Declan Ivory, VP of CS at Intercom - AI and the new age of customer support

In this episode, our host Anika Zubair chats with Declan Ivory, VP of Customer Support at Intercom about how AI is transforming the customer support department. 

The future of customer support departments is rapidly approaching. So, what changes can we expect in 2023 and beyond? What skills will support agents need to keep up? And what key performance indicators should we be tracking when proactive support becomes the norm?

Podcast enquiries: [e-mail]

Episode Transcript

Speaker 1 (00:04):
Hello everyone, I'm your host Anika Bert , and
welcome back to the nextepisode of the Customer Success
Channel podcast, brought to youby Plan Hat , the Modern
Customer platform. This podcastis created for anyone working
in or interested in thecustomer success field. On this
podcast, we will speak toleaders in the industry about
their experiences and theirdefinitions of customer success

(00:27):
and get their advice and bestpractices on how to run a c s
organization. Today we will bespeaking with Declan Ivory, who
is the VP of customer supportat Intercom. He's an
experienced senior leader witha passion for building and
developing high performingteams and applying digital
technologies to supportorganizations through major

(00:49):
business transformation. Priorto his time leading Intercom's
customer support team, Declanhas held senior support
leadership roles over the last10 years with Amazon Web
Services, Tableau Software, andGoogle Cloud. And today we will
be chatting with Declan allabout the transformation of
customer support in 2023 andhow AI will be changing the
landscape of customer supportin the future. Welcome Declan

(01:12):
to the podcast. I'm so excitedto have you with us here today.
Before we get into today'stopic and into the depth of our
conversation, I would love foryou to tell us in your own
words a little bit aboutyourselves, what it is you're
doing at Intercom, how youstarted to work in support.
Just give us a little bit ofbackground about who you are
and what it is you're doing.

Speaker 2 (01:34):
Hi Anika , delighted to be here and thank you for
the opportunity to share sometime with you. Uh , a little
bit about myself. So DeclanIvory, I'm VP of support at
Intercom. So I kind of maybetalk a little bit about
Intercom at the end, but talk alittle bit about how , how I
got to Intercom, which is kindof how I got into customer
support . So I'm an engineeroriginally by profession, but
went to work in the techindustry straight from college

(01:55):
and pretty soon I kinda had arealization that ultimately I
only get paid at the end of themonth because of customers. And
if we don't keep, if we don'tkeep our customers happy, then
you know, it , it's bad for thebusiness. So from very early on
I've had this kind of verystrong customer obsession,
customer focus , and generallyI've had support related roles
literally for the , the entire35 years that I've been in the

(02:18):
tech industry from kind ofbeing, you know, a technical
specialist originally, thenmoving into kinda managing
internal support teams and thenmanaging external teams
initially in the telco, butthen moving on from there to
work for the likes of AmazonWeb services, Tableau software,
Google Cloud. And I've ended upin an intercom at a pretty
interesting time. I think interms of intercom's evolution.

(02:42):
So Intercom is a , an AI firstfull function customer service
platform. And for me it'sreally interesting cuz I get to
use the same tool that I'msupporting from my customers.
So to some extent I experiencea lot of the, the challenges
and issues firsthand that mycustomers experience and it
means that I can be a verygenuine voice of the customer
back to the rest of theorganization and really

(03:03):
influence the , the productroadmap. And that's what really
makes this role in Intercom Ithink probably more interesting
than any other role that I'vehad. Cause I can have a very
significant impact on theproduct roadmap.

Speaker 1 (03:13):
Awesome. And I love what you just said, that you
realize really quickly thatkeeping customers happy is how
you get your paycheck at theend of the month. I think that
that needs to be like a , aposter that goes up into the
offices. But you mentioned thatyou had an engineering
background and you kind of justjumped into that right after
school, but then you've nowended up in customer support.

(03:34):
What kind of spurred thattransition? Why customer
support? Why not into thebackend side of things? Why,
why customer facing ? As

Speaker 2 (03:45):
I say , I've always been kind of obsessed by
customers and realizing that,you know, any business doesn't
have any value without theircustomers. So I've always
wanted to make sure that, youknow, delivering well for
customers. So I really prefercustomer facing roles than
non-customer facing roles. Uh ,apart from anything else, like
when you deal with customers ona day-to-day basis , you really
get to understand what are thereal challenges and issues that

(04:06):
they have. And because I'm anengineer, I've kind of a , a
very strong problem solvingkind of background. So I love
to solve problems forcustomers. That's ultimately
what gives me the mostfulfillment and and
satisfaction. So that's why I'min customer support. Cause
every single day there'sproblems to be solved for
customers. ,

Speaker 1 (04:23):
I completely believe you and I've been in a number
of organizations where theonboarding of a new employee
has been to work in customersupport for a day and to
actually see tickets and whatit looks like from a real
perspective of what ourcustomer's real problems. And I
think it's a huge learning andI think that, like you just
said, there's a problem toalways be solved and you're

(04:43):
always upfront and close withwhat your customer actually
wants from your product, whichI think is key. And you
mentioned that you also useIntercom at Intercom, which is
great. I've been at a a feworganizations where we do that
so we can really understand andlive and breathe our, our
product and what our customersare doing. But tell us a little
bit more about thisorganization at Intercom, what

(05:04):
you're building, what, whatyour remit is. I'm sure it's
quite large Intercom being asupport tool as well. But let
us know what is it , uh, thatyou're doing there and, and
what does your team look like?

