Latest / The DEX Show: A Show for IT Change Makers / Reality Bytes #46: AI, Magic, and Nexthink (w/ Mohamed Kafsi)

Transcript

Reality Bytes #46: AI, Magic, and Nexthink (w/ Mohamed Kafsi) · The DEX Show: A Show for IT Change Makers

0:01
Welcome Item change Makers to the Deck show with Tim Flower and Tom McGrath.
0:07
Let's get into it. Hello, change Makers, Welcome to Reality Bites show
0:14
within a show within a show where the Reality Bites team and I in this
0:18
case, we've got Dina and Shawn with us get to talk to fascinating Next
0:23
Thinkers. You know what it's all about by now, and we have got
0:26
an absolutely fascinating Next Thinker today. I don't know if you've noticed, but
0:31
the show is increasingly preoccupied with AI and DEX and it does feel like just
0:36
an inexhaustibly fascinating subject. We're really enjoying syncing our teeth into it, and
0:42
we've got a special emphasis on it across the month of June, in fact,
0:47
so keep your eye out for other episodes coming along in this vein.
0:52
And we are joined very appropriately by Next Think's director of AI, and that's
0:58
Mohammed Cafsi. Mohammed, welcome to the show. Really excited to have you
1:02
here, sir, Thank you, Tom, thank you for having me happy
1:04
as well to be here today with you. Fantastic and we had a couple
1:08
of conversations with you at this point. Always really enjoyed them and I'm really
1:12
looking forward to getting stuck in. But I guess a great place to begin
1:19
is to hear a bit about your professional background. How have you come to
1:22
be a director of AI, Mohammed? What brought you to next thing?
1:26
Yes, Tom, So let me start from the beginning. So I was
1:32
born and raised in Tunisia and North Africa, and after high school, I
1:38
felt that I needed to discover the world. So I went, I traveled,
1:42
and I studied in different countries in Switzerland and the US, Germany and
1:48
Finland. And along the way, I mean, I knew that I was
1:56
really passionate about first learning and also about technology as many of us and many
2:01
of the listeners here today. And I was exposed throughout the way through to
2:07
AI, starting in like twenty twenty five years ago with like data and well
2:14
AI was a little bit narrowars I went also through you know, different winters
2:19
and summers of AI. And I've worked for different companies where bisicy the focus
2:23
was to make to produce insight out of the data, to automate to have
2:29
a better understanding of the world through AI and data. And I worked for
2:37
Yahoo, I worked Nay, I worked for Swiss Com which is a tere
2:39
Com operator here in Switzerland, and I always since I've been in Switzerland,
2:45
I heard a lot about Next Thing. Actually funny enough, actually the first
2:51
time I heard about it is through Yes which also a fellow Tunisian, and
2:55
who's the chief strategy officer today? And then I think we got in got
2:59
A three or four years ago and we discussed about nexting and opportunities at Next
3:02
THINGK And this is how I joined because I knew Next Think was I did
3:07
a company and also an AI company since its inception, so I was very
3:13
exciting there saw a lot of opportunities to join and to solve the interesting problem
3:17
together. I was going to say, like Next Think, you go back
3:22
more than two decades in your your interest in AI, right, and the
3:27
sounds of the regular listeners one there Next being founded of course during our CEO
3:32
and founders AI postgraduate study. Just before we move on, Mohammed, what
3:38
what was it about Ai've attracted you to it? Particularly those years ago when
3:43
it wasn't it wasn't such a ubiquitous topic of discussion and point of interest.
3:52
What attracted me? I mean the fact in addition to the fact that I
3:54
was already attracted to technology. It's the fact that we're AI works. It
4:03
feels like magic, right, the kind of thing that an AI is able
4:08
to do. I mean we see it now recently with everything around generative AI
4:12
with ch GPT, but also before if you go come back, when you
4:15
see like an AI, I don't know, hot performing humans and games and
4:19
chess, there is like a feeling of magic right when you see it working
4:24
right, And for me being very curious, there was always a need and
4:29
a drive to understand what's behind and how this somehow the black box is able
4:35
to make sense out of data, to train and to learn from data and
4:40
to perform cognitive tasks that are usually associated with humans will increase complexity. This
4:46
attracted me like a lot to AI and a need to understand what's behind and
4:51
how we can apply to solve like meaningful problems right in our field today and
4:58
dex but also beyond that, right for health, for education, like very
5:02
central topics that are very important to humans. I believe personally, I continue
5:06
believing that AI can ring as a lot to make progress that's really well put.
