All Episodes

July 23, 2025 52 mins

Zendesk CEO Tom Eggemeyer joins Michael Krigsman to discuss the impact of AI CX on customer expectations. They explore real-world lessons from the Zendesk AI journey, including what has worked and what hasn't. Learn how customer service is being reshaped by ChatGPT and artificial intelligence and machine learning.

Key Topics Covered:

  • The shift from traditional chatbots to reasoning AI agents
  • Why Zendesk only gets paid when AI successfully resolves issues
  • How AI is more accurate (and sometimes more empathetic) than humans
  • The surprising ways AI is changing customer service metrics
  • Real-world examples of AI successes and failures
  • The future of work in an AI-dominated landscape


🔷 Newsletter: www.cxotalk.com/subscribe

🔷 LinkedIn: www.linkedin.com/company/cxotalk

🔷 Twitter: twitter.com/cxotalk

🔷 Episode takeaways: https://www.cxotalk.com/episode/zendesk-ceo-on-ai-in-customer-experience-what-works-what-doesnt-and-whats-next

00:00 🤖 Introduction to Zendesk and AI in Customer Support
01:04 ⚙️ The Realities and Challenges of AI Implementation
05:32 🌊 Zendesk's Transformation and the Future of AI in Customer Support
09:21 🔄 Transitioning to AI: Challenges and Reactions
11:23 🤖 AI in Customer Support: From Reactive to Proactive
16:38 🔍 Transparency and Use Cases in AI Decision-Making
19:12 ✅ The Role of Quality Assurance in AI-Driven Customer Service
21:39 🤖 AI Adoption and Industry-Specific Applications
28:05 🤝 Balancing AI and Human Touch in Customer Service
31:16 🤖 AI as a Strategic Differentiator in Customer Experience
35:22 📊 Evolving Metrics and Challenges in AI-Driven CX
38:27 🌐 AI's Broader Implications and the Future of Work
42:33 📊 The Role of Data in AI Development
45:33 🤖 AI's Impact on Jobs and Customer Experience
47:52 🔒 Preparing for the AI Era and Data Privacy

#AI #CustomerService #CustomerExperience #Zendesk #ArtificialIntelligence #CX #CustomerSupport #DigitalTransformation #CXOTalk #TechLeadership #cxotalk

Mark as Played
Transcript

Episode Transcript

Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Welcome to CXO talk #886 I'm Michael Krigsman and we are
cutting through the hype around artificial intelligence and
customer experience with Tom Eggmeyer, CEO of Zendesk.
Tom will share honest, real world lessons from Zendesk's AI

(00:21):
journey, what's worked, what hasn't, and how AI is reshaping
customer expectations. Tom, welcome to CXO Talk.
I'm delighted to chat with you. Great to be here Michael, and
I'm delighted to be #886. Everybody has received messages

(00:41):
from Zendesk. We submit a support request and
we get something back from Zendesk.
So just briefly give us some background.
Tell us about Zendesk. Zendesk is AAI first customer
support software company. We're all about helping
companies resolve their customers, their employees, or
their fellow business issues. What's going on with AI and

(01:07):
customer experience today? When we think about AI from a
customer support perspective, wethink there's just huge, huge
gains for companies, employees and consumers.
From an AI perspective, the it'snot hype, it's actually real.
So when you say that AI is disrupting, can you elaborate on

(01:28):
that? AI is actually solving customers
problems. The hype, I'm old enough to know
20 years ago we talked about machine learning or predictive
analytics. We're going to be able to solve
customers problems, quite frankly, without human
intervention. And it's been a little bit of a
hype cycle for the past 20 years, but we're actually seeing

(01:48):
that happen right now. We have one product called AI
Agents Advanced where you put a bot into one of our customer
environments and with three clicks of a mouse, taking some
knowledge from the company, we can solve almost instantaneously
more than 30% of customers interactions.
And what's interesting is the customers are more satisfied, it

(02:12):
costs less, it's more accurate and you can do it with three
clicks. And that's kind of the promise
of AI is always on fantastic accuracy, lower cost for
companies so they can go put their finite resources into
maybe solving the the root causes that are causing them
having problems with their customers.
There's so much discussion around AI and you've just

(02:35):
presented. Can I call it a a kind of
idealized version or is it totally realistic?
Or where's the gap? The gap is a couple things.
Number one, you know, I see somepeople in the marketplace, you
know, guaranteeing 65 to 85% of interactions can be resolved

(02:55):
with AI within a matter of days.And I think that's hyping a
little bit. There are some of our customers
that are getting 85 or 90% resolutions, but there's some
customers with more complex use cases that are getting 20 or
30%. So I think part of the hype is
it kind of depends on the customer type, what kind of
interactions having with their employees or with their

(03:16):
customers on what kind of resolution rate you can get.
The second thing there's a little bit of hype is right now
you need to have something called a knowledge base or a
bunch of information or other knowledge, you know, structured
data so that you can get resolution rates quickly.
And that's not always the case. And so we got to be honest with
customers and honest with consumers about that.