Speaker 2 (05:14):
Tell me a little bit about Intercom. So it , it's an
Irish founded company, about 11years old and its mission , uh,
was to to make internetbusiness personal , uh, but
very focused on how businessesengage with their customers
with a particular focus oncustomer service, customer
support. And at the moment ourfocus is to really build an AI
first , uh, full functioncustomer service platform that

(05:37):
is considered best in class inthe market. And you know, the
philosophy in Intercom hasreally been to use innovative
technology throughout itshistory. So a lot of, you know
, uh, the , the Messengertechnology was quite new and uh
, Intercom product to themarket. Uh , Intercom was very
quickly into chat bots and ,and evolving there , applying
machine learning techniques aswell within the product and has

(05:58):
obviously adopted some of therecent changes in ai,
particularly generativegenerative ai and a particular
, uh, chat gpt and GPT four .
In fact, we were the firstcommercial organization to
announce a product after GPTfour was launched as would
error . We , we had a productannounced, so we'd be very
focused on innovation withinthe , the product set . So for
me then, from a customersupport point of view in

(06:21):
heading up the customer supportteam, which is a global team,
it's based out of Dublin,Chicago, and Sydney. So we can
provide full follow the sunsupport for our customers, but
we, you know , get to use thetechnology, which is great. You
know, we , we uh, can be earlydoctors of the technology but
we also then have to make surethat we have all the skills in
place to support the , uh,complexity of the environment.

(06:44):
Cause it is a complexenvironment but it's complex
because it provides a lot offunctionality, a lot of
capability. And again, it meansthat I have to really focus on
uh , on the skill sets withinthe team, making sure that we
can handle all of the variouskind of , uh, complexities
within the product itself andalso how our customers
integrate that product withother tools and other systems
that they have. So it's reallyinteresting space in that point
of view. Uh , there's a widevariety of kind of , uh,

(07:07):
customer questions and issuesthat we handle on a day-to-day
basis. So, you know , part ofthe role is, is ensuring that
we are upskilling the team allthe time and that we have, you
know , full visibility andknowledge of all of the product
changes that we're bringing tomarket.

Speaker 1 (07:20):
Awesome, awesome.
And customer support issomething that isn't new but I,
like you mentioned, there's somany different facets to it and
it's evolving so much and thereis a real transformation
happening in customer support,whether it's with AI or with
other parts. But I know we'regonna talk more in depth of
that, but I think support hasalways been seen as support

(07:40):
agents answering tickets,triaging tickets, you know,
reacting to something that'sgone horribly wrong within the
product or you're always firebiting and you're always trying
to like just answer somethingjust in time before a customer
really explodes at you. Which I think is
traditionally what everyonethinks of support. But I know
that we're here today to reallytalk about the transformation

(08:02):
of support, what it isbecoming, how AI is changing
that, how your organization ischanging. And you've obviously
seen some of that change happenat Intercom recently, but I'm
sure with the time you've beenthere, what are some of the
first steps that you took whenyou moved from let's say
reactive to more proactivesupport? Cause I know that's
what you guys are doing there.
The

Speaker 2 (08:21):
Big challenge is to change your mindset. And you
kind of touched on it, youknow, very much traditional
support is around almost atransaction driven environment.
Like you've got an issue comingin, you gotta solve it , you're
trying to hit an average handletime target, you're trying to
just close out the , the , theissue. And when you have that
kinda transaction focus, you'renot really looking at it well
what is the overall customerexperience? What is this
particular customer experiencein using my product or my

(08:44):
service? Right ? And that's thebig switch that I've tried to
make. You know , try , try tomake sure that people are very
much focused on what is theoverall customer experience and
, and one of the benefits of aplatform like incom, you do
have full visibility of all ofthe customer interactions and
once you have that visibility,you get a really good sense
then of what has the experiencebeen like for this customer.
And, and really getting peopleto take that customer

(09:05):
perspective rather than atransaction perspective is
probably the first thing thatyou need to do in moving from
reactive to proactive. Nowmoving from reactive to
proactive, you also gotta makesure that you're enabling your
support team to understand wellwhat's the next kind of bit of
advice or guidance they cangive to a customer that's going
to be kinda proactive. And asan example, we have implemented

(09:27):
what we call customer milestoneframework where we actually
track how well customers haveadopted our product and you
know, there are kind of what wethink are best in class steps
around how a customer should beusing the , the product. And we
actually make sure that thatinformation is available to
anyone on the support team whena customer opens up an issue or
a conversation. So basically wehave an app built into our

(09:49):
platform, it bends all thisinformation around where the
customer is and the customermilestone framework. We have
predefined actions that a , asupport agent can take to say
to the customer, Hey, I noticedyou're at step, you know, two
outta three, oh , here's whatyou need to do to get to step
three in terms of enabling thisfunction or this capability.
And that kinda almostconsultative practice support

(10:10):
is really valuable to , to thecustomers. Cause in providing
it, you're very often, you know, ensuring that there isn't
gonna be an issue down theline. So from our our point of
view, we are probably avoidinga contact down the line. We're
helping the customer acceleratein terms of adoption or use of
the platform , which ultimatelydelivering value to them from a
business point of view. Causeyou know, they've made an

(10:30):
investment in the platform,they wanna use the
functionality and sometimesthere are blockers to them
doing that and having thiskinda more proactive approach
and particularly where you'relooking at the overall customer
experience, you can really addvalue in that context and you
move from this transactionmindset to actually a customer
experience mindset. That's

Speaker 1 (10:48):
Such a great thing to share, but also so simple in
its nature. I think that wealways think of like support,
like you said, transactionallike problem solution. And I
think that that happensnaturally when you're in that
mode of like, I'm presentedwith something, I need to
figure out the problem and thesolution to it. But I think
that the full customer team nowI'm used to working on the

