEAKER_00: Got it.
• 5:12
  SPEAKER_00: So it's a surgical approach rather than a sledgehammer.
• 5:14
  SPEAKER_00: It's all about measurable relief.
• 5:15
  SPEAKER_01: That's exactly right.
• 5:17
  SPEAKER_00: Which brings us to a final thought for you to mull over.
• 5:20
  SPEAKER_00: What if the ultimate benchmark of
• 5:22
  SPEAKER_00: a successful AI deployment isn't how conversational or,
• 5:27
  SPEAKER_00: you know, smart the bot sounds?
• 5:29
  SPEAKER_01: Oh, I like where you're going with this.
• 5:30
  SPEAKER_00: But rather,
• 5:31
  SPEAKER_00: what if the benchmark is that the customer never even realizes they interacted
• 5:35
  SPEAKER_00: with AI at all?
• 5:36
  SPEAKER_00: They just hang up knowing their problem was solved instantly.
• 5:39
  SPEAKER_01: That's the dream scenario right there.
• 5:41
  SPEAKER_00: If you found this helpful,
• 5:42
  SPEAKER_00: subscribe to Connectivity Without Chaos on Substack for more insights on eliminating
• 5:47
  SPEAKER_00: the hidden IT tax.
• 5:48
  SPEAKER_00: And if you're dealing with a connectivity or infrastructure challenge, send it our way.
• 5:52
  SPEAKER_00: We may cover it in an upcoming episode.