(03:38):
And then third, there are still mistakes that AI makes just like
humans, and we don't want to underestimate that.
And we want to make sure that weget as close to 100% accuracy as
possible. What's interesting is we've done
some AB testing and we see less hallucinations from AI than we
see mistakes from humans. But it's interesting that we use

(04:00):
different nomenclature. We call AI agent, you know,
mistakes, hallucinations and human agent mistakes, mistakes.
When things are set up in the right way, that the data is in
the, is organized properly and so forth, then the AI does a
great job, is essentially what you're saying.
Exactly, I was with a customer about a month ago in Australia,

(04:21):
for instance. And what they, what we found out
working with them is they had five different policies, for
instance, on from different parts of the organization on
what constituted a return. It was an e-commerce company.
And when you have data like that, of course the AI is not
going to get it perfect because it's getting input at 5
different, you know, policies that are slightly different from

(04:43):
each other. And so really having good
policies, having good knowledge,having good information is
really, really key. Now we can help and you know,
other players in the market can help create that knowledge.
Like one of the things we do nowis we see a bunch of the
customer interactions and we seeyou have a gap right here.
In the past, we would have said you have a gap, you should go

(05:04):
solve it. Now we say we, you have a gap,
here's a recommended. We write it for you, OK?
The AI writes it for you. A recommended article or
recommend a policy, Do you want to accept?
Do you want to decline? Do you want to edit it?
So the world is kind of flippingon itself right now.
But like you said, data is really structured and

(05:24):
unstructured. Data that's consistent, that's
publishable, and that's readableis really important in this AI
journey. You mentioned this phrase AI
journey. I was really interested.
At what point did you recognize that AI and customer support was
so important that you had to? Would it be?

(05:47):
Is it too strong a term to say, transform the company around
this? Absolutely transform the company
because the way we look at it asZendesk, we can be the disruptor
and AI customer support or we'regoing to be disrupted.
And we were the disruptor 17 years ago with customer support
helps the help desk, and we wantto do that again.

(06:07):
I remember when Open AI launched, I think it was October
2022, and there was an amazing amount of buzz about it.
And just thinking about the implications of it in the summer
and the fall of 2022 and I was over in our Lisbon, Portugal
office. I remember this vividly in
December 2022 with some of our lead, we had a small group of AI

(06:30):
data scientists and engineers, about 10 of them, and I asked
for our road map. I just started that month as CEO
and we had a bunch of great ideas, but we didn't have a road
map. And I think talking to industry
experts in the summer and the fall of 2022, you know, I came
to the conclusion that just likeInternet, just like mobile, just
like premise to cloud, there wasgoing to be a major, major

(06:53):
disruption. I think some people came to that
maybe 6 or 12 months ahead of that and we did not have a road
map. We did not have enough people on
it. We did not, we're not thinking
about it. And so we spent a couple days
writing on a whiteboard. I wish I would have taken
pictures of it in Lisbon, Portugal, about what kind of
first product that we wanted to launch.
And so that was kind of the aha moment for me.

(07:14):
You know, I would have thought it would have been the Bay Area,
but was actually in Lisbon, Portugal.
You mentioned disrupting yourself.
That's a hard thing to do. Most of the time businesses get
disrupted because there's a competitor, there's something
going on out there and they haveno choice.
But in this instance, you saw the tidal wave coming and so you

(07:36):
took proactive steps and said, hey, we need to change.
It seems like that's what you were just describing.
I don't think we had an option, honestly, Michael, you know, if
I'm really honest, because the customer support, like we have a
stat, we think 80% of customer interactions with companies are
going to be through AI within five years and a 100% are going

(07:58):
to involve AI in some way withinfive years.
Many humans will use AI. So we didn't have a choice, but
we have a lot of great Zandeskians employees that saw
this wave coming and saw that, hey, we were the disruptor 17
years ago. Mikkel and the whole Zendesk
team disrupted the help desk andcustomer service and we wanted
to be, we wanted to do it again.So I would say we're fortuitous

(08:23):
that this was coming, that we had some of the building blocks
and pieces in the foundation. But also it's just an imperative
that if we didn't do it. And I had seen in the premise to
cloud contact center world, I work for a company called
Genesis that was 100% Prem. It's now one of the leading
cloud providers of contact center.
They went, we went through that metamorphosis where the other

(08:44):
two leaders of Contact Center Prem didn't go through it.
And so it's easier to do as a private company.
But there's definitely an imperative to do it because
there's going to be winners and losers here.
And I really feel responsible first for our customers to give
them value, second to our employees and finally to our
shareholders that we need to disrupt ourselves.
I want to remind everybody that there is a tweet chat taking

(09:07):
place right now on Twitter X andyou can ask your questions using
the hashtag CXO Talk. And if you're watching on
LinkedIn, just pop your questions into the LinkedIn
chat. And we have some questions that
are coming up. But I'm I'm still really
interested in this transition. So as as you were moving from

(09:32):
the pre AI world into AI, did you get any resistance?
Did you from from stakeholders, from your board, from employees,
from your customers? Because it's a big change, and
change is always hard I. Don't know if there's
resistance. I'll be honest with our board,
it's actually been how can you go bolder, How can you go

(09:52):
bigger? Because they see the
transformation. So if anything, I, we have a
great board and I say they do strategic agitation and their
agitation again is go faster, think bolder, think bigger on
how you can really give value toyour customers.
I think there's different levelsof maturity within our customer
base. Some people are like, want to be
cutting edge. Some people were over the last

(10:12):
couple years saying, hey, is this going to be real?
Is this really going to disrupt?Are we really going to be having
AI agents copilot generative search for knowledge bases,
really transform customer support?
I think the vast majority of customers are saying right now,
you know, after these two 2 1/2 years that AI is real, AI is
disrupting customer support. So I think it's gone from a

(10:33):
question mark 2 1/2 years ago with our customer to a
consensus. The board has already been there
and employees I think that ask real good questions is like, how
are we going to be the disruptorand not just be disrupted?
How are we going to do all the core things that we need to do
on our core customer service Omni channel platform while
adding AI? Am I going to be left behind if

(10:53):
I'm not part of the AI revolution?
And so we have, you know, coffeechats, we have Amas where we
talked to the employee base. But if anything, I think the
employee base is really, really excited about being the
disruptor, being number one in customer service and really
serving our customers and getting real, true, tangible
value to our customers with AI. Now would be an excellent time.

