What's Up with Tech?

Agentic CX, Real Results

• Evan Kirstel

Share episode

Interested in being a guest? Email us at admin@evankirstel.com

Tired of chatbots that stall when customers need real help? We dive into how agentic AI flips the script by taking actions inside your systems—issuing refunds within limits, updating orders, authenticating users, and escalating with intent—while humans handle complex, high-empathy moments. Joe Anderson, Senior Director and Head of CX and Digital Transformation at TaskUs, breaks down a practical playbook: choose the right high-volume use cases, embed guardrails and escalation paths, and integrate platforms like Decagon and Regal to ship in weeks, not quarters.

We walk through safety by design, from prompt hardening and policy constraints to continuous human evaluation that tunes agents over time. Regulated industries like healthcare and financial services come up often, and we outline how to meet compliance standards while delivering faster resolutions and better customer outcomes. The conversation also tackles a persistent myth: AI won’t erase your support team. It will elevate them. As tier-zero and tier-one tickets automate, human experts shift to white-glove support, retention saves, and nuanced sales where judgment and empathy pay off.

The business model is evolving too. Instead of hourly rates, TaskUs leans into outcome-based pricing and gain share, aligning incentives around first contact resolution, containment, CSAT, and cost per resolution. For leaders just starting, we offer a clear path: pick a major contact driver, launch to a small percent of traffic, measure rigorously, and scale as accuracy climbs. You’ll learn how to convert insights from AI interactions into product fixes and journey improvements that reduce avoidable contacts and lift satisfaction across the board.

If you’re ready to transform customer experience with real action, not just conversation, this episode is your blueprint. Subscribe, share with your CX and product teams, and leave a review telling us which high-impact use case you’ll tackle first.

Support the show

More at https://linktr.ee/EvanKirstel

SPEAKER_01:

Hey everyone. Fascinating chat. Today we're diving into how Task Us is helping companies bring a Gentic AI to life. Joe, how are you? Doing good, Evan. Thanks for having me on this morning. Well, thanks for being here. This is a quite a hot topic, and you're right in the middle of it. Before that, maybe introduce yourself, your role at Task Us, and what's the company's mission these days?

SPEAKER_00:

Yeah, thanks, Evan. The senior director and head of CX and Digital Transformation here at Task Us. And we help our clients and organizations reimagine how they deliver customer experience. Really, for some of the world's leading brands, that involves combining humans and technology to deliver great customer experiences that transform the way our clients do business.

SPEAKER_01:

Brilliant. Well, there's a lot there to unpack. Maybe walk through how exactly Tascos helps clients deploy AI, Agentic AI in production. What does that look like on the ground?

SPEAKER_00:

Yeah, thanks, Evan. Agentic AI has been one of the hottest topics in business over the last year or so. And at Task Us, we're taking an interesting approach. You know, we have helped some of the world's leading brands over the last 17 or 18 years deploy our teammates to provide customer experience and customer support services for their customers. We're taking those same learnings that we have being operational experts for our clients, recruiting, training, hiring teammates, building processes and policies, and employing that same expertise to train and deploy AI agents on behalf of our clients as well. And this is really the next evolution in customer experience. It's going beyond those deterministic chatbots and IVRs that say press one to get to this queue or press two to speak to an agent. We're building um really interactive AI that can take action on behalf of customers and resolve inquiries.

SPEAKER_01:

That's fantastic. And you know, in customer experience, AI has been around for a while in various forms and fashions. Um, how is your sort of modern approach to AI with agentic, you know, different from traditional approaches?

SPEAKER_00:

Yeah, and this is, you know, Chad GBT created a lot of frothiness in the AI market three years ago when it was first launched. And there were a lot of applications that came around and ended up being a wrapper on top of an LLM. Um, you know, we saw some traction in the market with some of those, but um the the next evolution in a gentic AI is actually allowing us to go beyond content synthesization and creation and move toward taking action. Um, we're proud to partner with a couple of leading software platforms in the space, Decagon and Regal, and adopt their solutions and um deploy those on behalf of our clients. So that's where um you know the rubber really meets the road is our Agentic AI Consulting and Transformation Services that we offer. We are packaging our human support services with technology from partners and are able to solve 100% of our clients' CX needs using that model.

