AI-Enhanced Member Services: What You Need to Know | 73 - Sidecar Sync

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Amith (00:00):
It's interesting because associations tend to have this
pattern of infrequent but deepengagement.
Let me put it another wayPeople come to your event, but
then they don't think about youthe rest of the year.
So think not just about whatyou can do with AI that you
currently do but things that youcould not do were it not for AI

(00:21):
to begin with.
Welcome to Sidecar Sync, yourweekly dose of innovation.
If you're looking for thelatest news, insights and
developments in the associationworld, especially those driven
by artificial intelligence,you're in the right place.
We cut through the noise tobring you the most relevant
updates, with a keen focus onhow AI and other emerging

(00:42):
technologies are shaping thefuture.
No fluff, just facts andinformed discussions.
I'm Amit Nagarajan, chairman ofBlue Cypress, and I'm your host
.
Greetings everybody and welcometo the Sidecar Sync, your home
for association content focusedon the emerging, exciting and
crazy world of artificialintelligence.

(01:04):
My name is Amit Nagarajan.

Mallory (01:07):
And my name is Mallory Mejias.

Amith (01:09):
And we're your hosts, and today we are going to be
kicking off part one of aspecial two-part episode all
about AI-enhanced memberservices.
This is content we think you'llfind useful now and probably
well into the future, becausethe concepts really are about
how you can bring memberservices into the present with

(01:29):
AI and build for the future, andwe're breaking it down into two
different parts.
Before we get into part one ofthis series, let's take a moment
to hear a quick word from oursponsor.

Mallory (01:42):
If you're listening to this podcast right now, you're
already thinking differentlyabout AI than many of your peers
, don't you wish there was a wayto showcase your commitment to
innovation and learning?
The Association AI Professional, or AAIP, certification is
exactly that.
The AAIP certification isawarded to those who have
achieved outstanding theoreticaland practical AI knowledge.

(02:05):
As it pertains to associations,earning your AAIP certification
proves that you're at theforefront of AI in your
organization and in the greaterassociation space, giving you a
competitive edge in anincreasingly AI-driven job
market.
Join the growing group ofprofessionals who've earned
their AAIP certification andsecure your professional future

(02:27):
by heading to learnsidecarai.
Amit, how are you doing today?

Amith (02:34):
I'm having a great day so far.
It's beautiful here in NewOrleans today.
I'm back Just got backyesterday from a week-long ski
trip, and that always puts me ina good mood.
Of course, I've got the usualcatch-up game to play with being
out for a little bit, but I'mdoing great.
How about yourself?

Mallory (02:50):
I am doing pretty well myself.
Unfortunately not beautifulweather here in Atlanta, but
hopefully the rest of the weekwe'll get some sunlight out and
less of this rainy cold.
But I've been pretty excited todo this AI Enenhanced member
services episode because I feellike it's been a minute since
we've done a more evergreen typeepisode.

Amith (03:09):
Indeed it has, and this topic, I believe, is going to be
of value for people now and foryears into the future.
Certainly, the AI techniques,the AI tools that you'll use
will vary and they'll certainlyimprove over time, but the idea
of improving the fundamentalcustomer service function of an
association is so key.
Many association folks in myexperience get excited thinking

(03:31):
about how AI can save them timeand improve their internal work,
and that's, of course,critically important.
But if you can alsofundamentally improve the way
you serve your members, the wayyou engage with that audience,
that's incredibly interesting,right, because it potentially
levels up the quality of theexperience, that can help
improve member retention, thatcan attract new members and

(03:54):
ultimately just creates morevalue all around.
So that's what these twoepisodes are, and to me it's a
very exciting, evergreen topic.

Mallory (04:03):
Yep Excited to kick that off.
I know you mentioned you justgot back from some travel, Amit,
but I think you have some moretravel coming up right, Because
we've got those DC and Chicagoinnovation hubs coming up.

Amith (04:14):
I have far too much travel going on this spring.
So, yes, I'm bouncing around alot.
I'll be in DC, I'll be in Vegas, I'll be in a couple of other
places over the course of thenext, I think, four or five
weeks.
So it's going to be busy.
So it's fun.
I enjoy getting out there andmeeting with people too, but
it's a lot right now.

Mallory (04:33):
And remind me the DC one is.
It's this month, it's March25th, 23rd, that's correct.

Amith (04:38):
Yeah, so Blue Cypress and Sidecar have an innovation hub
in Washington DC on March 25th.
It's a full day event and thatevent is going to be fantastic.
We're going to cover topicsreally all around innovation and
at the moment, a lot of thatinnovation talk, of course, is
AI related, and after we wrap upin DC, two weeks after exactly,

(05:01):
we will be in Chicago on April8th doing the same thing.
So we have DC Innovation Hub onMarch 25th and the Chicago
Innovation Hub on April 8th, andyou're not going to want to
miss these events.
They're in person, but just theday, and it's an informal
gathering of people who arereally interested in driving
their associations forward withall kinds of innovation.
Again, ai is the center of alltopics.

(05:23):
It seems to be across the wholemarket right now, but it's
going to be really exciting tojust talk about a variety of
different practitioners sharingtheir practices and what they've
been able to innovate.

Mallory (05:33):
Yep.
So if you all are looking forsome fantastic content to
consume in March and April, Ihighly encourage that you check
those out, and maybe you'll evenget to meet Amith while you're
there, so it's a win-winsituation, all right.
As Amith mentioned, we aretalking about AI-enhanced member
services today.
In this part one episode, we'regoing to kind of set the stage

(05:53):
for why we believe this iscritical right now.
We're going to break down theresponsibilities within the
member services function anddiscuss ways that AI can augment
each of those member servicesfunction and discuss ways that
AI can augment each of those.
In our part two episode, we'llfocus more on implementation, on
actually getting started, howto turn those ideas into action.

(06:13):
We'll also be talking about thechange management piece and
kind of having thatuncomfortable conversation
around AI and job displacementthat we tend to avoid.
So, first and foremost, we knowwe're exploring what we believe
is a critical transformationopportunity for associations,
which is modernizing your memberservices with artificial
intelligence.
To prepare for this episode, weused Google's deep research to

(06:37):
analyze hundreds of memberservices' job descriptions
across associations.
So we're going to break downprimary responsibilities and
duties from those jobdescriptions, task by task, and
discuss ways that AI can augmentor transform them entirely.
I think we're all in agreementthat member services has
transformed from handling basicadmin tasks into a strategic

(06:59):
function that directly impactsthe way your members engage with
you.
And retention that directlyimpacts the way your members
engage with you.
And retention this is yourorganization's frontline, your
direct connection to members andprimary brand touchpoint.
The challenge associations facetoday is this growing
experience gap.
So members aren't comparingyour service to another
association's.

(07:19):
They're comparing it to bigbrands they're interacting with
daily that offer things like24-7 availability without
waiting, immediate, accurateresponses, proactive rather than
reactive service,hyper-personalized experiences
and seamless self-serviceoptions.
The gap, yes, is frustratingfor your members and it directly

(07:40):
impacts retention, satisfactionand perceived membership value.
We believe AI is offering aunique opportunity to not only
close this gap but potentiallyleapfrog traditional consumer
experiences while reducing costs.
So, amit, you have been talkingabout friction since I met you,
so I know it's at least threeyears.
I think we can agree.