(11:10):
customer success side ofthings, but the full customer
team, anyone who is customerfacing , if you are looking at
the full journey, like you'remapping out and really
understanding, hey, if you arewith our product for six
months, these are kind of thetendencies that you should
have. These are the use casesyou should be doing.
Understanding that wheneveryou're customer facing ,
whether you're an accountmanager, a customer success

(11:32):
manager , uh, a renewalsperson, an upsell person, a
support person, I think itjust, it makes a big difference
to change that mindset of,okay, I'm here to do my job
versus I'm here to make thefull experience flawless for
the customer. And I think weget so fragmented in the
customer journey and and and inwithin an organization as well

(11:52):
. Well if you think about it,we're all, you know, there to
do our day job and we almostforget what is that customer
journey or like you said thecustomer milestone framework.
Can you give us a little bitmore insight? I know it's hard
to just talk about thingswithout examples, but what does
that framework look like? Whatare those milestones? How did
you guys come up with that?

Speaker 2 (12:11):
Again, it was strong collaboration across the
organization. Like , uh, youknow, a lot of it was driven by
the customer success team who,you know, deal with our
customers on an ongoing basis.
And also we have onboardingspecialists as well. So a
combination of onboardingspecialists, success managers,
and support people, we wereable to kind of take a view of
well, you know, what does goodlook like in terms of how a
customer is using our productand what's the expectation of

(12:33):
what features and capabilitiesthey should be using after, you
know, two months, three months,four months, et cetera . So you
can actually map out veryclear, you know, to activate
the product then to get valuefrom it, then to have a kind of
mature implementation. Andwe've been able to build out a
whole framework around what'sthe expectation of where a
customer should be at a pointin time. So it's really looking
at, you know, it , it may bewhat particular features are

(12:55):
they using? So have the , youknow, in the first few months
there'll be very basic featuresyou'd expect them to use like
the inbox, the messenger, etcetera . Then over time you
expect them maybe to start touse some of our AI capabilities
to drive automation. Maybe someof our , our more complex
features like workloadmanagement, et cetera . So you
have all that mapped out interms of what's the , the
timeframe for our customer,like what level of activity

(13:16):
should they expect after aperiod of time based on the
number of seats that they have. And so there's a whole lot of
different metrics to feed into, basically say this is what a
healthy customer looks likeversus this is what basically,
you know, unhealthy looks likeand we probably need to help
this customer and , and , andencourage them to move further
down the the adoption path.
Cause if say ultimately it'sabout them getting value from

(13:37):
the investment that they'vemade, right? Uh , and you know,
if they're not using featuresand capabilities, they're not
leveraging the value thatthey've made the investment for
. So, you know, it's reallyimportant to give value to the
customer through this process.

Speaker 1 (13:49):
Yeah, I love that.
And it comes right back to thatcustomer journey that we're
talking about and it , theimportance of all of that and
just the importance of comingback to that, it's as simple as
that. Sitting down withcross-functional teams mapping
out what the heck your customeris doing with your product and
what you would expect them todo in an ideal world, how would

(14:10):
they perfectly use your productand then work backwards from
there. Or like you said, fillin the gaps based on each
different department and howthey can help reach those
goals. And I think that if youever get confused or if you're
ever unsure how to really beproactive with your customers,
really map out the perfectjourney for them and then work
in that way. And um, I thinkthat you've talked a little bit

(14:32):
about how you've shifted thatreactive to proactive side of
things, but I think thetraditional way of delivering
support is like get a ticket,enter a ticket. That is, that
is what I think of and that isall , most of my career. How
support agents work. Is theresome ways that you guys are
changing that mindset atintercom or doing different
ways that are supportingcustomers that are outside of

(14:55):
those ticket answering ways?

Speaker 2 (14:56):
Maybe take a step back. So you've described, you
know, the whole transactionpiece and you're waiting for
things to come . It's almostrelentless from a support point
of view is the term I use. Andpart of what you're trying to
do is how do you free up timefor your support team to
actually spend time with thecustomer , uh, in a more
consultative way. So for usthat has been really looking at

(15:18):
what workload is our teamdoing, what workload is so
valuable and important that itneeds the skill and expertise
of a human to handle it and itneeds that level of empathy and
interaction with the customerto really, you know, give value
versus what part of ourworkload should we be actually
solving within the product IDissues should never occur in
the first place or what part ofour workload can we actually

(15:41):
automate away ? Whether that'strue kind of , you know ,
traditional machine learningtechniques or whether it's
using AI kind of advances thathave happened in the last few
months. So we've focused a loton basically taking work outta
the system, you know, througheliminating it originally or
handling it in an automatedway. So we actually free up
cycles for our support team tospend time and , and focus on

(16:02):
understanding, I say where thecustomer is and being able to
think about what's the, thenext best action to recommend
to the customer to kinda , youknow, get , get them more value
of the product. So freeing uptime has been the most
significant thing that, that wehave done. Cause you know, it
is relentless and aligned withthat . We , we are now , we're
in the middle days , we haven'tdone a bottom yet. We need to

(16:23):
change our whole view ofmetrics as well for the support
team. Like if you just leaveyour traditional metrics in
place, then you're not reallygetting your team into a space
where they feel they can spendtime and effort consulting with
the customer and maybe taking alittle bit longer than they
might have done in the past. Sothat is really important that
we think about metricsdifferently. And I know we're

(16:43):
gonna move on to talk about AIas well. But you know, the
other thing that's reallycoming to the fore as well is
that you want your support teamactually to build a lot of
knowledge as well based on howthey interact with our
customers. You know, ultimatelythey should think of every
customer interaction they have,ideally will be the last time
that particular issue occurs inthat you drive through root
cause you develop eitherknowledge or a solution to make

(17:05):
sure the problem doesn't happenagain. But again, to free up
your team to do that means yougotta get bandwidth, you gotta
make sure that they have timeto do that and the reward is
for it as well. Like themetrics should take account of
that. And then, you know ,there are some skill sets as
well that they need to be ableto do both the consultative
piece and the knowledgemanagement piece as well. So
there's a lot changing in thisenvironment and a lot of it is

(17:27):
driven by some of the advancesthat we've seen recently in the
whole AI space.