5:12
And I actually have said before here that first time I used chat GPT
5:16
a year or so ago, I guess I was struck by its magic in
5:21
that moment, that first ever question and answer, because I gave it quite
5:26
a complex task, you know, it was hard on it in a way
5:30
I haven't in my whole life been struck by a technology, because I feel
5:33
like all other technologies around us have been relatively incremental, hugely impressive, but
5:39
relatively incremental. But AI seems to have sort of for people who aren't already
5:43
immersed in it seems to have just sort of shot out of the blue somewhat
5:46
and really can kind of blow you away. And so that as being a
5:50
director of AI. Next thing, what does that break down into for you
5:54
today? What? What? What? What do you oversee here? And
5:58
just to come back to your point, and then I will answer, especially
6:00
if you if you don't mind, if you you know, very very quickly
6:05
we got used to chatipety and now for us it's given, and you know,
6:09
we're naturally less less impressed as we use it, right, because somehow
6:13
our baseline is getting higher and higher. But if you take someone from three
6:17
years ago right and show this person like stragipity today, they will not believe
6:25
what they what they see, right, I mean the jump we had and
6:29
the ability to understand the user and to answer with something that makes sense,
6:33
and the the ability also to almost mimic reasoning. Right sometimes now we say
6:39
it's mimi in reasoning, but this is like it's a big, big advance
6:43
we've made in the recently. And your question one sorry about yes, how
6:50
does how does being director of AI breakdown? And next thing what do you
6:55
ever see? What are your what are your areas of OSI? What is
6:59
your jump? Yeah? Uh, I mean generally it's really to to to
7:09
as we said at the beginning. And next thing is is a data company?
7:13
Uh, data and a companies since it's an inception, but if we
7:17
think of the all the data we gather and and next thing, I think
7:23
if you want to make sense out of the data, produce insights. I
7:28
mean AI is a key technology for that, right and me I am involved
7:32
in different initiatives And next thing where we where we work with data in order
7:40
to produce insights, but also to to to to make our products easier and
7:48
simpler and more enjoyable for the user interaction, help the user have a more
7:55
like a simpler journey in the in our product. And we have also some
7:59
other projects are forward looking, but so how I see it today is really
8:03
my me. I'm trying to help her. There is like value with AI
8:07
and trying to to to make it happen and work with different teams in order
8:11
to integrate AI in different parts of the product. So we have the example
8:16
of cloud andags right where we and this is more classic AI. That means
8:22
we we make sense of the data we have, we learn from all the
8:24
data we have, and we provide recommendation, a personalized recommendation to our users
8:31
and customers. Right if we think of it as like and and the consumer
8:35
market is like Netflix looking at all movies it has and providing very personalized recommendation
8:43
to you. And this is what we do with cloudations really at a at
8:48
a very large scale. And then there is everything which we do around generative
8:54
AI. If we think of assists really having a copilot in the product,
8:58
that helping our users navigate the product in the simplest way to interact with the
9:03
product, like in a conversational manner, right, and also generate content in
9:09
the most natural way. Yeah, I know, we've got to show and
9:16
dealer with us as well, both of them itching to ask you some additional
9:18
questions on this female and it's I'll have the mic ovid to shown for his
9:22
first question. Yes, hey, Muhammed, I was curious about asking you
9:30
just how you think AI in DEX particularly is going to pan out in the
9:37
next couple of years and specifically, like what's its potential as an AI driven
9:41
platform to improve employee engagement and productivity. I think the interesting interesting aspect around
9:52
AI in how it will impact DEX in particular, but also other like other
9:56
industries. I think what we what we can do with AI is really provide
10:03
a very personalized experience right to the to the end users and to the employee
10:09
by making sense out of the making sense of the data we're collecting, right
10:16
in order to provide very like personalized and intuitive like DEX experience for employee.
10:26
And this is like the personalization side is basically what we think of classic AI,
10:30
right, as I said before, everything around the recommendation that we provide
10:33
that you can see in the consumer market, so basically offering you the right
10:35
product at the right moment for you as a user. And then the second
10:43
part is it is really about what we see today with generative AI, right,
10:48
that's also all today dominating the discussions is really to allow users and employees
10:58
to interact with our product and our offering in the most natural way. It
11:03
means we have the ability today and a natural language to understand the user intent
11:13
and also provide it with the like with the right answer at the right one.
11:16
So those two elements together like a personalized recommended a personalized experience, but
11:22
also a very natural interaction the understanding of the user intent are two elements that
11:28
allow to to have this great experience for users and employees. Excellent, and
11:37
actually continue with that theme, I'm going to skip a question and then come
11:41
back because of what you just said, So do you have like any specific
11:46
examples than of AI customizing the workplace to individual needs like with its user intent?
11:50
Can you maybe just sketch a common use case for us so variously interesting
11:58
use case I would like to share with you today on whichual working is.