(08:01):
It's always been relevant, butwhy do you feel kind of this
sense of urgency now?

Amith (08:08):
One of the ways to think about business is that often
people come for the product butthey stay for the service.
So they wouldn't necessarilycome to you if they didn't need
your product.
That doesn't make any sense,right?
But oftentimes retention thewhy people stick around is
because they value the service.
That's true in the softwarebusiness.
In my experience across a lotof software companies, you tend

(08:32):
to have really high customerretention when you provide a
great, empathetic, high quality,responsive and really value
additive service.
Function and the product itself, of course, remains important.
If your product is no longervaluable, that's an issue, but
you can have the best product inthe world.
But if your service is terribleor if your service is simply
non-responsive, that's a problem.

(08:54):
So to me, that is one coreprinciple that I do think
associations completely get.
But I also think that peoplehave become accustomed to their
challenges as an association,their choke points internally,
the number of member services.
Personnel is the number onething, right?
How many people can you affordto have employed by your

(09:14):
association, on a full-timebasis or on a part-time basis,
who can help your members whenthey have questions, right?
So that's the issue is thatpeople the members coming in
looking for help, have a limitednumber of people who can help
them, and so sometimes you knoworganizations, you hear them
talking about a goal of having a24-hour response time as an

(09:37):
example, and that's their goal.
Oftentimes that's not reached.
So it's typical in manyassociations to have a multi-day
wait time to get feedback formember services.
Obviously, there's exceptions,there's some organizations who
are able to be far moreresponsive than that, but that's
an issue, right, because havingto wait a day, even if that's

(09:58):
considered the best case, or twodays or three days, that's not
really acceptable in the worldof 2025.
That's a big problem.
And then sometimes, other partsof friction are even with the
so-called self-service aspectsof member services and, by the
way, member services, you couldreplace the word member with
customer or vendor or volunteeror event.

(10:19):
It's service, and so havinggreat service matters across all
dimensions of your brand,persona, all aspects of your
business.
But let's talk about thatself-service concept.
Associations often will have awebsite these days, of course,
and those websites will allowpeople to, let's say, register
for a webinar or perhaps renew amembership.

(10:41):
But oftentimes the amount offriction there is so incredibly
high it's almost comical, youknow, to register for a webinar
with some association websitesmight take seven or 12 or 15
steps instead of simply likeputting in your email address
and hitting register, and sothis idea of user experience,
customer experience isunfortunately a lower priority

(11:05):
in practicality than it shouldbe, and so my point would really
simply be put this way Peopleare not going to put up with
this anymore.
They live in a world wherethings are instant and really
good.
So you know why would I put upwith that when I have
alternatives?
In a world where you have amonopoly in your particular
domain, maybe you can get awaywith those kinds of things, but

(11:26):
that's not true in a competitiveecosystem.

Mallory (11:30):
That makes sense.
Amit, I want to take thecounter position here to you and
kind of to myself, with what Isaid when I was introing this
part of the episode.
When I'm interacting with bigbrands, right, I do expect
minimal friction.
I do expect near perfection.
As an example, I'm a big fan ofWhole Foods delivery through
Amazon.
I hate going to the grocerystore, so that's something I use

(11:51):
all the time.
Just placed an order before thisepisode.
The service is quick, and theonly downside with delivery
groceries sometimes you getproduce that's not great or
spoiled, but their refundprocess is so efficient.
I just I don't even interactwith a person, right.
I say I didn't like thisproduct, they automatically
refund me.
On the flip side of that,though, if I were calling up my

(12:13):
local DMV or OMV and asking arenewal question about my car,
I'm going to expect pain.
I'm going to expect long waittimes, multiple transfers, all
the terrible things, right, thatcome with customer service, so
I don't hold the DMV to the samestandard that I do Amazon, for
example.
Do you feel like there's anargument there for members of

(12:33):
associations?

Amith (12:35):
Well, I mean, it kind of goes back to the question of
what are your choices and youknow if you live in the state of
Georgia.
You have one DMV you get to goto.
You know if you live in thestate of Georgia.

Mallory (12:44):
You have one DMV you get to go to.

Amith (12:44):
You really can't, unless you just want to not have a
license or not register your car, which some people may choose
to do that.
But the reality is is that youdon't have choice.
They have a monopoly on thatservice category.
Perhaps another category thatis worth discussing, kind of in
the same vein though, of whatyou're describing, is airlines.
Very few people would simplysay that airline customer

(13:06):
service is fantastic.
It's pretty rare.
I mean, southwest used to be,kind of, you know, the solo
player in the market that peopledid say positive things about
customer service, andunfortunately that seems to have
declined.
I'm not a big Southwest flyer,but I just hear people generally
say they've gone downhill inthat area over the last 10, 15
years.
But my point there would be thatthere is a semi-competitive

(13:27):
space.
Right, because there are not alarge number of airlines, but
there's enough of them andthere's enough choice,
particularly for you living inAtlanta, where it's a major hub
and there's a lot of choice.
A little bit less so here inNew Orleans, but you kind of do
put up with bad service theretoo, because there is a little
bit of friction in switchingcarriers, particularly, let's
say, if you have frequent flyerpoints with Delta or with United

(13:50):
or one of these other ones, youtend to put up with a lot of
pain because there's kind ofwhat we'd refer to as a
switching cost, as both a powerfrom a strategic lens and also
as a pain point from a customerviewpoint.
So I think that there's kind oflike this inverse correlation,
if you will, between thewillingness to accept friction
and people's long-term desire tostay with the brand innately.

(14:13):
Ultimately, people do want tohave a low friction, high
quality experience if they canget it without having to undergo
enormous switching costs.
So for you, the switching costsother than being an unlicensed,
unregistered driver which I'msure you're not, mallory,
however, would be to move states, right, you'd have to leave the
state of Georgia.
That's a very high switchingcost.

(14:36):
Now, in the context ofassociations, of course, the
switching costs are.
Historically they have beenhigh.
If you're an association in aparticular domain or profession.
Historically there haven't beenthat many choices in your exact
category and perhaps yourcategory plus your geography
combined.
So a state association innursing or a international

(15:00):
association in another countryin architecture, right, there
tends to be like one of thosethings in each of those kind of
verticals plus geographies, butthese days geographic borders
have kind of melted away.
Content is ubiquitous andvaluable across borders, and
it's not just associations thatare in the content game.
Tons of other players are outthere who may not have content

(15:23):
that is 100% as good as yourpeer-reviewed fantastic journal,
right, but it might be goodenough.
And so this friction problem isa real problem because there
are alternatives.
I think associations firststarted to feel this pain with
Google when search enginesbecame free and ubiquitous.
You know, 20 years ago it was aproblem for associations

(15:46):
initially because people couldfind stuff, but it wasn't that
big of a problem because thestuff that people were finding
wasn't really all that great.
But with AI, the answers peopleare getting are truly
outstanding.
And we were just talking aboutCloud 3.7 right before this call
right, and you could probablygo ask Cloud 3.7 quite a number

(16:06):
of very complex, highlyspecialized questions in a lot
of domains and get a good answer.
And that's true, of course, ofdeep seek, it's true of open
AI's latest stuff.
It's true for a lot of theseAIs that have been so broadly
trained on so much outstandingcontent and, by the way,
synthetic data, which is a wholeother conversation that
actually these general purposeAI tools are competitive and

(16:31):
they're displacing forassociations core services.
So that's an issue.
And if you can get an answerfrom a generic AI model on your
phone or on a website in secondsand it's either free or
extremely cheap and to get thesame answer from your
association requires severalforms of backflips, you're not
going to go to that associationfor very long.