Speaker 1 (17:31):
Yeah, and I wanna get into that, but one more
question around just supporttickets and how you're freeing
up time for your agents and ,and what you guys are doing
there. Is there a way that youare being consultative when
you're either closing out aticket? I'm just wondering
because even a customer thinksof a support ticket
transactionally, are you doinganything different when you're

(17:53):
closing out a ticket, whetheryou're answering a ticket, any,
anything that you're doing toalso shift the customer's
mindset? Because I think it's,it's a two-way street in the
sense that a customer sees itas transactional. How are you
changing the customer's mindsetwhen it comes to support ticket
and reacting to, to thosetickets?

Speaker 2 (18:10):
So I , I think the customer milestone framework
that I mentioned earlier, it'sa two-way street. Like , you
know, the customer has toengage with, you know, the, the
interaction from the supportteam member as well. So it is
actually encouraging thecustomer to think slightly
beyond the current problem. Uh, you know, so that's how we're
changing, you know, customerbehavior over time as well. Uh,
like we typically have alwaysclosed out our conversations

(18:32):
with a very open-ended, isthere anything we can do? Is
there, you know , any otherissues that , so we do invite
the customer to go beyond thecurrent transaction. We've kind
of done that historically, butnow we're actually putting some
meat on the bones for what abetter way describing it where
we're actually sharing withthem, hey, here's something
that we think you, you mightwanna look at or that , you
know , maybe you're looking atalready, can we help you with

(18:53):
it? So you're beginning to makeit a little bit more specific
then for the customer cause ofthis information that we can
now provide to the support team. Something like the customer
milestone framework is atwo-way street . Like the
customer has to engage with it, has to understand it , and I
think it opens up their mind aswell then to really think
beyond the current transactiondoes take time to sh to change
people's minds. Like, you know,we don't get full engagement

(19:15):
with it from all our customers.
Uh , we're all the timethinking about how , how can we
make that engagement a littlebit better, get more of our
customers to , to uptake , youknow , the recommendations
we're making. We still gotpretty strong performance outta
it, but we're all the timetrying to tune it and see, you
know, how can we get ourcustomers to engage more, more
proactively with us as weengage proactively with them.

Speaker 1 (19:33):
Awesome. And just one other question on that, as
you guys are being moreconsultative, I'm just thinking
how that's almost crossing alittle bit into the CSM land,
but I'm sure that there'sboundaries around what support
is doing versus what a customersuccess manager is doing versus
the onboarding specialist youmentioned. Is there ways that
it almost crosses boundarieswhen you're becoming almost too

(19:55):
consultative and you passing itback to another team member?
How do you guys manage that sothat it's not, let's say
confusing both to internalteams but also to the customer
of who they should beinteracting with? Yeah ,

Speaker 2 (20:06):
It it's a , it's a really good question and I , I
don't have a complete answerfor it to, to be honest, right?
Because I think there areblurry lines between customer
success, onboarding andsupport. I think as we move
into this particularly AIdriven world, I think there
there's more activities fromthose three domains can
actually be automated, canpotentially be handled by a
single human, you know ,somewhere along the line. But

(20:28):
then what it's doing is it'sfreeing up other people,
whether it's an onboardingspecialist or a customer
success manager to actuallyundertake the activities that
are really value adding fromthe customer perspective. Cause
again, if you look across thosethree domains, there are
activities that are kinda moretransactional in nature that
are , you know , very open forautomation and very open for,

(20:48):
you know, part of the activityto be driven by someone in a ,
in that consultative supportrole as opposed to having to be
handed off to a onboardingspecialist or a success
manager. But it doesn't reallyundermine the, the value that
those two roles provide for ourcustomers. But it is a little
bit blurry and I think as youknow, predict this technology
evolves. There will beactivities that can be

(21:09):
centralized into a singlehuman, you know, whether we
call them a support person orwhat our service person, I'm
not sure what we call them ,but they'll be able to
undertake maybe more activitythan they can do today. But
there's still this high touch ,high value piece that success
managers deliver and onboardingspecialists deliver.

Speaker 1 (21:25):
Awesome, awesome.
And we've touched upon thisalready a few times and I wanna
dive much deeper into itbecause I think when everyone
thinks of AI or machinelearning, they think of it
almost replacing support or youknow, what's the point? We can
have something like a bot or anAI machine learning person
that's gonna come in and reallytransform support. Now I know

(21:48):
you have your views on this,but how are you guys utilizing
AI at intercom with the supportteam where it's something
that's supporting the supportteam rather than replacing?
Yeah,

Speaker 2 (21:58):
No re really good question and , and maybe a
little bit of context. So likeI've been involved in support
for a long time and I've beenreally frustrated by the slow
pace of the support industry toadopt technology, particularly
when it comes to , uh, machinelearning and , and AI et cetera
. And I think what has happenedin the last few months, the the
technology has matured to adegree that you can now

(22:20):
actually think aboutimplementing it pervasively,
but it's not implementing it toreplace humans, it's
implementing it to complementthe human support experience.
Cause it's different dimensionsthrough which AI can be
applied. And I'll can talk alittle bit about what we're
doing here at Intercom, butthey're , they're three
dimensions for me. And thefirst dimension which gets most
focused is customers dealingwith AI via some kind of bot uh