12:03
And we are really in the exploration phase to see how how we can bring
12:09
value to our customers with that. So everything around personalized it right if we
12:13
think of for example, let's take the example of hardware. If you want
12:20
to recommend like a computer or our laptop to an employee. Right, we're
12:28
not all the same. We don't have different needs. We have different digital
12:31
needs. And I think with the scale of the data we have about like
12:37
laptop usage, performance performance with certain applications, we have the data in order
12:46
to give you the best recommendation for the best hardware for an employee. Right,
12:52
So we're working on what we call in personalize it in order to understand
12:56
and to recommend for every employee what would be the best where for them to
13:01
have a great user experience. So this is one example among others where we
13:05
can basically a personalized it T give you a recommendation that suits you the best.
13:13
And then there is also everything out personalized IT resolution, right, I
13:22
mean in the sense if we if we understand the users, and we know
13:35
the users, we know that certain aspects, like certain IT issues might impact
13:39
their day to day work much more than others. Right. We are not
13:41
all impacted equally by an issue I don't know without or an issue with like
13:46
a given software. Right, some are more impacted than others. And I
13:50
think this fact you can think of it if you think of it as like
13:56
an it T agent or that knows you well, that knows your habit that
14:00
knows what are the things you use the most, so everything are out personalized,
14:03
making the it support more and more personalized to provide you with the best
14:09
experience is also something where AI can will play hero excellent. Thank you for
14:15
those specific examples and now kind of conversely, this might be a bit more
14:20
esoteric, but how do you think those things like contribute to a person's well
14:26
being in job satisfaction? I think, you know we're talking about I mean
14:37
more, I mean the more I see today like the today, like the
14:43
you know, the the the frontier between the personal life and work life is
14:50
like it's very malleable and it's evolving a lot, but it's I think it's
14:56
very important for us to be to be well in our personal life and as
15:01
well in our work life, right, And I think the ability to have
15:07
a having a great like digital digital experience at work is like it is one
15:15
of the key aspects of coming to work, like you know, happy and
15:18
relaxed and not frustrated. And I think those parts of having providing a very
15:26
great dig digital experience two employees also providing a very personalized, you know,
15:33
personalized, it very personalized. It support something where things are running smoothly is
15:39
definitely like a key contributor to the to the well being of the employee and
15:43
then at the end the well being of a person and absolute right. So
15:48
all those aspects I think of personalization also the aspect of having interaction and the
15:54
and the the most natural way in order to to to either make our experience
16:00
more enjoyable or solve certain issues we have will definitely contribute to the to our
16:06
well being. How much so, what are the emerging trends in AI that
16:15
could further impact the digital employee experience going forward? So we know now,
16:22
I mean with with with what we see in l LM's like a large anguage
16:26
model h this is already given right, the fact that we're interacting with HGPT
16:34
people I mean they, I mean they see value and that the ability to
16:37
be able to to to express in ourselves in the most natural way in natural
16:45
language and get answers. I mean, this is today already there and we
16:51
see it more and more integrated and different parts of the product we use either
16:55
in personal life or the work life. I think what's most exciting for me
16:59
what's coming next is what something that's basically building on top of large language models
17:04
ability to reason. Right, And this is what they call like agents.
17:08
Right, it means agents. Simply think of it as like it's a virtual
17:14
agent that is using a large language model as a reasoning engine. Right.
17:19
It means an agent is able to take an input like an issue, and
17:26
then reason come up with a plan in order to solve this issue. The
17:33
plan is basically is a set of steps and every step is using potentially a
17:38
tool. A tool it can be a web search, it can be like
17:41
nexting as a product, it can be an API, a public API,
17:45
and then come up with a plan in order to solve this issue. And
17:49
here's exciting because first it's not an easy problem, right, it's a problem
17:57
when we ask the agent to be able to reason and to iterate and to
18:03
solving a problem. And then it's exciting because it's basically it's a promise for
18:08
us to automate more and more tasks we we we don't want to do.
18:15
And if we think of this ability of agents now of reasoning in order to
18:19
solve a problem, reasoning and iterating, right, and then after that it's
18:26
comes a step where we have like let's say, an ecosystem of agents right.
18:29
Let's imagine an agent is an expert and nexting like it's like, think
18:34
of it as a power user of nexting, right, and you have another
18:38
agent that is a power user of an i SM system. Then a third
18:44
agent that is an expert I don't know in hardware and how now can we
18:48
have the technology in order for them to collaborate to fix to solve a given
18:52
problem? Right? And here I mean it's a it's a very active research
18:57
field where the research community industry is also pushing a lot. But I think
19:07
it's quite exciting and I think it will have a big impact also on the
19:14
dext industry. Right because as an example I just get before, if you
19:18
are able to build like an agent that's the best and next thing it will
19:22
be and the agent that is the best, the best like it support agent,
19:26
they will be able to interact to solve issues in a really autonomous manner.