(11:16):
To subscribe to the CXO Talk newsletter, go to cxotalk.com.
We'll keep you updated on upcoming shows.
Tell us about the changes that you drove into the product.
So how is customer support rooted in AI different from what

(11:36):
came before? There's a couple of things where
we are now and where we're goingin the future.
Where we're going right now is we think we can automate a lot
of customer interactions. So for the customer, again,
that's what it's all our customers.
Customer it's all about. It's quicker, it's more
accurate, and they're able to solve their problem in a really
good way. For companies, they're spending

(11:57):
less on it. And So what has happened right
now in AI is a little more reactive.
And I'll give you a couple of use cases.
In the future, it's going to be more proactive.
But in the past we have about 30,000, 1000 knowledge based
customers. So think you go into a company's
website and you say how you typein how do I return these pair of
shoes? A knowledge base in the past

(12:17):
would come up with 5 or 10 blue links and you'd say and you'd go
pick one on how you return shoes.
What happens now is we've launched something called
generative search, and this is free part of our product for
knowledge bases where like a Google Gemini search, you
actually get the answer, you know on how to return the shoes.
And it's getting even better with a gentic AI where you might

(12:39):
be able to do that whole process, including going back to
an ERP, for instance, all you know the actions all through an
AI agent. So that is kind of reactive
still. You then go to an AI agent,
maybe from the degenerative search and an AI agent can act
like a human agent and do all the things a human agent can do.
Sometimes generative search and an AI agents cannot going to

(13:00):
solve the problem. And so you go to a human agent,
but that's all reactive right now.
We're kind of flipping that right now in customer service
where it's proactive, where we're taking all this
information from the searches you make, Michael, the
interactions that you have with an AI agent and human agent.
And we're saying these are 7 root cause issues that are going
on. We keep sending out black shoes

(13:22):
when Michael won it, red shoes. And then we have customer
interactions. What do we need to go do and
like have an insight into the supply chain?
Why do we have a problem there through AI?
And so it's getting being flipped on its head right now,
customer support from reactive to more proactive, from reactive
to more root cause analysis and from reactive to actually
actions of solving root cause problems that a company is

(13:45):
having interacting with its employees or customers.
And again, I want to come back to this, this comment you've
made a couple of times on the accuracy.
Can you talk about that? Because of course this is a key
dimension of everything here. We've done a lot of AB testing
and on let's call it the top 50 or 100 use cases.
Usually an AI agent is as accurate or more accurate than a

(14:09):
human agent. When you get to really complex,
maybe a customer asks 3 different things in a query,
they say in an e-mail or a chat or a phone interaction where
there's voice AI. Now you say how do I return my
shoes? I want a credit, I want to order
a new pair of red shoes instead of my black shoes, and I want to

(14:30):
go order a sweater. When you get 3 or 4 queries in a
row, like in one message, sometimes AI does not nail that.
I think a human does better still, but it's getting better.
The other place where I think AIis not quite there is a really,
really complex corner case that the, again, the data does not
solve where a human needs to getinvolved.

(14:52):
On the other hand, I think AI, it's interesting, some of the
data that we've had some research recently, data is doing
just as good or better. The data says on empathy, which
is surprising. I would have always thought
humans are going to be more empathetic, but there's some
interesting research piece I'm forgetting where I read it
recently where doctors, AI agent, doctor and AI or a human

(15:13):
doctor. AI agent doctors are considered
more empathetic and so we're starting to see more empathy,
more caring from AI agents, which is something I never
thought we would have had. We have come a long way from the
chat bots, which were simply prescriptive and frankly for a

(15:34):
lot of users, honestly a waste of time.
Things of There's a very different universe now.
Exactly. So those chat bots that you
probably have in your mind that would drive me, you know, we're
they'll be frustrating for me, frustrating for you.
We're kind of decisions tree. You know, if then you go yes,
and you kind of go down the decision tree and a lot of times

(15:56):
again, the inquiries were more complex and you get stuck at
part of the decision tree and then you get handed over a
human. The human wouldn't agent
wouldn't have the context of that chat bot.
What's different now with the gentic AI is there are these AI
agents are basically autonomous and can reason and that's the
fundamental difference from the chat bot experience everyone
had. And So what I I tell everyone, I

(16:18):
know you've probably dealt with chat bots in the past, but if
you drew you deal with a true agentic AI agent right now, I
think you're going to be wowed by the experience.
It's not perfect, just like human being AI.
Human agents are not perfect, but that's the fundamental
difference. You're going from a kind of a
decision tree pre baked to something that's agentic.
That's something that's reasoning.
What we think is really important.

(16:39):
And as part of this is that you need to be white box, not black
box. And if I were a company or
consumer, like how did the especially a company, how did
the autonomous AI agent, the agentic AI agent make decisions?
What was their thought process? One of the things we do is we
lay that all out, you know, for every single interaction an AI
agent or a bot has with a customer so that companies can

(17:03):
go refine that by giving naturallanguage instructions to it to
keep getting better and better. I think that's really, really
important because if you're black boxing it as a company,
you don't know why the AI agent is doing something.
So it's harder to improve. I use large language models,
multiple LLMS every day all the time and for the reasoning

(17:25):
models, when they expose their logic, it is so useful.
I agree. I'll give you an example of my
personal life. My wife's grandfather, he was
born in the US, but he grew up in Ireland and my wife wanted to
know can she get Irish citizenship?
I took ourancestry.com, gave thepassword to a large language