Mallory (16:51):
Well, with that, I want to move into this task by task
breakdown.
As I mentioned at the top ofthe episode, we are using a
Google Deep Research report andwe're going to mention, for each
of these tasks, the currentstate, what we believe is the AI
opportunity, the potentialimpact and then a more practical
example for you to wrap yourhead around For each task as

(17:14):
well.
Amit, I'm going to ask you someclarifying questions and I also
want to get your take on thedifficulty to implement it.
So low, medium high.
And then also the perceivedimpact low, medium high.
First and foremost, man we canprobably all relate to this in
some capacity.
But responding to inquiries and,as Amit said, we're focusing on
member inquiries, but thiscould be vendors, various other

(17:37):
groups, volunteers as well thecurrent state you have staff
manually handling emails andcalls during business hours with
varying response times.
Potential AI opportunity hereis something like a 24-7
intelligent chatbot andautomated email responses that
can understand complex questions.
The impact of this, I think wecan all imagine, would be huge.

(17:59):
So significantly reducedresponse times, consistent
quality and the ability for AIto detect member sentiment,
which could be important forvarious kind of downstream
things.
And then, finally, a practicalexample maybe.
This is an AI-powered memberservice agent that can handle
inquiries across all channels,including questions about

(18:20):
certification requirements,upcoming events, membership
benefits, at any time of day.
So that is our first taskresponding to member inquiries.
In theory, amit, this soundsgreat, sounds easy, sounds
perfect.
But in actuality, I certainlyunderstand that email inquiries
can be more difficult to resolvethan you think.
From the Sidecar perspective,we do have people reaching out

(18:43):
to Sidecar on occasion, askingkind of straightforward
questions like tell me about theAIP certification or tell me
about the AI learning hub formembers offering you have.
But oftentimes, when customersreach out to us, it can be a
little complicated, like can youissue me this certificate for
this virtual event that happenedtwo years ago?
Right, it takes multiple steps.

(19:03):
It's not a quick one and, doneso, do you see this task as
something that can really befeasible for associations in
terms of AI enabling it?

Amith (19:15):
Well, I think so, because what you're trying to do is
break down the parts that AI cando well and the parts that AI
maybe not, is maybe not bestsuited for, at least at the
moment.
And so, if you kind of break itdown, I like this idea of you
know, build a pie chart in yourmind of the distribution of the
tasks that are complex versusnot complex.

(19:35):
Or perhaps another way to breakit down is AI suitable versus
not suitable for AI?
And what I think you'd find isthat certainly FAQ kind of
inquiries will be not onlyhandled well by AI probably
handled better than by a human,because the AI doesn't get tired
of responding to the samequestion over and over again,
whereas humans do tend to gettired of that kind of thing, and

(19:56):
maybe they're copying andpasting the response, but it
tends to get difficult.
So that kind of thing iscertainly easy for AI to handle.
I think that quite a few of thetransactional things like can
you swap out my guestregistration from person A to
person B, right Very commonkinds of things that you get,
particularly leading up to largeevents.
It tends to be an overwhelmingvolume of both FAQ style

(20:19):
inquiries but also minor kind ofincremental transactions that
people are looking to do as youget close to a major event.
Those are things AI can handlebeautifully already, and the
thing to remember we say this alot, mallory, on this pod is
that the AI you're using todayis the worst.
AI, you're ever going to getright.
It is the worst, even thoughit's amazing.
It's cloud.

(20:39):
3.7 is both amazing and it'sthe worst you're ever going to
get.
So that's the thing to thinkabout.
You don't want to build yourplans based around today's AI,
because you know we all wantassociations to move faster.
But in reality, if you weregoing to implement an automated
member services function likethis, I don't think you'd
probably be doing it tomorrow oreven next week, or possibly not

(21:00):
even next month.
It might take you six months toplan and implement something
like this.
It doesn't need to, by the way,from a tech perspective, but
from just an organizationalperspective, a cultural
perspective.
Budgets, all that.
It takes time to do stuff, andevery six months, you get a
doubling in AI right in terms ofAI capacity, ai power, and that
is equivalent to also halvingthe underlying costs of the

(21:22):
models.
So AI is going to be able to doa very large percentage of
these things.
I'll give you a couple ofexamples, too, that AI can go
beyond just doing what we do.
So we tend to think of like theuniverse of possibility is that
which we as humans can do.
Okay, so that's cool.
We definitely want to solve forthat, because that's the

(21:42):
expectation is, people want youto be able to do at least what
you do with people in order torespond to these inquiries.
But what about what peoplecan't do?
Your member servicesrepresentatives, as wonderful as
they may be, are not experts inyour domain typically.
Maybe some of them have someexperience as surgeons, as
accountants, as lawyers buttypically not.

(22:03):
Typically the people doingmember services are focused on
member services.
They're not people with domainexpertise.
But AI can be a domain expertand not only deal with kind of
rote FAQ style inquiries butalso domain questions.
So people can start emailingyou, texting you, calling you
about things that are actuallyhelpful to them in the lane of

(22:26):
their professional work.
Now why is that interesting?
It's interesting becauseassociations tend to have this
pattern of infrequent but deepengagement.
Let me put it another wayPeople come to your event but
then they don't think about youthe rest of the year.
That's a problem in terms of abusiness model, because it's
easy to get forgotten and easyfor people to choose to not use

(22:47):
you if they don't use youfrequently.
So if you can provide in-domainprofessional knowledge and
response to inquiries about thatprofessional domain of
knowledge, that's leveling upright, because you don't offer
that currently.
So think not just about whatyou can do with AI that you
currently do, but things thatyou could not do were it not for

(23:09):
AI to begin with.
So to me, that's a bigopportunity.
Around member services andmember inquiries.
It is, in fact, a process verysimilar to what you described,
where you have people respondingto emails, responding to text
messages in some cases, andcertainly some phone calls, and
many of those things can behandled by AI.
One key thing, though, I dowant to point out I said it
earlier and I want to drill intoit just for a second you do

(23:31):
need to account for, in thesesystem designs, things that AI
can't handle.
So if you try to get the AI toanswer something that it's not
suited for, it's probably goingto piss somebody off.
So a good example of that issomeone who is not happy with
you.
So let's say I write an emailMallory to Sidecar and I go hey,
mallory, your customer servicereally stinks.
And let's say it's the AIversion of Mallory that's

(23:53):
responding, and that's the thingthat I'm complaining about.
Do I really want an AI responseto my complaint about the AI?
Right, maybe not?
Right, maybe not?
And so I probably want to speakto a representative.
Maybe I want to have thatforwarded to someone else.
So the first thing you want todo and this is something AI can
indeed do is you want an AI todecide whether an AI should

(24:14):
handle it or if it should behandled by a human.
So, in that case and youactually probably want to be
fairly thoughtful about that, sothat you narrow the use cases
where the AI attempts to respondand leaves the rest of the
messages for people to actuallyrespond to- I like that
distinction between AI suitableand not AI suitable.