SPEAKER_01:

Brilliant. And you know, what kinds of companies or industries are kind of leading the way in adopting your agentic AI approach? Is there a typical profile or you know, vertical where there's a great fit?

SPEAKER_00:

We really see it across industry verticals. You know, AI has become a horizontal that applies across companies. Some of the biggest, though, that are where we're seeing impact are those that are traditionally pretty highly regulated. When AI first came out, um, you know, there was some hesitancy in industries like healthcare and financial services to really embrace AI because of the risks involved. With a gentic AI in place now and the ability to define very specific guardrails and safety standards, escalation pathways for when a human should be involved, um, we're able to see success in uh the more regulated industries now, too.

SPEAKER_01:

Fantastic. You mentioned a couple of partnerships there. Maybe describe the arrangement there, the tech players you're collaborating with. How do these solutions come to life with you leading the way?

SPEAKER_00:

Yeah, we're really excited to partner with Dekagon and Regal, who have best in class agentig AI software platforms. We rely on them as the builders of the technology and then act as a system integrator and partner for our clients to help them deploy that technology across their enterprise and at scale. Really, we find that clients who are and companies who are trying to build technology on their own are not able to really have the impact as quick as they would like. And so that's why we turn toward um platforms out there in the market that are demonstrating success and that we know we can deploy quickly and at scale and have a measurable impact, you know, in a time frame we measure in weeks as opposed to months or quarters.

SPEAKER_01:

Brilliant. And so it is quite easy for AI uh to go off the rails, as we've seen uh many times, uh even in production, and just as users personally, we've all seen it, uh hallucinations and beyond. Uh talk about your approach to kind of safety by design, the your principle there, but also what does it look like in practice in the real world?

SPEAKER_00:

Yeah, you know, embedding safety, human oversight, and the right guardrails in any deployment of AI is key to getting it right and developing trust amongst the users of those platforms. So I think it first starts with selecting the right use cases, which ones are ripped for automation versus which ones have more risk or require more human judgment and more EQ that AI probably shouldn't touch. So we build the escalation pathways in place to make sure that those specific use cases that are kind of human-only ones are going to reach a human. AI is not going to try to automate or touch those. Next, we have the guardrails in place that are set up to kind of monitor the interaction and make sure that AI agents stay on topic, are not prone to any sort of prompt injection or um you know, kind of bad act or attacks. And then after the fact, we employ um rigorous reporting and monitoring um solutions where people are looking at the results on a regular basis. There's human evaluation of those and monitoring so that um we understand should anything um need to be adjusted, we're able to go back and fine-tune those solutions. Fantastic.

SPEAKER_01:

And, you know, tell us about the human workers in the mix. Um, how has the Gentec AI kind of reshaped their roles and what's been their reaction to this new paradigm?

SPEAKER_00:

Evan, I'm glad you asked. This is really the most exciting part of the transformation. You know, as AI takes on the more simple contact types, you know, kind of the tier zero, tier one things like help me reset my password or where's my order? You know, these are um issues that uh are easy to automate. They don't provide a lot of value through having a human in the loop where that human empathy and judgment can really shine. So um our teammates are excited to be transitioning toward those more complex interactions, um, things like um, you know, white glove support, VIP or premium offerings, or sales and retention work that we do that really require a lot more judgment, empathy, and nuance in the reaction that only a human can provide. And so we see a significant portion of our workforce moving toward those more complex roles. We're employing a lot of upskilling programs and training on how to use AI, even within those roles, um, to help our teammates succeed.

SPEAKER_01:

Fantastic. Um the end of the day, it's all about business results, and there's been lots of talk around cost savings and productivity savings. But what what what are the actual results you're seeing with companies from adopting your agent AI approach?