(17:46):
model, connect it to an e-mail account that I made-up and
create it. Made-up create it and said can
you go research and grab recordsto support the claim that my
wife can go get Irish citizenship.
It was amazing what it was able to do.
It requested parochial grade school records from an Irish

(18:07):
Parish School. It got me all this information,
then wrote a legal memo on why it thought my wife was able to
get Irish citizenship. The only problem kind of to your
point, Michael, is it got the legal conclusion wrong.
It looks I it said the legal conclusion was, yes, it's
probably a 5050 corner case thatshe got it, but it exposed some

(18:28):
of the logic and some of the reasoning and that's how I
caught it might have been wrong on its conclusion.
So that's why I just think it's first of all, cool business
case, right? You can take your tree from
ancestry.com, Create an e-mail account, get a legal memo,
document supporting it by just giving some prompts.
But on the other hand, that's why it's so important to give
that white box treatment to really understand and the logic

(18:51):
behind it, so you can still see flaws because there's flaws and
mistakes just like human beings would make.
When the AI makes a mistake and a customer service agent is
actually working with a customer, do you have ways of
trying to trap that or how how is that get managed?

(19:12):
Because the mistakes can be insidious, right?
The LLMS speak with the full confidence of quote UN quote,
someone who knows they're absolutely right all the time.
So one of the things we think wedifferentiate and we think it's
important for any player in customer service is we have
something called quality assurance.
And quality assurance is using adifferent algorithm than a large

(19:35):
language model. And we actually score every
single AI agent or human agent interaction for accuracy, for
tone, for resolution. And we find through that where
humans are making mistakes or AIagents are making mistakes and
we surface that. And so again, the white box and,

(19:56):
and you're totally right, one ofthe flaws right now is AAI agent
will say with absolute certaintythat my wife can get Irish
citizenship where it's a little murky, but be 100% confidence.
And, and so we have this real strong point of view.
You should be as a company doingquality assurance, not on a
sample size. Right now, most companies are
doing like 1% of interactions. They're analyzing, we say to

(20:21):
analyze 100% of interactions andyou can get summaries of that
now based upon AI. And that's how you're going to
spot issues, whether humans or AI agent.
And what's really key about thatis sometimes you're going to see
the AI agent is handling a question or a query better than
the human agent. But sometimes you're going to
want to have with like my 80 year old parents, my mom will be

(20:41):
upset. She's not quite 80 yet, but my
father, my father's 80. You're going to want to have
them talk to humans. They're not going to want to
talk to an AI agent. And you want to get that AI
agent accuracy into the humans. And so we really think quality
assurance systematically is so important in the AI era.
So the AI is also enabling you to have a far more granular type

(21:05):
of ongoing QA than was possible in the past.
Allowing us to have more granular, more ongoing and
what's the great thing is surface those insights where
there are mistakes and propose actions, whether go build this
policy, go change you have you have policies that are
conflicting with each other. Go make sure that you create a

(21:26):
policy here or a piece of knowledge to go solve it.
So has more granular insights, it's more personalized.
And with the Gentic, AI can propose actions and make actions
to go hit the root causes of customer dissatisfaction.
Let's jump to some questions. And the first question is from
Greg Walters. He actually asks a couple, but

(21:46):
I'm going to go to his second one and he says regarding
disruption, AI eliminates the middle person.
Do you see a time where Zendesk is more than a component of and
a tool for customer service becoming quote customer support
and service the department? I know some people, some

(22:08):
companies are out there. There's something called BPOS
that will say you can outsource a lot of your customer
interactions to us and we'll be in charge of the technology and
the people to solve that. I think in the future, our
platform needs to be reactive for our companies, but also
proactive. And again, it's going to be
interesting. It's going to change the

(22:28):
dynamic. Right now, we're all about the
reactive customer issues. Greg, I think that's a great
question. By the way, in the future, I
think we're going to be solving problems before they even occur.
You know, with our companies. I don't know if we're ever going
to be the customer service department itself.
I think we're going to be a platform to help the customer

(22:49):
service resolve issues before they happen and when they do
happen, solve them. But we have debates and I tell
everyone this every week, customer service is massively
changing. Every week there's new AI
innovations and we have had discussions on how does our
platform evolve and are we goingto be taking more and more of
the load from customer service departments in the future.

(23:10):
I think it's a really insightfulquestion.
Another question from LinkedIn and we'll jump to those
questions on Twitter shortly. This is from Brandon Byrd.
And, Brandon says, are there specific verticals where Zendesk
plans to double down with AI to gain a first mover advantage?

(23:32):
We think over the next five years that vertical industries,
the more deep that you can get into tasks and procedures and
expertise, the better you're going to be able to solve your
customers, employees or your ourcustomers customers issues.
And so we right now have a really, really good foothold in
e-commerce, fintech manufacturing and I can go on

(23:56):
with the list tech. And so we are going deeper and
deeper and taking our learnings from those verticals and
applying them to our customer service AI driven resolution
platform. And so I think that's going to
be one of the secret sauces where companies help with help
their customers get more value by differentiating on a vertical

(24:16):
or industry basis. And so it's already happening
now. I think we're in the early
innings or the early part of thematch for your international
audience. But I think you're going to see
over the next five years a lot more industry specific solutions
and industry, industry specific knowledge and industry specific
AI. We have a question from Firestar

(24:40):
who says it feels like we're finally past the AI is coming
phase and into actually making it work for customers.
The technology works, but I think where we are in the, the,
the, the, the, the next evolution is getting it adopted.
It's interesting we see customers maybe don't want to

(25:01):
adopt it on the, on the one end.And so one of the things we tell
our customers to do, I've seen some, and this is learning from
other customers is you might want to say, I have an AI agent
available right now for you or do you want to wait 5 minutes
for a human agent? And so a lot of people will test
the AI agent because they've hadthe bad experience, Michael, you
had with the chat bots in the past.
And so you kind of want to nudgethem to go try the AI agent.