(22:43):
, experience, right? And , andthat's definitely a , a really
good use case for , for AItechnology and there's lots of
very simple transactions thatcan be automated through this
approach and actually customergets an answer far more quickly
than they would ever get ifthey had to go through to a
human support person. So that'sa real advantage for the
customer and it's a realadvantage for the support team

(23:06):
cause it's freeing them up fromall those kind of mundane
routine questions that theykinda get, they they by day
that aren't very fulfilling toanswer and are kind of
consuming bandwidth that theycould be using more
beneficially for the customer.
So that's kinda the AI from thecustomer lens. Then there's AI
from the , uh, teammate lens aswell because AI can assist

(23:28):
teammate . Even simple thingslike summarizing a very long
case or a very longconversation, you can use AI to
summarize it . Reallyimportant, if you're doing a
follow the sun model and you'rehandling handing over from one
support agent to another, itcan change the tone of
response. You might craft aresponse, but you might wanna
make it more formal. It canchange the tone of the , the
response to be more formal. Youmight wanna make it more
casual, you can change it to bemore casual. You can take

(23:50):
bullet points and expand themout into kinda a , you know , a
more comprehensive reply for acustomer. You can use what we
call smart replies where, youknow , AI can suggest to a
support agent here's, you know,possible answers to the
customer question, not exposingdirectly to the customer but
allowing support agent to takethat as a guide and help them
craft the final response to thecustomer. So AI helps the, the

(24:12):
support agent as well. And AIalso helps, you know, people
who are running support likemyself, support leadership and
management in that, you know,using AI for example, you can
do very comprehensive analysisof all your cases or
conversations. You can gain alot of insights, you can help
pinpoint where you need todrive improvements. You can
generate things like sentimentanalysis across every single
case, every single conversationyou have and not just relying

(24:34):
on CSAT responses. So it doeshave the opportunity to really
transform how you deliversupport. And people do focus a
little bit on, well, you know,it's gonna drive cost
efficiencies the way I term itis it changes the economics of
delivering support is what thistechnology does. Uh , but it
actually frees up a lot of timeand bandwidth for your support
team to actually add a lot morevalue from the customer

(24:57):
perspective and actually becomereally value adding from a
business point of view as well,as opposed to many
organizations which viewsupport as a little bit of an
overhead today cuz it istransaction focused . So that's
kinda a , a kinda philosophicalview of , of where I see the
technology and and where it canbe applied here at Intercom.
You know , we, we've done twothings. So pretty soon after

(25:18):
chat g PT was launched, we kindof assessed the technology and
kind of said this is actually afundamental change in terms of
how AI works and it's somethingwe should really try and
understand where does it playwithin our own product and how
can we deliver value to ourcustomers? And the first place
where we looked at it was whatwe called AI in our inbox. So
inbox is where we managedconversations with our

(25:40):
customers. So we launched AIfeatures, I think it was back
in February this year, whichincluded that text , our case ,
uh, summarization, ourconversation summarization
piece, the tone change , uh,tools as well . And I was
really delighting, my teamdidn't view that as something
that was inhibiting them in anyway. In fact they, they felt
they were being empowered byhaving these tools available.

(26:02):
They actually felt it , it madethem more productive, et cetera
. So we got really positivefeedback from our team in using
the AI in the inbox tools. Andthen we decided to launch an an
AI bot based on GPT fourtechnology. And again, my team
were very instrumental intesting the technology before
we ever delivered it to, to ourcustomers. Uh , we were able to

(26:23):
provide a lot of feedback tohelp shape the technology and
basically what it has done isit has allowed us to, because
we, we were kind of basicallythe first beta customer for it
. Uh , we were able to deployreasonably quickly cuz this is
the only thing that, you know,customers have. I want to be
kinda , you know, veryconsiderate about how, how I
apply this. And yes, you doneed to be considered about how

(26:43):
you apply it , but in our case,like we re trialed with one
customer segment, we saw thepositive reaction from our
customers. We saw the positivebenefit in terms of resolving a
lot of our mundane work , uh,automatically. And we rolled it
out literally within a matterof days to the segments . Now
the one thing about the waywe've deployed technology is
like , you know, AI or Chachi ,bt it can hallucinate, right?

(27:05):
It can make stuff up andclearly in a customer
SupportPoint or uh ,environment , you don't want to
make stuff up for yourcustomers. So we've kind of
constrained our AI bot in apositive way. It will use the
knowledge space that you havein your system. So in our case
it's our help center. All ofthe information in the help
center is used by the AI bot todetermine an answer for the

(27:27):
customer. So we've constrainedit to a knowledge base that is
already verified, you know,that you know , is high quality
and ensures that theinformation being provided to
customer is actually accurateand it , you know, basically
reduces or close to eliminatesthe opportunity for the
technology to hallucinate. Sothat's one of the, you know,
the key guardrails that we putin place. Cause you know, I'll

(27:49):
say when chat BT came out andthe large language models it ,
the , they do invent stuff up,right? It's just the nature of
how the technology works andputting those guardrails around
it has been really fundamentalfrom our point of view. And I
can very simple terms likewe've deployed the technology
and , and without too muchtuning and without kinda really
thinking about , uh, augmentingour knowledge base in , in , in

(28:10):
any kind of comprehensive way,we've taken out about 25 , 30%
of our kinda mundane workloadliterally out of the box. So
that is transformational. Youknow, the stuff that we'd done
before was kind of , you know,you take out two or 3% of your
work, you know, through aninitiative, but to have an
initiative where we'vebasically been able to deploy
something in a matter of daysand take out 30% of your work,

(28:31):
that is transformational.