SPEAKER_00:

Yeah, the most exciting thing, Evan, we're able to do is um combine our human plus technology solutions into a cohesive model. And this allows us to change the dynamic on the commercials. We can move from the kind of hourly rate pricing that's traditional in the BPO industry toward outcome-based models, where our clients are paying based on results, um, whether that's per contact solved with humans or technology, or um in game share models too, where these benefits from automation accrue back to task us and our clients in a gain share fashion. So we're really excited about being able to offer um these types of innovative models to our clients, and we're seeing a lot of traction and interest.

SPEAKER_01:

Fantastic. You know, there are a lot of misconceptions about agentic AI, and uh including that, oh, this is like a science project, it's really not ready for prime time. I think you might be proving that wrong. What are the misconceptions that you you see out there? How do you tend to address them?

SPEAKER_00:

You know, I think one of the biggest misconceptions is that agentic AI is here to solve everything. Um, you know, we see really lofty projections by um some of the industry analysts out there that agentic AI is going to totally do away with customer support needs, it's going to eliminate um customer service organizations. It's really not. While it may solve a portion of kind of the simple contact types, there's always the need for humans behind the scenes to make sure that they are managing those agentic AI solutions, um, forcing them to come and improve on a continuous basis, taking those um insights gained by agentic AI and feeding that in back into the overall customer experience ecosystem. You know, the the user experience, the the customer journey, the product roadmap for our companies. There's a lot of information to be gained from um from those insights that Agentic AI generates. And having humans who are in place to evaluate and create a feedback loop based on those results is always going to be super important. And so the role of humans is not going to go away anytime soon.

SPEAKER_01:

Fantastic. Good to hear. Um, so we're a lot of companies are still in a traditional contact center, you know, CX model, uh, haven't yet took taken the leap into AI or Gentec to say the least. How would you suggest they start down this journey? What is a good practical first, second, third step uh to get started?

SPEAKER_00:

Yeah, Evan, I think some companies approach it with um caution and say, look, I'm only gonna pick a very small area to start and deploy the technology. I would actually challenge those listening out there to think big. Um if we pick off you know very small use cases or pilots, should those succeed? And when they do, once we move to scale, um then you won't see the same ROI as you deploy a solution more widely. If you know the use case isn't gonna have the kind of broad scope of impact across your organization. So you know, a good example is you know, you're you might try to automate a customer contact reason or your use case that has two percent of your contact volume. Well, if you successfully do that, um you may have automated two percent of your contact volume. But if you deploy an AI agent that is seeking to um automate a use case that has 40% of your contact volume, that's the use case to start with. Maybe you only deploy it on a subset of that initially to gain confidence and um you know fine-tune the AI agent and the prompt to make sure it's getting the um results right and um to manage the expectations. But you know, start with a drip of the contact volume on that major contact driver and then scale toward um larger volumes within that contact type, and you see a much greater ROI at the end of the day.

SPEAKER_01:

Fantastic. Um, great advice. So, what's next for Task Us? Um, what other kind of opportunities or technologies are you exploring over the next year or two? I won't ask for more longer than that, given how fast things are changing.

SPEAKER_00:

Yeah, and it's amazing how fast the industry is moving. Trying to stay on top of all the new developments is almost a full-time job in and of itself. But um, you know, we'll continue to deepen our partnerships with Regal and Dekon. We're really excited about their product roadmaps. Um, we will continue to invest in our Agentic AI consulting and transformation practice to bring deeper expertise to our clients um across industries. And we're excited to also um be able to offer those alternative commercial models, like I talked about.

SPEAKER_01:

Well, exciting times. Uh, it's going to be a busy 2026, to say the least. Congratulations on all the success so far, onwards and upwards.

SPEAKER_00:

Thanks, Evan. We're excited to ride the rocket ship.

SPEAKER_01:

Absolutely. And thanks for listening, watching, sharing this episode, everyone. And be sure to follow us at techimpact.tv as well for the TV show on Bloomberg and Fox Business. Thanks, Joe. Thanks, everyone. Thank you, Evan.