(25:24):
We've also have the human agentswhere there's something called
Copilot they're using and they've been using other tools.
And so how do you go nudge them to go get adopted?
And once they get, they see the power of Copilot, they'll start
using it more and more. And then just implementing with
the customer to make sure that you're getting the data right,
you're structuring the policies right.

(25:44):
You're really taking a gentic AIand starting to run through your
business process and workflows and creating new processes in
workflows with AI. There is still some adoption
challenges in the marketplace and just the fact that you can
do a lot with three clicks, but it might take you a couple weeks
or a couple months to get the full value stack or beyond that

(26:05):
journey is another problem with adoption right now.
So I think we've gone from does the technology work to how do we
get employees, customers and autonomous AI agents, all three
really embracing the technology,the adoption curve, the
implementation adoption I think is the next frontier that you
know, Zendesk is dealing with and other companies are dealing

(26:26):
with. It's a great point.
And of course, across different industries and different
companies, as you mentioned earlier, there are varying
degrees of maturity which has a direct impact on that adoption
capability. Definitely.
And it's really interesting, we're seeing, you know, when you
talk about a small customer, they're looking for

(26:48):
instantaneous impact and they'relooking for the three click
solutions. When you talk to an enterprise
customer, they're really lookingfor embedding AI deeply and
tasks and procedures, changing their workflows, inventing new
workflows, having actions that take say like I used the return
example, it's used across the ERP, across the supply chain to

(27:12):
really interconnect it. So different customers are
asking for different things. And I kind of look at it,
there's a maturity curve for customers.
There's a segment whether you'reSMB, mid market, enterprise.
And there are also these geographic differences people in
the Bay Area, for instance, are ready to embrace.
I'm originally from Ohio, my parents are from Kentucky.
When I talk to customers in Ohioand Kentucky, they might be a

(27:35):
little more, you know, where theBay Area was a year ago.
So you have this kind of Rubik'sCube based on geographic region
segment and maturity to the customer where they're ready to
go. But I think largely we're
getting over is this hyper real and people are accepting that's
real. I have a lot of questions of my
own, but I'm going to, I really like to prioritize the questions

(27:55):
from the audience, The audience you guys are, I always say this,
you guys are really smart. We have a question.
I'm still on Twitter X and we'regoing to jump to LinkedIn in a
moment. Question from Elizabeth Shaw on
Twitter who says as AI enabled customer support functions
increase, especially with AI agents and then agentic AI, how

(28:21):
important will the human touch be and when will human touch be
the right thing? So how important is the human
touch and when do you need to loop in the human from the
electronic agent? We still think 10 or 20% of
interactions are going to be human LED and they're going to
be the most complex interactions.

(28:41):
They're going to be segments of your customer base.
Maybe it's your top customers and, and there could be people
that just want to talk to a human.
And so we think there's going tobe two or three areas where
humans are going to be still thecentral focus point of a, of a
resolution. And sometimes, even though I
said sometimes the AI agents aregiving more empathy, if a

(29:03):
customer is really upset, a lot of times a human being can do
that really, really well. So I guess that would be the
first point that humans, we believe are still going to be in
the loop. The second thing that we think
is really exciting is I know there's been a concern that
people are just going to start cutting human agents.
And what we've seen is somethingdifferent.
We call it there's been a service deficit up until now.

(29:26):
I think most people would say when they interact with brands,
whether they're a business or consumer, they might not get the
best experience. And so there's this service
deficit. We think that AI is going to
give a service dividend where the humans are going to be able
to free up to spend more time onthe root cause of some of these
bigger issues. They're going to be able to

(29:47):
manage the AI agents. So you might have a human agent
managing a team of AI agents to what we talked about before,
Michael, to make sure that there's accuracy using QA and
other tools. And we think the service
dividend is just going to go increase customer satisfaction
again, whether that's B to B usecase or B to C use case or an
employee, because you're going to be dealing with more the root

(30:07):
cause and preventing things before they happen and you're
going to get more personalized. And so we think there's going to
be a couple reactive use cases. We believe, you know, people
need to remember that there's always a human at the other
side. Even though, and I'm sure we'll
talk about it, Michael, in 12 to18 months, you're probably going
to be using your own bot to go talk to a company's bot.
At the end of the day, there's still that human on the other

(30:29):
side of the interaction, and we need to remember that.
And this is from Preeti Narayan.And she says as AI reshapes
customer experience, how do you see industry verticals like
retail, healthcare and financialservices evolving their customer
engagement models? Are there industries where AI

(30:52):
driven CX has become a strategicdifferentiator rather than just
an operational tool? And how is Zendesk positioning
itself to lead in those transformative shifts?
And I'm just going to make one request to people asking
questions, keep your questions simple because I'm reading your

(31:13):
questions and I need to make sure I get it.
OK. So AI driven CX becoming a
strategic differentiator rather than just an operational tool
and how you're positioning to lead in these transformative
shifts? What's really interesting is we
are seeing some customers see itas a a transformational tool
that can really differentiate their business.