Mallory (24:35):
Amit, I know we have a good bit of influx of new
listeners and viewers on theSidecar Sync podcast.
I realize we mentioned BlueCypress without maybe clarifying
what that was, but for our newlisteners and audience, Blue
Cypress is the parent company ofSidecar and Blue Cypress has
been working on a product thatdoes exactly this right Responds
to, can respond to memberinquiries using AI.

(24:58):
Can you talk a little bit aboutthat?

Amith (25:00):
Sure Happy to so.
Within Blue Cypress, we're anincubator for new technology
ideas, and we provide a varietyof different professional
services for associations aswell, and one of our newest
products that we're excited tobe launching right now, as a
matter of fact, is an AI agentcalled Izzy I-Z-Z-Y and Izzy is
named after Isidore Sharp, whowas the founder of the Four

(25:22):
Seasons, and we were looking forpeople to honor with this
particular AI tool, and very feworganizations have a service
standard quite at the level of aFour Seasons.
We also thought aboutorganizations like Nordstrom's
and others that haveextraordinary customer service,
and we asked ourselves well,what if you could have AI
deliver a Four Seasons orNordstrom's or plug in a

(25:44):
different brand that you thinkis aspirational from a quality
of service perspective, everysingle time to your members
across all associations?
It would be remarkable.
So that's what Izzy is, that'show Izzy got named, and Izzy is
an AI agent that can plug intoany number of your channels,
includes email, includes SMS Ifyou're international and you

(26:05):
have people on WhatsApp or onFacebook Messenger, you can plug
into that as well as well assecure communication modalities,
izzy can process those messages.
Izzy is an asynchronous agentright now, meaning that Izzy is
not suitable for real-time voicecommunication today.
That will be something Izzywill be able to do, probably by
the end of the year, but rightnow AI is not quite fast enough

(26:27):
to be excellent at that.
It could just be like okay, sowe're not tackling that yet.
We're providing onlyasynchronous so email, sms, et
cetera.
And Izzy also is very, veryclose friends with another one
of our AI agents named Betty.
Betty is the domain expert,right, so Betty's a knowledge
agent.
Betty surfaces herself on theassociation's website or in apps

(26:47):
.
What Izzy does is Izzy's ableto talk to Betty whenever an
inquiry comes in.
That's about domain knowledge.
And then Izzy also is giventools, just like you would give
a human agent tools to work withyour AMS or your LMS.
You train Izzy to work with thevarious software tools you
already have in place, like yourcurrent AMS or LMS, and Izzy is

(27:08):
able to use those tools.
Again, you have to providepermission and the specific
capabilities to do things likeprocess member renewals,
register people for events,cancel registrations.
So you essentially give Izzy avocabulary, that is, a set of
tools that Izzy can use toprovide service to your incoming
inquiries, and Izzy takes careof it.
All this is done with logging,so you have complete tracking of

(27:30):
everything, so it's a reallycool product.
We're excited about Izzy.
The goal of Izzy and I knowwe're going to touch on this
later in this conversation thegoal is not to displace member
services personnel, but ratherto increase the level of quality
, both through responsivenessand better answers that your
members get, and to enable yourmember services team to do the

(27:52):
high-touch stuff that they'rebest suited for, which right now
you can't do.
Simple example of that isproactive outreach.
What about actually callingyour members or emailing them to
talk about their needs beforethey actually ask you for
something because they're upsetabout something?

Mallory (28:09):
Aziz sounds fun.
I like the names.
Amit and I are always big fansof infusing some fun when we can
, so I'm glad it's not calledservice bot or service agent.
In your opinion, Amit,responding to member inquiries,
having an AI augmented processfor that, what would you say is
difficulty to implement?
Low, medium, high, and then thepotential impact of that.

Amith (28:27):
I would rate the difficulty as medium, because
you do have to integrate whetherit's Izzy or something else.
You could build it yourself.
There's other products that dothis.
You have to integrate thisservice capability with systems
you have in place and withknowledge, so it's not low.
I don't think it's high either,though, because the level of
integration is not super, superdeep Impact.

(28:48):
If there was a category abovehigh, I'd probably pick that.
It's enormous.

Mallory (28:54):
All right, moving to the next task processing
applications and renewals.
So the current state is manualverification, payment processing
delays and limited follow-upcapacity.
The AI opportunity here isautomated application review,
intelligent document processingand anomaly detection, the
potential impact, streamlinedprocessing workflows and

(29:16):
improved conversion experiencesfor your members.
Streamlined processingworkflows and improved
conversion experiences for yourmembers, and an example here
would be a system that canautomatically verify credentials
, process payments and sendpersonalized renewal reminders
based on member behavior.
Amith, my question here.
We probably have a lot ofpeople in our audience who are
saying we have some prettycomplex membership categories

(29:37):
and eligibility requirements.
Can AI really understand thatkind of nuance when processing
applications?
Yes, full stop, no explanation,just yes.

Amith (29:50):
Yeah, ai is capable of understanding those kinds of
nuances as well as the averagehuman can.
At this point, quite franklyand probably that's an
understatement what I would tellyou is this that I think that
the nuance is actually where AItends to shine really well.
So things that are like this isa bit of a fallacy in all of
our thinking, including my own,that when you think about

(30:10):
computerizing or automatingsomething you tend to think of
like the rote, repetitive stuffthat's the same every time,
because that's historicallywhere computers have played well
.
Right, if you can give it a setof very rigid rules, then the
AI sorry, not the AI, but thecomputer pre-AI would just do
those things over and over again, right?
So it's very easy to processthe same transaction over and

(30:31):
over again if all the parametersare exactly the same.
But the minute somethingchanges or requires some
judgment call, then traditionalclassical computing tends to
just stop.
And that's when you need ahuman to do a review of
eligibility requirements.
Or perhaps let's say it's anapplication not for membership,
but someone's submitting aproposal to speak at an event.
That requires a high degree ofsubjectivity.

(30:53):
You have to review text andread what they've submitted and
determine whether or not youalready have a speaker on that
topic and whether that topic isaligned with what you want for
that event and whether thatperson has credentials that you
think are suitable to be able toaddress your audience on that
topic or in general, and on andon and on Right, and actually
that's where I really that's,that's the missing ingredient

(31:15):
from traditional automation.
Traditional automation requiresa highly rigid set of constructs
or rules.
Ai can operate in the gaps.
Ai can operate where the nuancelives.
So this is also an excitingpiece, both because I don't know
really anybody that I've evermet that enjoys processing

(31:35):
applications.
So if that's part of what youdo or part of what your team
does at your association, helpis eventually going to be on the
way and it's available reallysoon if you want it, and so you
can take some of those steps outfrom a manual process.
Now, does this mean that youdon't review anything ever again
?
No, it is like a trust butverify kind of mindset.

(31:56):
So when the AI does the review,does that mean, for example,
that you have the AI completelyprogram your upcoming annual
conference and choose all of thespeakers, align all the
sessions and nobody doesanything by hand, of course, not
Just like if you hired a reallysmart recent college graduate
to do that job for you.
You would absolutely reviewtheir work before you published

(32:16):
it on your website or before youconsidered it done.
Similarly, with AI, you canhave anywhere from like a random
sampling, where you just checksome of the work, to an actual
review process where humansactually look at the work of the
AI and approve it or reject it.
So let's talk about one specificconcept within this to maybe
put a little bit more concretestructure around know structure

(32:39):
around what we're talking about.
So let's just take that speakersubmission concept right and
that's probably more part of theevents department.
If you're a larger associationthan it is membership.
But membership often has likevolunteer applications that they
are dealing with and sometimesapplications for membership
itself.
In some associations it'sliterally anyone who wants to
join can join as long as theypay.