Speaker 1 (28:33):
Definitely.
Definitely. And I love howyou've already highlighted how
ai, machine learning chat sheetpt , like everything that is
happening in our world rightnow is something that's
complimentary, something that'shelping , something that is
transforming the way we work,but in a way for delivering
support in a better way. Iwould say, I think when
everything came out around ai,especially when chat c p t

(28:55):
like, you know, went viral andeveryone's like, what is this?
What are we doing with this?
Everyone was so fearful thatthis is gonna replace our jobs.
It's going to suddenly takeaway from our day-to-day or
make our jobs redundant orjust, you know , fear of, you
know, technology replacinghumans. But like you said, you
have to put up the guardrails,you have to decide how AI is

(29:16):
gonna work with your product,with your customers, with your
team in order to see itsuccessfully transform your
department. Which I think is socritical for anyone who's
trying to implement AI orthinking about using AI in
their customer supportorganization. I think it's
something that's complimentary,but something that you have to
decide how you use it and howit's gonna empower your team.

Speaker 2 (29:39):
Absolutely. And I say it changes the economics of
support as well. You know, anexample in the past, like I've
been trying to scale supportteams and literally you could
not hire people fast enough,right? And yeah , and with this
type of technology you canscale your business without
having to scale your supportteam at the same rate. Like
that's a really fundamentalchange in the economics of
support as well.

Speaker 1 (29:58):
Definitely. Yeah.
And I think that it's somethingthat is very relevant in
today's economy in today wherebusinesses are really looking
at every dollar spent and alsohow can you optimize every
dollar spent? And when you havesomething like AI driven
support and AI behind the wayyou're interacting with
customers, like you said, youcan be bringing on tens of

(30:19):
thousands of customers acrossthe world, but you guys are
lucky enough to follow the sunmodel, but some people might
not be. And how can youimplement AI to be able to
compliment that as you grow?
It's not something that'llreplace everything down the
line, but as you're scaling andgrowing have like a backup plan
with AI in place,

Speaker 2 (30:37):
and , and the other thing , at the end of
the day, the technology is onlyas good as the knowledge that
it's fed, right? And we stillneed people to develop that
knowledge, develop thatexpertise, we need to have
subject matter experts andthere are many situations where
customers will want to engagewith that subject matter
expertise. So at at one levelthe support role is actually
going to evolve to be moretechnical, you know, you're

(31:00):
gonna need more subject matterexperts, you're gonna need
people with a strongertroubleshooting skills,
stronger customer engagementskills. So yeah , like the ,
the , the transformation isimpacting the human support
role as well in a very positiveway. The always gonna be far
more fulfilling for peoplegoing forward.

Speaker 1 (31:16):
Definitely. And you've kind of already segued
into another question I had,which is around the support
agent skills that are needed in2023. I think, you know, a year
ago very different skills wereneeded definitely 10 years ago,
but I think a lot of people seesupport agents as someone who's
transactional, someone who canproblem solve, someone who can

(31:38):
quickly answer a question. Andwe've already talked about some
skills that are, let's say,newer or transformative to a
support agent role likeconsultation and thinking
beyond the ticket itself, whatwould you say are the key
skills needed, especially withAI involved that a support
agent needs to have in 2023 ?

Speaker 2 (31:58):
Really good question. So I , I think the
first one for me istroubleshooting skills. Like
the, the issues that are gonnacome true to a support agent in
any business are gonna be morecomplex, more nuanced cause the
easier stuff has been takenout. So trouble, sorry,
troubleshooting skills aregonna be critical. Uh , and any
way that you can help your teamto hone and improve their

(32:18):
troubleshooting skills, that'sgonna be pretty key. Uh , the
second thing is that people aregonna become subject matter
experts in particular areas ofa product or service or your ,
whatever your , your businessoffering is. That's the second
thing. You know , rather thanbeing generalists, I think
people are going to developsubject matter experts in , in
components or parts of , ofyour service. And the third

(32:39):
area is really having, youknow, I call it kinda almost
like a curiosity mindset liketr trying to think through why
did this customer have to comethrough to me in the first
place? What was the issue andhow can I ensure that issue
doesn't happen again ? And alot of that will be around
building some kind of knowledgeartifact that goes back into
your help center that thenhelps ai, you know, resolve

(33:01):
more problems down the line.
There are kinda three of thekey areas that I see it's
around troubleshooting,becoming subject matter experts
and having that level ofcuriosity that you can actually
build and develop the knowledgethat ultimately feeds the AI
machine in the background.

Speaker 1 (33:15):
Yeah, I love those skills though because I'm just
smiling to myself thinking,wow, that's something that I
hire for in customer successtoo, like curiosity, subject
matter experts. So I thinkcoming back to those blurred
lines, it's interesting to seehow we're all becoming one
customer organization and howwe're able to service our
customer but with differentlike skill sets in mind. So I

(33:37):
really, I do appreciate yousharing those and as support
evolves and AI machine learninggets more involved, I'm sure
some of the KPIs that you guysare tracking are going to
change, like, like you said, tobe consultative. It's no longer
important to answer a ticket asquickly as possible close out
or resolution as clo quickly aspossible, but at the same time

(33:58):
you do wanna find a resolutionas quickly as possible. So what
is it that you guys aretracking from a K P I
perspective when you'rethinking of proactive support?