(31:34):
And I'll give you an example. We have more and more customers
that don't want to be customer references because they don't
want to let their competitors know that they are getting 70 or
80% of their interactions solvedby AI agents and they're having
this service dividend that they can go reinvest.
And so some of our best customers don't want their name

(31:55):
out and don't want the results out.
And to me, that's kind of the proof in the pudding that they
are getting amazing automated, automated resolution rates.
So the AI is solving the problembefore it goes to human and they
see that as a competitive advantage.
The way we're trying to do that is we really focus on what we
call the resolution platform or automated resolutions and the

(32:17):
whole system of the resolution platform.
So we are looking with our customers on a day-to-day week
by week basis on what percentageof your interactions have you
resolved and have you automate it.
And again, like we talked about Preeti, this was a good question
before. We're getting deeper from a
vertical perspective on that. And when we get those high
automation rates, we really think you can go into that

(32:40):
service dividend land. And so that's how we're
positioning ourselves with our customers.
We're not perfect, but we're really trying to drive our
customers to higher and higher automation rates.
And that's why we've come out with our new pricing model
called automated resolutions. We only get paid if we solve a
customer's problem with you, if we resolve it, but it has to if

(33:02):
it goes back to a human, we don't get paid for that
interaction even though it costsUS money.
And that's a way for us to kind of lead the way in customer
service and put our, you know, put our operating model totally
in line with our customers because they care about driving
up the automation rate, the automated resolutions.
And we're in the exact same boatbecause about how we price.

(33:24):
I like that, putting skin in thegame.
Historically software companies,you know, going back to on Prem,
they sold the product, threw it over the wall, and if the
customer used it or didn't use it, they still paid.
And now you're going to the opposite extreme where Zendesk

(33:45):
has a stake in the customer outcome as opposed to the
process of getting to that outcome.
We talked about transformation and culture a little before and
employees. This is a big transformation for
our customers, for our employeesand for our shareholders because
how we look at operating model. In the past it was seats.

(34:06):
We still have some seats. Some companies made the switch
to interactions and now the nextevolution is to outcome based
pricing or really aligning with your customers.
I'll give you an example. I was on an airline, I won't say
we're in the world recently and I noticed that I was going to
miss my connection. I got on their app.
I think they had a beta going onwith AI agents and I got about

(34:29):
90% done solving my problem of re booking my connection
unfortunately took off, I turnedoff my phone, no Wi-Fi, I
landed. The AI agent hadn't solved my
problem. I'm confident that airline,
maybe they didn't get charged but they would have gotten
charged $2.00 for an interaction.
Even though I was more dissatisfied because I called a

(34:49):
human agent then and the human agent didn't have any of the
context from the AI agent. And so I felt like I was
repeating myself. And oh, by the way, I had to
wait like 8 hours at the airportbecause I missed all the
connections because they got booked in the meantime.
And so that's an example where if you're on a seat based or
interaction based, the software provider probably got more money

(35:09):
from my interaction with that AIagent, even though I was more
dissatisfied. And that's not what we want to
do as a company. We really want to align our
customers outcomes to our outcomes.
Here's an interesting one from Ryan Smith.
This is really nuts and bolts, he says.
As AI begins to resolve tickets faster and automate more

(35:29):
interactions, how should CX leaders rethink success metrics?
Do legacy KPIs like customer satisfaction and FCR first close
resolution still matter or do weneed a new framework?
I'm going to give you an example.
There's a lot of companies rightnow that use average handling

(35:51):
time. So how long does a human agent
stay on, whether it's a phone and e-mail, a chat, you name it,
a web form, how long does it take for that human agent to
resolve the problem? And you want lower average
handling time. One of the things that we tell
our customers is once you implement automation, automation

(36:12):
takes let's call all the easy and medium hard interactions
away. And so the humans are left with
the most difficult ones. And so we tell CX leaders, your
average handling time is going to go way U OK, and you're going
to be like, why did this happen?I just imlemented AI agents and
automation and then I gave my human agents copilot, which
should make them more efficient,but their average handling time

(36:34):
is going up. And so we're starting to develop
more frameworks with our customers.
So they're not surprised when you explain that, you know, a
customer CX leaders like of course, because they're getting
the 10% or 20% of interactions that are really, really
difficult that take a long time.OK.
And so we're encouraging more ABtesting and different
frameworks. I do think still NPS or CSAT

(36:57):
matters. I think there's going to be some
new frameworks and some new metrics like how much of how can
you reduce interactions not for reducing interactions sake
because you're solving the root cause.
And so I think over time how quickly you reduce the
interactions that keep coming upis going to be a metric in the

(37:18):
future. Average handle of time is going
to be less of a metric. I still do think first contact
resolution is important. If you look at the psychic, the,
the, the psyche of customers, they rather wait longer to
interact with an AI agent or human agent, read a rate on
hold, quote on hold for 5 minutes and get it solved
immediately rather than being moved around two or three times

(37:40):
between different humans and AI agents.
So I think first contact resolution, but Ryan, I think
your prep supposition that we'regoing to have an evolution in
metrics is absolutely going to happen with CX leaders.
It's so interesting how AI is driving change in so many

(38:00):
different directions. The the tentacles are far
reaching and the implications are very profound.
I mean, just just look at what you just described with your
pricing policy. I mean, that's such a dramatic
shift. I don't think we know, Michael,
all the implications of AI and Ithink people that are certain on

(38:22):
how it's going to go over the next, you know, 15 or 20 or 25
years are probably wrong, but itis profound.
You know, one of the things I dois every two weeks, I do a deep
dive on what's changed in AI in the last two weeks and how could
it impact our customers and how can we provide more value to our
customers. And I would tell you, you know,
if you're in the customer service industry from 2009 to

(38:45):
2019, like I was at another company, like the changes you'd
have in years are about the samepace that we now have in weeks.
And so there's just so many profound implications on the
future of work. You know, I have a 22 year old
daughter and a 20 year old son, and they asked me with AI, what
should I do? And that's like kind of an
existential question about what skills and capabilities they