(33:00):
But in some organizations, ofcourse, there's a whole series
of steps you have to go through.
But let's take the speakerproposal submission process for
a moment, because it's the samebasic workflow.
So one of the first things youwant to do is determine whether
or not the person's content thatthey submitted is relevant to
the event.

(33:20):
So if I have an event all aboutAI for associations and someone
submits a session onagriculture or Is that relevant
to my event?
Probably not.
Does that require a human tomake that determination?
Really it doesn't right.
You can take that even to somevery low level, low capability

(33:42):
AI models and say here's anabstract from this individual
who submitted it and here is anevent definition of what we're
looking for.
It's essentially like a rubricthat says these are the topics
and these are the criteria forevaluating a submission.
And I was extremely good atgiving you that high level rough
cut of is it a good fit or abad fit?

(34:02):
Right, and Mallory, you couldgo do that right now without
programming.
You could go do that one by oneIf I gave you a whole bunch of
documents that were submissionsand if I gave you a rubric, you
could go and personally do thatinside Plot or ChatGPT, right,
and those tools would give you apretty good answer saying yay
or nay?

(34:24):
And then probably actually giveyou, if you asked for it.
In the prompt it would alsogive you an explanation of why
it's a good or a bad fit.
So that first round of review.
My example is kind of silly,right.
So typically people aren't thatfar off.
There's a little more nuance toit, but you know that first
phase often can cut out 20 to 50percent of the submissions.
The next thing that people oftenwant to do is they want to
compare that speaker's topic toother topics to determine if

(34:48):
they have too much of aparticular thing, right, right.
So how many people are going tobe speaking, let's say, on AI
for, let's say, audio generation, at the upcoming Digital Now,
which, by the way, is November2nd through 5th in Chicago?
So Digital Now, we want to havea fairly diverse array of

(35:12):
content around AI forassociations as the overall
theme, but we don't want to havethe same content over and over,
right?
So if we have six speakers thatall submit a proposal to speak
on using 11 labs for podcastgeneration or whatever, probably
wouldn't be great.
So that AI now has to likecompare that new proposal
against proposals that havealready been accepted, or maybe
just group them together so youcan more easily analyze this,

(35:33):
and the list goes on and on Eachof these topics.
Think about what they're doing.
They are essentially takingunstructured data, which is the
submission, and they're tryingto gain some structure out of it
.
They're trying to basicallypull some structure and, in some
cases, make some, effectively,judgments around that

(35:54):
unstructured content and, aswe've talked about on this pod
and many other venues, that'ssomething that AI is incredibly
strong at.
Already.
Here you're just dropping itinto your workflow as an
association.
So the manual process that Isuggested Mallory could do
without doing any programming,of course, is an option and
that'll probably save you sometime.
But what will save you a lotmore time is if you build AI
into the workflow.
So if you have a website thattakes in a submission to build a

(36:16):
little AI, step directly intothat to do that first cut of
review and even give real-timefeedback to the submitter saying
, hey, this isn't quite rightfor us, which is really valuable
to the person that's taking thetime to submit something to you
to begin with.

Mallory (36:31):
With this task, Amit, so processing applications and
renewals, or even extrapolatingand maybe using it for the event
speaking proposal process.
What would you say isdifficulty to implement and then
impact?

Amith (36:55):
current process.
So some people just have awebsite that dumps data into
like a Google sheet or into anExcel document with all the
submissions.
You could take that Excel fileor that Google sheet and
directly drop it into any of themodern AI tools and ask it to
compare each of thosesubmissions relative to a rubric
or a criteria, essentially, andgive you an answer, and so you
could do that manually and thatwould be quite useful.

(37:16):
I don't know that it wouldreally save you an enormous
amount of time ultimately.
I mean you would save time andthat you wouldn't have to
manually read every one of theresponses or the submissions,
but you'd still be doing somework by hand.
But that version is low effortbecause you could do that really
today with any of theseconsumer grade tools and the
impact would be probably low tomedium.
I think it'd be useful.
Low effort because you could dothat really today with any of

(37:37):
these consumer grade tools andthe impact would be probably low
to medium.
I think it'd be useful, but itwouldn't really like cut out a
ton of time probably.
I don't know.
Do you agree with that or?

Mallory (37:44):
Yeah, I agree with that .
I mean, I think it would maybehave a high impact on your staff
who, like, don't necessarilywant to do that, but maybe not
necessarily on the members, notquite as much as handling those
inquiries and having, like, aknowledge assistant.
I think that would be prettyimpactful.

Amith (38:00):
No, I think that's true and, at the same time, the next
level up, the version of thisthat would likely have a very
high level of impact, is if youput some automation in place
right where you don't do itmanually, where you say hey, you
know, we, through the manualprocess, we've proven to

(38:21):
ourselves that in fact, I can doan extraordinary job of
evaluation of content orwhatever it is in this
application process.
Now let's weave it into ouractual business process.
Right, so on our website whenyou submit a proposal, when you
submit a membership applicationto have AI directly woven in in
an agentic way, as we've talkedabout.
Now.
You're talking about a mediumto high implementation effort
but also a much higher level ofimpact, because you both reduce

(38:41):
staff time.
But also, in this case, it'sthat dual win or the win-win,
like we talked about with theresponses to inquiries.
Here you now have peoplegetting reactions from you as a
system much faster, right,possibly real time or within
minutes, whereas it might havetaken them days, weeks or longer
or maybe never, to get aresponse to their proposal

(39:02):
previously.
And actually just think aboutthat for a moment.
On the association side of theequation, you have this
seemingly endless stream ofstuff you have to review, either
in-house or, a lot of timesactually through a volunteer
committee which you don'tdirectly control.
These are again volunteers, andso that's frustrating on a lot
of levels.
But think about it from theother perspective for a minute.

(39:22):
So it's the famous Amazon emptychair in the conference room
which represents the customer,and so in this case, if you
think about the personsubmitting the content as maybe
not the customer, but they're astakeholder in your process
they're not getting a greatexperience, either because
they're waiting a really longtime.
A lot of times, the feedbackthey get is either yeah, you're

(39:43):
in, great Congratulations, orsorry, you're not in.
And they might have to waitthree months to get a response
like that, which really stinks,because they might not know if
they're going to come to yourevent until very close to the
event.
They might not ever hear backif they're getting a no, that
happens in some cases, but it'sreally really bad.
And well, you want the bestcontent, don't you?

(40:05):
You want your association to bethe premier place for people to
want to come speak or to submitcontent to your publications,
and so making it as low frictionas possible is really valuable,
because you have people thatare more likely to come to you
with their ideas if that's thecase.
So I think it's a reallyvalue-added process from that
viewpoint.