Speaker 2 (34:08):
It's evolving. Like we're we're in in the kind of
the process of changing yourKPIs and kinda more traditional
to, to different KPIs, youknow, as , as an example when
it comes to that proactivepiece. So we measure if the
customer that was engaging withthe agent, what percentage of
the time where there was aclear kind milestone
opportunity, did we actuallyprovide that opportunity to the

(34:33):
customer, right? So we , wetrack the percentage of time
that the agent will make arecommendation. Now there are
some situations where it's notappropriate for the issue to
make the recommendation causethe customer might be very
alright , right ? May not be ina kind of a , a mindset that
it's actually gonna beproductive to say, Hey, did you
think about implementing thisparticular feature? So there is
a judgment called by the agents, but we do track what

(34:54):
percentage of time where therewas an opportunity to make a
proactive recommendation. Didyou make a proactive
recommendation? So that's theone KPI that we track and then
we track of those customerswhere you made a proactive
recommendation, what percentageactually took action based on
that recommendation. So thereare kinda two metrics that we
track for that proactive piece.

(35:15):
Now we still have the challengeas you say , like you still
have to worry about averagehandle time and productivity
because you still havedimension, you're , you're
supporting. So we're actuallyin the process at the moment of
reevaluating all of our handletimes. Uh , because I say of
the work that we've taken out,the mix of work that's coming
into us is different and we'reactually relining all of our

(35:37):
average handle times righthere, right now at the moment
where we're in the process ofit. And we're gonna start to
build up with what what doesour new capacity plan look like
as we look into the future andwhat does it mean in terms of ,
of how we might pitch themetrics for our team . And
really, you know , we're gonnaend up with a bundle of metrics
. There are many supportorganizations have like almost
like a key metric, like, youknow, how many conversations

(35:58):
did you handle per hour? Howmany cases did you close per
week or , or whatever. And whatI'm trying to move is we have a
bundle of kind of metrics forour support team that will be a
mix of some of that transactionkind of mindset because there
still will be things comingthrough that you , you wanna
handle in , in , as you say ,in a faster manner as possible.
And then you wanna encouragethe proactive piece as well. So

(36:20):
the metrics have to, there willbe some tension in the metrics,
but naturally where you'reallowing almost agents or
support team members to makeautonomous decisions around,
okay, I'm gonna consciouslyhelp this customer . I know it
impacts one kpi, but it'sactually gonna improve another
kpi and I know, I know I'mgonna be measured on the bundle
of the KPIs now it's gonna takea lot of time to tune that and

(36:42):
get that right, but we gottamove away in , in the support
world from a single metric thatwe drive our team off to a
bundle of metrics where you'reallowing people allow a level
of autonomy to do the rightthing for the customer and it
doesn't impinge on theirperformance.

Speaker 1 (36:58):
Yep . I completely agree and I think I love those
metrics about looking at theholistic approach rather than
the quickest response. But youhave to find balance, like you
said, there's no way that youcannot have a support to get go
unanswered as quickly aspossible, but you at the same
time have to think of thebigger picture and that
framework that we talked aboutearlier. But again, just

(37:20):
thinking about how can we slotinto the bigger picture of
customer journey, how can weprovide better value and
service and how can we make ourcustomer use our product even
more by answering this ticketin that way? So thanks for
sharing all those pieces. Iwanna ask one more question
around the AI piece that I justcame to mind, but you were
saying how AI supports the, thecustomer agent and how it's

(37:43):
something that you have to putguardrails around something
that you have to really build,but a lot of people also see
bots and AI as something that'sunfriendly service or very un
like something that's nothuman, something that you're
interacting with that is trulya bot. You're, you know, that
you're not gonna get reallygood interaction with. I think
about it all the time.

(38:04):
Sometimes I'm chatting and I'mlike, can I just talk to
someone? This is this , thisbot is really not gonna do much
for me. But how are you guysfinding that balance of AI is
helping personalize customerservice rather than make it
more transactional? Yeah,

Speaker 2 (38:20):
It's a , it's a really good point and I think
it's one of the big benefitsthat's happened with the whole
large language model and thekind of G P T technology that
technology can obviously searcha lot of information and bend ,
you know, a response. I thinkthe key change in that
technology is the way that itengages with a customer. Like
our , anyone engages withtechnology, it is very

(38:42):
conversational. It will buildthe conversation with you. So
if you ask a question, it canask clarifying questions, right
? So straight away , e evenwithout thinking too much about
it, the , the nature of thetechnology is the conversation
or the interface with thecustomer has radically changed
and it does feel more naturalthan traditional chat bots. So
that's just one , one of the ,the benefits of how the

(39:03):
technology has evolved.
However, you've gotta look atthe customer experience end to
end , uh, because it can be acombination of a bot, it can be
a combination of human supportand it's not a case of, you
know, today in the supportworld we quality assured the
agent performance and we say,did the agent do a good job?
Like was the tone right? Youknow, did they , uh, were they

(39:24):
polite to the customer ?
There's a lot of things thatfeed into how you look at the
agent performance from aquality point of view and Peter
saying , yeah, we now need toquality assure the bot
performance. Yeah, you do, butlet's look at it differently.
Let's quality assure thecustomer experience end to end
. And let's think about fromthe time the customer decided
to engage with you to the timethe issue was resolved, what

(39:45):
was the customer experience?
And one we've gotta findmechanisms to measure that, you
know , uh, cause it's not justaround the csat , you know, if
it goes through , through humansupport, like what happens if
it's just handled by the bot?
What was the CSAT for that partof the journey? Um , also, what
was the flow like? Like wasthere a natural flow from, you
know , the bot to humansupport? Like that needs to be

(40:07):
seamless, it also needs toretain context. So everything
that the customers providedupfront should be available to
the human support agent. Andyou need , uh, you know, an
ecosystem behind , uh, thesetools to do that. Like you
can't just , uh, you can'timplement an an AI bot
isolation and not think abouthow it links each your
ecosystem. So all the contextof that customer is available
to the bot available to humansupport. And in terms of that