(39:06):
should learn. Because I think the nature of
work is going to just massively shift over the next 10/15/20
years for instance. As CEO of Zendesk, what does
this rapid change? It's extraordinary that you have
to look every two weeks to see what's changed in this AI world
that could have some type of potentially profound impact on

(39:30):
the market and on your company'sZendesk.
It really changes that. You need to make sure you're
getting a lot of signals from customers in particular.
And like when we're doing well and when we're failing, I think
is really, really important fromthe industry at large is really
important. And I think it really what I
encourage employees is you've got to be in a learning
employee. You've got to learn AI, you've

(39:51):
got to embed it into your daily workflows.
You got to stay on top of it because that is a skill.
And I think it's going to differentiate employees going,
including CEOs in the future whoreally understands AI.
Who's embedding it in their daily life?
Who's using it strategically andtactically?
And so I think always learning, I, I give advice to people, you
have to always be learning and you have to be always learning

(40:14):
about AI. It is going to impact your job
in some shape, way or form over the next five years, no matter
what job you're doing right now.This is from Arsalan Khan, who
is a regular listener. He asks great questions.
He says. When will we be able to replace
humans with AI? What happens to people?

(40:35):
Will manager AIS be able to write performance reports for
other AIS and humans? I'm an optimist about this
point. I think AI is going to enhance
people's lives and I think we'veseen with every major revolution
in the world it create more jobsbut create different jobs.
So I'm still an optimist here onwhat AI is going to do.

(40:59):
I think, you know, it's an insightful question about
performance reviews. I think right now where we are,
I think AI can give summaries. If you get a lot of 360 feedback
summaries for performance reviews, insights for
performance reviews, it's probably going to happen in
151015, maybe even sooner that you are going to have
performance reviews written by AI and edited by a human.

(41:20):
I think it happens right now. I've talked to some startups
that are doing that right now. And so I think, again, we just
got to really understand at the end, the other end of that is a
human getting a performance review.
And you've got to be really, really confident in the
technology that it's getting allthe data sources, all the
signals, you know, honestly in agood way and transparent on,

(41:41):
again, that white box theory, Michael, on what it got and how
it came to its reasoned conclusions about your
performance. We're already there with quality
assurance where we are evaluating 100% of interactions,
whether it's a human agent and AI agent and scoring them.
It's not a formal performance review, but it's like indicators
of success. So we're already there.

(42:02):
I think there's some really profound implications for human
society over the next, again, 5101520 years about how jobs are
going to evolve. Your long time listener, I think
alluded to it where he said you're going to have different
jobs. And we think there's going to be
managers, human managers, managing teams of AI agents

(42:23):
going forward. And that's going to be probably
one of the new jobs that's created from this AI revolution.
Arsalan Khan has been waiting a long time, so I'm going to pop
another question from him and hesays on Twitter, can you talk
about the role of data in your product and AI?
One of the things that we have an advantage as a company is we

(42:46):
do 4.6 billion resolutions a year.
And so there's a huge, huge dataset that can help post train our
models. So they're not just large
language models off the shelf. You post train those models to
get more accurate, OK, it's based on that 4.6 billion
interaction data set. It's based upon to the earlier

(43:09):
questions, deeper vertical information.
And we're able now with knowledge bases and pulling
things from Confluence or SharePoint or others, we're able
to get even more knowledge and more data.
But again, your models and your answers are only as good as your
data. So I think, again, he's hitting
on something really, really strong that you want to get as

(43:30):
much data ingested, but you want, even if it's unstructured,
clean scrub data that has some effectiveness in it.
And so I think we're getting to a world where the more
structured and unstructured datathat you have that's scrubbed,
the more precise with post training you can do to give
answers to customers or employees or other businesses

(43:54):
more accurately. And so I do think we're in a
situation where data is really, really important.
We've gone through a whole thinginternally at Zendesk the last 2
1/2 years that wasn't that sexy.But we've spent a lot of time
with our internal data, scrubbing it, putting it into
the data lake so that we could go use it in a bunch of
different AI applications. Until we did that, we kind of

(44:15):
had a bottleneck on using our data.
I'm not talking about our customer data.
I'm talking about internal data to go run the business.
I think smart companies have taken this on as a project
because it's not fun, it's not sexy, but it's going to have a
dramatic impact on how a business operates in the coming

(44:39):
years. It's the foundation.
It's the foundation and there's some really cool use cases right
now. I've been talking a bunch of
startups on reporting and analytics and we just acquired a
company to do some really cool stuff in reporting and analytics
that you're getting to the pointwhere instead of getting a
report, you can go natural querywhat is the year over year

(45:01):
increase in automated resolutions in Malta for our
customer base. And you can go get that data
with natural language queries that can give you really, really
insights. And you can go prompt reporting
analytics to give you more forward thinking things.
What should I be thinking about to run my business that I'm
concerned about Malta? And I'm concerned about how our

(45:22):
customers are getting value fromour product.
And you can get prompts and ideas back.
And I think that's again, shifting from old static reports
to real insights. And again, you can do that
through natural language. Brandon Byrd on LinkedIn says
when you increase self-service levels, how do you address agent
burnout? And I'm going to ask you to
start answering these, Tom relatively quickly because I

(45:44):
want to get to everybody and we have, we have a lot of questions
and I have my own questions. You can self-service, you're
answering the easy questions andhuman agent burnout is real.
It's one of the most stressful jobs possible.
I think you can help them get better at their job through
Copilot and so that you're serving them the answers, which
I think can go really help with burnout because they're getting

(46:07):
more accurate, more empathy, andthey're getting it right with
customers. Here is a question from Preeti
Narayan, who says in her region there is a lack of Computer
literacy. And what can a company like
Zendesk do to make those computer interactions easier for