(40:25):
The other thing that you coulddo is, let's just say, someone
submits a proposal.
So Mallory submits a proposalto speak at Digital Now, and
we're like, oh, mallory has agreat background, she's, she'd
be a great speaker for us.
She even has, let's say, priorspeaking with us.
That was well received by theaudience, or whatever, but she
submitted a topic that we'vealready accepted.
It's 95 percent similar tosomething else we already have

(40:47):
programmed.
So what would normally happen?
Well, first of all, it wouldtake a long time for Mallory to
hear back, and then then she'deventually get a no right, not a
hey.
Mallory, thanks so much for yourinterest in speaking at Digital
Now.
We'd love to have you involved,but that topic is not quite
right for us.
What about one of these threetopics right To give you
feedback and suggest analternative, and then let you
resubmit with the alternativetopic?

(41:09):
These are all things that AIcan do beautifully and very,
very rapidly.
So there's a lot of opportunityhere to level up in terms of
the quality of your content andthe way you engage these people
who are really trying to submitvaluable insights into your
organization.

Mallory (41:28):
The next task we want to dive into is database
maintenance.
So the current state of that ismanual updates, the duplicates,
inconsistent data.
That is, manual updates, theduplicates, inconsistent data.
The AI opportunity is automatedrecord deduplication, data
enrichment and qualitymonitoring.
The impact higher data accuracyDon't we all want that?
With minimal staff timeinvestment, and an example here

(41:49):
would be having an AI thatcontinuously scans and cleans
member records, identifiesoutdated information and even
suggests updates.
So, amit, something we hearover and over again with
associations is data quality isa barrier to AI implementation.
This one, for me, is a littlebit chicken in the egg, because
I know your answer is AI canabsolutely do this right now and

(42:11):
it's not that difficult toimplement.
That's what I think you'regoing to say.
But if you have to kind of getyour data in a good spot to
implement AI, but then AI canclean up your data, it's kind of
like which comes first.
So I'm hoping you can talk alittle bit about that.

Amith (42:26):
Yeah, well, I think there's elements of this that
are definitely low hanging fruit, right, and there's pieces of
it that AI can help you withfairly quickly but, at the same
time, to fully and thoroughlyimplement.
This is definitely not an easylift, because people have
typically dated infrastructure,to be kind in the way you
describe it.
You know it's kind of funny.
I remember my prior life.

(42:47):
I'm no longer involved in this,but for 20 plus years I started
and then ran an AMS company,and this software company
provided very large associationswith their core database system
, and I remember one of thethings we used to do when we
demoed the software is we'd goand ask people in the room hey,
do any of you have problems withduplicate data?
As kind of leading into areally cool feature we had that

(43:09):
would help you merge duplicatedata.
And, of course, everyone prettymuch said, oh man, yeah, no,
that's such a pain point.
So, and except one time, Iremember there was this guy who
was adamant that hisorganization had no duplicate
records, zero and I said do youhave no data?
Because that's the only way Ican imagine having no duplicates

(43:30):
.
What he was referring to was atechnical aspect, that the
primary key of the table was infact, distinct and unique and
guaranteed by the database to beunique.
Of course that doesn't meananything, because you can have
multiple records that do havedifferent ID values, but of
course they're the same personand that's fundamentally the
issue.
So, in any event, I thoughtthat was quite interesting.
That happened once over 20something years, but it was a

(43:53):
problem back then.
It's been a problem since thebeginning of time.
It's even a problem with paperfiling preceding digital right,
where you have the same file andsomeone fills out you know, I
don't know, maybe you've gone toa doctor's office and they've
asked you to fill out the samedamn patient intake form again,
even though you've been going tothat practice like for 10 years
.
We've all experienced thatright.
So that a version of of theanalog version of duplicate data

(44:16):
.
So what is the capability we'retalking about here?
The first thing is is that wewant to be able to detect,
detect duplicates, right.
So deduplicating or merging thedata, that's a complicated
thing because you probably havemultiple systems and even within
a given system, a lot ofsystems do not support the idea
of eliminating duplicates ormerging records, at least they

(44:38):
don't do it very well.
So we're not going to reallytalk about how to solve for that
, because your underlyingdatabase systems like AMS, lms,
et cetera.
It's really, really tough, andso modernizing that technology,
of course, is a good idea, butsometimes that's a tall order,
at least in near term.
So what we're going to talkabout right now for a minute is
how do you detect duplicates,because, let's say, for example,

(45:00):
from the customer perspective,the annoying thing is getting
three emails from you that allsay the same thing, because you
have them in your database threetimes.
Well, maybe you have them inyour database three times, but
you want to figure that out andsend them one email.
Right, and that might soundsimple, because you say, oh well
, I have a distinct list ofemails, but what if the emails

(45:21):
are slightly different?
Right, a lot of people haveemail aliases.
You might be Mallory atSidecarai, you might be Mallory
M at Sidecarai.
You might have several emails.
So it's a lot more complex thanit may appear.
So AI is capable of doing this.
So there's some different kindof layers that I described First
.
Number one thing here I'm aboutto say is make sure you're
working with a system that youtrust, so don't go and be a free

(45:45):
user of ChatGPT.
Don't send your data todeepseekai, because that's not
housed here.
But if you want to just take aspreadsheet of emails that
you're about to send out likefirst names, last names, emails,
organizations and then throwthem into a cloud again a paid
account with cloud or one of theother tools that you consider
trustworthy you'll be able toget a list back that's

(46:06):
deduplicated with a very highlevel of accuracy from a
consumer tool.
Now, the more columns you putinto that file, the more likely
the AI will be able to actuallyhelp you dedupe it.
Now, that's just done with areally simple language model.
Right, it's not nearly as goodas what you might have if you
use a more sophisticatedapproach, but there are tools
that you can use as just aregular business user not a

(46:27):
developer, not an IT person tohelp you a little bit at least
to minimize the impact on yourcustomer or member.
So that's a good thing.
To help you a little bit atleast to minimize the impact on
your customer or member, sothat's a good thing.
The part of it I'd say like thenext level up from there is to
implement an AI data platformwhere you bring in your data
from your AMS, your LMS, yourCRM, your marketing automation
system and so forth, and thenunify that data and then get a

(46:49):
more sophisticated AI technique,which is using something called
vectorization, to be able tocompare all of the data across
the board and find duplicateswith a much, much higher level
of precision.
That's probably a medium levelof effort.
It's not enormous, but it'snon-trivial.
You still have to set up an AIdata platform, which is a
somewhat technical topic, andthen you have to bring the data

(47:11):
in, then you have to run thisprocess for deduplication, but
it's well within the capabilityof most associations to be able
to do this.
It's just a project you have tochoose to undertake.
And then the last thing I wantto quickly say is that
duplicates, of course, are oftenthe beginning of the pain, but
they're not necessarily the end.
Another problem is decayingdata, right?

(47:33):
So a lot of times you have ascenario where you have a member
and that person worked atcompany A and then they decided
to leave and, shockingly, youwere not the first call they
made when they changed jobs.
You, the association, were notthe very first person, very
first organization they calledto inform, call to inform, but
they did update their LinkedInor maybe they updated something

(47:54):
else, right?
So what if we could have AIautomatically match people up
based on their publiclyavailable data and then update
our database very seamlessly sothat we knew where people went
right, tracking them down andupdating our database, and
that's also something you can do.
That's actually fairly loweffort level.
There's a variety of dataservices you can buy.
There's ways of doing matchingfairly easily, and then you can,

(48:17):
depending on your system.
Of course, you can update yoursystem with that new data.
So I think it's a reallyimportant concept.
That's actually kind of relatedto the duplicate data, because
one of the reasons you end upwith duplicates is because of
subtle changes.
Someone gets married andchanges their surname, someone
changes jobs, so they have adifferent email, right?
So these keys that we use totry to identify dupes oftentimes

(48:39):
are invalid because people haveslight differences If they're
literally the exact same record,even systems from the 1970s
could detect those duplicates toimplement could be kind of low
to medium range.