(40:30):
interaction with the customer,like we've hired in a new role,
which we call conversationdesigner. That particular role
is looking all the time at thatend-to-end customer experience
and tuning, you know, so we, wemake some changes around how
the, the customer's prompted ormake changes around the
conversation flow or thehandover, you know, cause it's
not just the AI bot we handover to other custom bots as we

(40:52):
call them , you know, that cando further investigation or
triage. And we're looking atthat whole kinda journey and
making sure that it is , feelsseamless from a customer point
of view and where it isn'tseamless that we're tuning all
the time and it's not a one anddone type effort. Like you've
gotta constantly tune itbecause the nature of your work
will , will change over time.
So that conversation designerrole for me is actually quite

(41:13):
critical. I think they can havea huge impact on the customer
experience through constantlylooking and seeing how we can
change the , the way thecustomer engages with the
technology. But the AI botitself, it's a much more
conversational field than thetraditional chat bots. And I
say it tunes itself, like itwill ask clarifying questions,
it will get to the root cause,which feels very natural for

(41:36):
people.

Speaker 1 (41:37):
Yep . I love that and I love that, again, coming
back to that customer journeyand really thinking out how are
you giving the best experience,whether it's thought , whether
it's agent, whether it's aconversation, whether it's
something that's been preemptedfrom a conversation perspective
or someone live answering, it'sall coming back to what is it
your customer is expecting andwanting, but also what is it

(41:58):
the best way to set yourcustomer up for success when it
comes to their full longevityof using your product. So
Declan, thank you so much . Wecould keep talking about this.
I could keep going and going,but we do have to wrap up and I
wanna challenge you with ourquickfire discussion questions
where I'm gonna try to ask youto answer these next few
questions in one sentence orless . I know it's gonna be

(42:18):
tough, but are you ready?

Speaker 2 (42:19):
I'll try

Speaker 1 (42:20):
. Okay , awesome. The first question I
have is, what do you think isnext for the customer support
or customer service industry?

Speaker 2 (42:29):
Wow , big question.
I mean, o obviously thepervasive implementation of AI
I think is, is immediate, but Ithink longer term it's seems
like , um, virtual reality hasa role to play somewhere down
the line, what exactly it isnot quite sure it could be on
customer onboarding, education,et cetera , but I think virtual
reality has a role to play here.

Speaker 1 (42:47):
I'm excited about that too. I think with all the
VR happening, it's gonna beexciting to see what's gonna be
the next step of interactionwith customers. The next
question I have is, what isyour favorite SaaS product that
you cannot live without as acustomer support professional?
I'm gonna say not Intercombecause I know you're probably
a little bit biased, but otherthan Intercom,

Speaker 2 (43:07):
That was, that was gonna be my answer was
intercom. , yeah . Causequite, quite literally, you
know , it's , it's been , uh, aproduct I've wanted all my
career and uh , you know , uh,I I just love it . So yeah, I'm
slightly biased. I would sayIntercom if it's not Intercom.
So we use a quality assurancetool from a company called
Klaus , and I love their kindof whole strategy and approach

(43:30):
at the moment. Again, they'relooking at how they kind of can
automate and scale qualityassurance. So I have to say
Klaus is something that I'm

Speaker 1 (43:38):
Awesome. Sorry, I know I threw in a spanner
there, but great answer . Um, and then what is your
favorite place or learningresource when it comes to
customer support?

Speaker 2 (43:50):
Okay, so I taught long and hard about this. There
isn't one I I , I like to kindof chip into as many different
avenues as possible, whetherthat's, you know, you get a lot
of good stuff on LinkedIn, youget a lot of good stuff with
various podcasts, webinars, etcetera . Uh , you know, there
are some industry people that Ikinda would look at and see
what their messaging is, whatthey're seeing in terms of the
future. So there's wholecombination of things and I

(44:11):
don't think there's one that Iwould kinda just say I , I go
to that above anything else.
Like there are some of theindustry commentators or uh ,
analysts are really producingsome good stuff at the moment ,
like Gartner , bcg , et cetera. So it's really Forrester as
well . Like it's really tryingto tap into as many sources as
possible.

Speaker 1 (44:28):
Yep . And I completely agree, there are so
many sources out there, whichis plentiful and great for
anyone who is looking forinformation. And my last
question to you today is who isinspiring you correctly or whom
do you think we should have asour next podcast gap ?

Speaker 2 (44:43):
.
Interesting. So there is a kindof a kinda support industry
expert guy called J Bar . Idunno where you've come across
Jay before, but, but, but Jayis quite an inspirational
speaker , uh, very large hislife character. Uh, like he has
some really, really good kindagrounded stories around what
good support is. He focuses alot on speed of response, like

(45:04):
he's got some really goodperspectives on, on the whole
support industry. So , uh, I'dsay Jay Bar would be well
worked , uh, time on yourpodcast.

Speaker 1 (45:12):
Awesome. I haven't met Jay before, but I will
reach out. Thank you so muchDeclan, for your time, your
expertise, your insights oneverything on how support is
transforming in today's world.
So thank you so much for yourtime. Really appreciate it.

Speaker 2 (45:25):
Thanks Nick . You really enjoyed the
conversation. Thank you.

Speaker 1 (45:27):
Thank you for listening to the Customer
Success Channel podcast today.
We hope you learn something newto take back to your team and
your company. If you foundvalue in our podcast, please
make sure to give us a positivereview and make sure you
subscribe to our channel as werelease new podcasts every
month. Also, if you have anytopics that you would like me
to discuss in the future or youwould like to be a guest on the

(45:50):
podcast, please feel free toreach out. All my contact
details are in the show notes.
Thanks again for listening andtune in next time for more on
customer success.