(46:31):
folks who are not computer experts?
We have a foundation and we havea program.
We, we've trained over 10,000 people that might not have the
same kind of access. There are, there is in other
parts of the world to become customer service agents, human
agents and giving them the computer skills and the customer
skills. And so that's one of the things

(46:52):
we're doing at Zendesk. You know, doing good, doing
right is doing good. And I think, you know, more and
more companies are going to haveto figure out how to help people
become more computer, a little bit more customer service
literate, and become great customer service agents.
Tom, how has this AI trajectory affected how you think how

(47:14):
Zendesk thinks about talent, team composition and jobs?
We haven't gone as radical as other people.
I know they've said, hey, you need to justify any job
requisition with why AI cannot do it.
But when we talk to talent now, I know one of the first
questions I ask is how are you using even for CXC suite

(47:35):
employees? How are you using AI in your
everyday life, you know, professionally and personally?
Because in this AI era, if you're not using the tools, if
you're not experimenting, you'renot getting better.
I don't think you're going to have the skills and capabilities
to succeed. So that's one of the things that
we've talked about. And then we're going to start a
big push. We've kind of had a bottoms up
approach up until now about how our employees use AI and we've

(47:59):
opened up like an enterprise license for Chachi BT I use it
everyday. I think you're going to see a
little more top down pushes to encouraging employees to use,
you know, these tools more because I think it's good for
the company, it's good for theircareer.
Can you talk about the future and where is AI customer

(48:20):
experience headed? We believe at Endesk that 80 of
interactions will be through automated resolutions in five
years. So do not involve a human in any
way, shape or form. We think 100% of interactions
will involve AI in some way, shape or form within five years.
That means that last 20% that humans are solving, they're

(48:41):
going to do that with the assistance of AI.
We do think there's another trend though that once you start
getting quick satisfaction, it'sthe opposite of dealing with
that chat bot. Michael, you and I talked about
earlier. We think consumers, employees
and other businesses are going to interact with those companies
more and more and have more interactions.
So we think interactions are actually going to explode,

(49:03):
particularly when you start using your own AI agent and
tasked them with doing something.
And so those are some of the trends that we are going to see
in AI over the next five years. What about the impact of AI on
jobs in general across the economy?
Is that something that you have thought about at all?

(49:24):
Where, where is this going from A from a jobs?
What jobs will be affected? What, what won't?
And so forth. Definitely thought about it and
I'll give more of a personal story and I alluded to it
before. I have a 22 year old daughter
who just recently graduated fromcollege and I have a 20 year old
son who's in college and they'veasked me with the advent of AI,
what kind of what should what job should I be doing?

(49:45):
I think an answer would have been 15 years ago going to
coding, right? Software engineering is a great
place to go. I'm really telling them that
again #1 be really, really knowledgeable about AI.
Use it in your personal life, use it in your professional.
That is going to distinguish youfrom other people #2 I tell them
it's going to go back to the basics.
If you can read well, you can write well, you can analyze

(50:09):
well, you're going to be fine. Because no one can predict what
the job markets going to be evolve over the next 40 to 50
years. And that's kind of their career
trajectory of, you know, young 20 something people is what's
going to happen in the next 40 or 50 years.
And we've seen what's happened in the last, you know, three
years with AI. So really encourage people to
get those basic skills. And 3rd, I encourage them to

(50:30):
learn how to prompt. And it's kind of tied to 1 and
2, how to use AI. And I think I've experimented
the last two or three years and my prompting is getting better
and better. I think that skill is going to
be great. And it's kind of interesting.
I think some of the liberal artseducation, when you write well,
you analyze well, you read well,is going to be more and more
important in prompting. On the prompting topic, folks

(50:52):
listening, we did a show specifically on prompting for
business leaders. So check out cxotalk.com and go
back a few episodes and you'll find that show.
It was really, really great very, very quickly.
Data privacy and security, cybersecurity of your data.

(51:12):
Can you talk about that? Because that's obviously it's an
extremely important topic. We take data privacy and it's
kind of that white box approach really, really seriously.
And so we are out in the forefront right now on trust,
the amount of certifications that we have to make sure that
we have a lot of confidence thatwe're going to protect with our

(51:32):
customers. Your data is really paramount to
us. 2nd, we want to make sure that we can personalize things
to you, but not in a creepy way with companies.
And so we work with companies onthat.
But I think we're going to be inan era where that white box
approach versus the black box approach and exposing that not
just to companies, but employees, consumers and other

(51:55):
businesses on how people are using your personally
identifiable information, how they're securing that is going
to be more and more of a topic over the next 5 to 10 years.
And but it's absolutely the fourfront of our AI approach.
And with that, I'm afraid we're out of time.
Tom, thank you so much for beinga guest on CXO Talk.

(52:17):
I'm so grateful to you and for your team at Zendesk.
Great questions, great flow. Really, really appreciate your
time, Michael, and I really appreciate the opportunity to be
on with you today. Folks, before you go, now would
be an excellent time. To subscribe to the CXO Talk
newsletter. Go to cxotalk.com.
We'll keep you updated on upcoming shows.
Have a great day everybody, and we'll see you soon.

(52:39):
Take care.
Advertise With Us

Popular Podcasts

Stuff You Should Know
Dateline NBC

Dateline NBC

Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Special Summer Offer: Exclusively on Apple Podcasts, try our Dateline Premium subscription completely free for one month! With Dateline Premium, you get every episode ad-free plus exclusive bonus content.

On Purpose with Jay Shetty

On Purpose with Jay Shetty

I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!

Music, radio and podcasts, all free. Listen online or download the iHeart App.

Connect

© 2025 iHeartMedia, Inc.