Mallory (48:55):
I really like, too, how you keep bringing up ways that,
like an individual can go anddo this right now.
That's really important toremember.
You don't have to like overhaulyour whole business.
You can go to chat to BT andClaude as a paid user and do
this yourself and then impact.
To me it kind of seems like itdepends because, I mean, if you
have really good data about whoyour members are and where they
work, right, you could do likesome excellent proactive
outreach to them.
So would you say this is kindof maybe like medium impact.

Amith (49:19):
I think it starts off as medium.
I think it could be very highif you implemented it kind of at
the organization wide level andmade it a consistent and nearly
automated process tocontinually have correct data.
Because think about it in yourown experience as our listeners,
the number of times you havepersonally experienced pain
because of duplicates, justdoing a lookup in your database

(49:40):
and saying, oh, I'm trying tofind Mallory, and you find eight
versions of the Mallory record.
You got to look through everyone of those to find the
transaction or the eventregistration or whatever that
you're looking for.
It's an enormous amount offriction internally.
It also tends to have a prettybig impact on external
perception or externalexperience, because if there's
multiple duplicate records theyoften have a way of kind of

(50:04):
making their way into theconsumer experience on the
website, where Mallory has aharder time logging in or she
can't find her whole history.
Not all of her transactions areon her web profile, only some
of them are.
But there's this whole otherlike phantom Mallory record
that's out there with a bunch ofother transactions.
So there are a lot of issuesthat come from this.
This is one of those.
You know, death by a thousandpaper cuts thing.

(50:25):
No association I know of hasever ceased to exist because
they had too many duplicates intheir database.
But the impact of thoseduplicate records could, in fact
, radically impact the qualityof customer experience, member
experience, cause them to needway more staff time put towards
like rote functions versushigher order things.
So I do think it's a pretty bigimpact area and, to your point

(50:48):
earlier, I really like to focuson tangible here and now, how an
average person listening tothis pod or watching us on
YouTube can put these ideas intoaction.
Today and many times we'retalking about like scaling up a
process, taking it to thesystems level.
It is actually better if youdidn't try to automate it to

(51:09):
begin with, right.
So think about, like this ideaof the automated review process
of an application.
I don't want to automate thatcompletely for an association
right off the bat.
I want to experiment with abunch of different manual steps
using AI to do the ratings, therankings, etc.
But to learn from that processto see what's working well and
say, oh, this is really good,these are great responses.

(51:30):
Now let's scale that up andautomate it Right and automate
it once we know the processmakes sense.

Mallory (51:36):
Yep, and that's exactly the process we follow here at
Sidecar Next up.
I want to talk aboutcommunications in general within
member services, but reallyservice communications and
updates.
The AI opportunity here istailored member service

(51:57):
notifications based on specificneeds and preferences.
Impact would include thingslike increased response rates
and improved member satisfactionoverall.
An example here would be anagent that can send personalized
renewal reminders with specificbenefits relevant to each
member's usage patterns, orcustomized service updates about

(52:17):
changes that affect theirparticular membership type.
Amit.
When we talk aboutpersonalization with AI, we're
talking about personalization ata one-to-one level.
I know Blue Cypress has startedexperimenting with this with a
few associations with a productcalled Rex.
Can you share any of like thepreliminary results that we've
seen from that or insights frompersonalized communications?

Amith (52:40):
Sure, rex.
So Rex is our personalizationagent and engine and Rex
actually is the brains behindthe Rasa newsletter.
So for those people that arefamiliar with Rasa's AI
newsletter, that's arecommendations tool, right?
So it just happens to be in theform of a newsletter.
Rex is the brain behind that,and so we've done a number of
interesting experiments with Rexin the area of personalized

(53:04):
networking recommendations andpersonalized event session
recommendations.
So for some organizations overthe last six months, we've set
up some very simple campaignsthat have essentially shared
through email, suggested peoplethat they could connect with at
an upcoming event and suggestedsessions that might be relevant
to them, and the response hasbeen incredible.

(53:26):
First of all, if you measuresuccess of those types of
campaigns by clicks and opens,the data has been off the charts
, you know, in excess of 100%open rates on some of these
campaigns, which sounds crazybecause people have been opening
these emails more than once.
So the distinct or unique openrates close to 100%, with click
rates that are higher than youknow, most email campaigns dream

(53:47):
of having for an open rate.
And why is that?
Why is that?
It's the same reason why AIpowered newsletters are
dramatically more effective thangeneric newsletters, because
people get stuff that's actuallyrelevant to them right.
At the end of the day, what wewant is what we want as
individuals, right, what I want,what you want.
We all have different needs andthose needs are constantly

(54:08):
changing, and up until recently,you had to be kind of at the,
you know, to be able to punch atthe weight class of an Amazon,
netflix or Microsoft to be ableto do personalization at scale.
The good news about AI gettingcontinually smarter and cheaper
is that the underlyingtechnology becomes more general
purpose, more powerful and lessexpensive all the time, and so

(54:29):
it is possible for associationsto do true personalization.
I do want to say one other quickthing that you know this idea
of segment-based approaches.
It's better than nothingsometimes, but sometimes it's
actually worse than nothing.
And so like, do you send likeone generic email to everyone or
do you try to like go throughand create personas and create
segments, which was, for a while, considered like

(54:50):
state-of-the-art both in termsof tech but also from a
marketing strategy perspective?
Right, there's entireconsulting firms who do nothing
other than like design personas.
The problem is, is they're onedimensional, meaning that you
can only fit into one of thosetypically.
So you know, you might have apersona for young professionals,
a persona for later careerprofessionals, and each of those
personas might have, you know,people, even like get pictures

(55:12):
of people and say, hey, this isJane, jane is an early career
professional, here's Harold.
Harold is a late careerprofessional, right.
And then they kind of likepersonify these people in a way
where they try to like say, ok,well, I'm going to put these
kinds of people, get this kindof content.
It's very, very, very genericand of course it's overly
generalized by definition, andso you miss a tremendous amount
that way, and sometimes I sayit's even worse in some cases

(55:35):
than like one size fits allcommunications, because at least
in the latter you're notexcluding content from people
who may find it useful, right.
So, for example, like someassociations I've talked to said
oh well, retirement planningadvice, we're going to focus on
early career professionalsbecause that's when they really
need to get started.
Early career professionals,because that's when they really
need to get started.
True, early careerprofessionals should absolutely

(55:56):
pay attention to retirementplanning right away, but a lot
of times they don't.
And so then the late careerprofessionals are like I've got
a problem, I want to retire infive years and I don't have much
of a 401k or an IRA put aside.
What do I do?
And the association often has arole to play in educating those
folks too.
So that's the problem withovergeneralization, and this

(56:16):
type of AI can really help solvefor that, which is really
exciting.

Mallory (56:21):
Difficulty to implement and potential impact.

Amith (56:24):
So the difficulty to implement, I would say, is
solidly medium.
For this, you need to implementan AI data platform of some
sort to get your data into alocation where a recommendation
engine Rex or anything elsethere's a bunch of them out
there, right?
You need to have your data inorder to do recommendations.
Absent the data, there'snothing to recommend.
So you have to get your datasomewhere and that is not a low

(56:46):
lift for most people.
And the impact I really thinkthe impact of this one's high
Mallory because it's so powerfulto get people stuff they want.
You know one little tidbit fromone of the campaigns I remember
hearing about from our team,about Rex and the event stuff we
were able to, through thistechnology, connect
professionals together in oneassociation that's in the

(57:08):
scientific domain who hadsimilar research that could
collaborate but had never meteach other even though they were
in a very narrow field ofscience, and that's really
powerful because you know thosefolks can collaborate in new
ways.
Maybe they, you know, dosomething truly amazing right in
terms of discovery.
So if AI can help bring peopletogether in that way, that's
like the fundamental thingassociations are about right to

(57:29):
associate.
So I think it's a really reallyexciting opportunity.
So I think it's a really reallyexciting opportunity.

Mallory (57:34):
The last task we want to break down here is upselling
and cross-selling.
So the current state of thisresponsibility within member
services is kind of ad hoc,probably a lot of missed
opportunities there.
The AI opportunity ispredictive product and service
recommendations and propensitymodeling, potential impact.

(57:56):
It would be an increase innon-dues revenue and enhanced
member value, and an examplehere would be an AI that can
suggest relevant courses orevents to members based on their
career stage and pastparticipation.
So this kind of falls in withthe prior one on personalized
communications.
Amit, my thought here is I wasinterested to see this included
in the deep research reportbecause I don't know that.

(58:17):
I personally have had a ton ofconversations with association
leaders around cross-selling orupselling.
So do you feel like this is animportant issue for associations
, a missed opportunity?
What are your thoughts?

Amith (58:29):
I think it's a massive missed opportunity.
It's hard to do at scale,certainly it's hard to do even
as an individual, because youhave to know a lot about the
individual that you're trying tocross-sell or upsell in order
to be useful, and you also haveto know your full catalog of
available things, right, whetherit's products or if it's events
or educational offerings.
So it's really an AI scaleproblem.
But think about it from theviewpoint of let's rewind in

(58:52):
time to, let's say, I don't know, it could be decades ago, it
could be hundreds of years ago.
Imagine you're a shoesalesperson and you work out of
a little store in a little townand you have a town of a couple
thousand people and you're thelocal shoe salesperson, right,
and you get to know all yourcustomers.
You know them, you know theirfamilies and you also probably

(59:14):
remember a lot of their purchasehistory.
You know when did they buy,let's say, dressier shoes or
when did they buy casual shoes,and so, therefore, when that
customer comes in, if you have anew product that's relevant to
them because let's say it's a,you know, super powerful like
break resistant shoelace thathas come out that you know my

(59:35):
teenager would love to havebecause he's always going
through shoelaces, right.
So and that comes out and I say, hey, that'd be great.
But, like, if the customercoming in doesn't use shoes,
that use shoelaces, maybe it'snot a great thing to cross sell.
That's probably a silly example, but the point is, is in that
old school example you had asalesperson who knows their

(59:56):
customer really well, he's knownthem probably for quite a while
and also knows their productcatalog of their 50 or a hundred
products, whatever they sell,really really well.
So they are the brain at theintersection of products to
offer and the customers they'retrying to sell them to.
They also have timing, you know, built in right?
Because if the customercustomer just spent $100 on a
pair of shoes which to me stillsounds like a lot of money, but

(01:00:18):
I realize shoes cost a lot morethan that now typically but in
any event, they're probably notgoing to buy another $100 pair
of shoes tomorrow because theyjust bought one today, right?
So all of these various factorsgo into how to make
recommendations work.
And upselling and cross-sellingis one of the most beautiful
aspects of it because, yes,you're generating more revenue

(01:00:38):
for your association.
But let's say that I knowMallory really well and I say,
hey, there's this course weoffer.
Mallory, that would be perfectfor you.
You are already really strongin AI and all of these
categories, but there's this onearea that I know you've told me
in the past you were interestedin learning more about, and we
just introduced a new course inthis area.
It's reasonably priced, it'savailable online at your

(01:00:59):
convenience.
What do you think?
And Mallory was like crap.
That sounds awesome.
It's like exactly what I washoping to learn and I can do it
on my own time, and it's only$15 or whatever, right it's like
reasonably priced.
So that's the beauty ofcross-selling and upselling is
that you're creating more valuefor everyone involved.
But once again to your firstpoint, mallory.
It requires goodrecommendations.

Mallory (01:01:19):
Yeah, impact.
I think for this one soundslike it could be high,
incredibly high for yourassociation Difficulty to
implement.
I feel like if you need all ofthat background, that backstory,
like in the shoe salesmanexample of your customer or your
members, that seems like intheory it would be a high
difficulty to implement.
What do you think?

Amith (01:01:37):
I generally agree with that.
I think there's lower versionsof that that are easier, where
you can do it through like,let's say, emails, but, if you
want to like, integrate thiswith your website.
That tends to be an area that'sharder for people to implement.
Prerequisite number one getyour data house in order.
Get all of your data into an AIdata platform of your choosing.
That means wiring your datainto one data environment from

(01:01:59):
your AMS, your LMS, your CMS,your FMS, all the MSs right.
Bring them all in.
Then you can run arecommendation engine to
understand the relationshipsamongst all these.
You know disparate entities ofdata and then you can build the
functionality to say now I knowwhat I should be recommending.
Let me connect people withproducts or people with people
or people with events in theright way.

(01:02:20):
But you can't really upsell andcross-sell intelligently until
you have the recommendations inplace.
So I think you're right.
I think it's effectively.
It's a medium to high level ofimplementation difficulty, but
the impact is extraordinary ofimplementation difficulty, but
the impact is extraordinary.

Mallory (01:02:41):
Well, everyone, thank you for tuning in to part one of
our AI Enhanced Member Servicesepisode.
We are really excited to get topart two, where we'll take all
this inspiration that you'refeeling and kind of talk about
the practical next steps of howyou can implement this Reminder.
We broke down everything taskby task for ease of explanation,
but the real transformativepower here is when you start
thinking about bundling some ofthese tasks into one kind of

(01:03:02):
overarching way to transformyour member services.
So with that, we will see youall in part two.

Amith (01:03:10):
Thanks for tuning into Sidecar Sync this week.
Looking to dive deeper?
Download your free copy of ournew book Ascend Unlocking the
Power of AI for Associations atascendbookorg.
It's packed with insights topower your association's journey
with AI.
And remember, sidecar is herewith more resources, from
webinars to boot camps, to helpyou stay ahead in the

(01:03:33):
association world.
We'll catch you in the nextepisode.
Until then, keep learning, keepgrowing and keep disrupting.

All Episodes

AI-Enhanced Member Services: What You Need to Know | 73

Episode Transcript

Popular Podcasts

Are You A Charlotte?

On Purpose with Jay Shetty

Dateline NBC

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}AI-Enhanced Member Services: What You Need to Know | 73