19:33
So we're not there yet, but it's quite exciting and I think it's
19:34
something we should continue to investigate. Super exciting. It's as if you're building
19:41
subject matter experts of all of the amazing topics dynamically letting them speak in one
19:47
ecosystem at the end of the day, right, instead of going to multiple
19:49
channels to get answers, or like you mentioned earlier, like the reasoning mindset
19:55
of adapting and advancing. I think one of the most interesting and exciting ones
20:00
that are coming up is also the fact that they keep on learning, so
20:03
you don't have to reteach them every time the contextual reasoning. This gives a
20:08
more advanced experience. So with a lot of speculation and even anxiety for some
20:14
people about the potential impact of AI on their careers, whether it's an IT
20:19
or other industries, it's discredible in your view. And what do you think
20:23
is the secret to effective AI and human employee collaboration? Yeah, I mean
20:30
this is a question that's but I personally got a lot and also a question
20:36
that's very I mean very valid right about. I think it's beyond AI,
20:41
right, it's also about technologies. When you have a new technologies, it's
20:42
always perceived as like as a potentially a threat for certain for certain jobs.
20:52
Me and how I see it personally, and it's my optimistic view of the
20:55
world. I don't see AI as automating. I see it as really automating
21:02
tasks, okay, and tasks that we would be happy not to do in
21:07
order for us to focus on on on something where we're uh more fulfilled and
21:12
happier. Right, So me, I see that as really more an opportunity
21:19
rather rather than than a threat. And it means if we have the opportunity
21:26
of AI automating our tasks like either my task or yours or me you know,
21:30
I'm I'm a big fan and biased of using like CHGPT to automate many
21:36
tasks I do, right, so let's keep it here. But like to
21:38
write emails I don't and then to to iterate, to brainstorm and to those
21:45
are the kinds of tests I'm happy to automate, to to to AI so
21:48
I can focus on something that I enjoy more. So I see it really
21:52
has an opportunity and not not not a threat, and I see it as
22:00
a as a way to augment us really not to replace as as humans.
22:07
You know, I discussed before our efforts and the research effort we're doing the
22:11
next thing AI in order to explore how we can how we can improve in
22:18
the automated certain parts of it support how I see it is that for me
22:22
as it as a way of augmenting like for example at support agents. Right,
22:26
providing the right information at the right moment to solve an IT issue is
22:30
definitely something that allows to augment and not necessarily to replace. Totally agree.
22:37
It's complementary to what we do, and I think it's not going to dehumanize
22:41
us in any kind of manner. On the contrary, it will give us
22:45
more time to be humans, right. And I think that's the key point
22:48
that when we're teaching people more and more about integrating AI in their day to
22:53
day life, we need to keep on like bringing it back. They're not
22:56
going to be humans. They are just complementing you just said exactly, taking
23:00
the mundane tasks so we can do the more fun stuff. Yeah, And
23:06
as I said, I think you have also a very valued point that I
23:08
didn't mention. I think we have to see it as an opportunity and also
23:14
for us to say, look, this is a new technology. It's a
23:17
tool I could use to automate certain tasks. So I should like, you
23:23
know, focus, learn and see how I should use it in my day
23:26
to day dependently of the job I'm doing, and then you know, be
23:30
able to focus on certain parts I enjoy best. So I think education and
23:36
not feeling the tool itself, but learning about how to use it is a
23:38
very important point. Completely agree what excites you the most about the next thing?
23:44
AI roadmap in the coming years. So as I said, I think
23:53
what expired me the more is really to explore how we can use new technology
24:00
and so large language models but also agents in order to offer the best copilot
24:10
for IT. Support it means building you know, it's like today if we
24:22
think of IT support, most of the job is really about understanding, okay,
24:26
what's the issue? Right, So discussing with the user it means discussing
24:30
with the employee who has the issue, but also checking different parts. Either
24:34
it can be next thing, but it can be other products. And then
24:37
in order to have a clear understanding of the problem and once this is done,
24:41
then acting on the problem. Right, can we have that, like
24:45
say, IT support copilot that is able to do this job for the IT
24:51
agers and automating the resolution. Because once an ITE agent have seen a problem
24:56
one or two times, they know I mean, when you see the third
25:00
time, it's really you know, repetitively, they have to do it again
25:02
and again. So can we have a compiler that's able first to understand the
25:07
problem and represent it to the IT agent and then allow the IT agent to
25:11
automate once it's like the resolution is clear, right, and can this IT
25:21
virtual agent right take care for example, of the interaction with the user to
25:25
clarify the problem right, perhaps it can be once in the user have an
25:30
issue, you know, usually like the issue description is very short, so
25:34
the can we automate this part where we have like a virtual agent triggering the
25:41
discussion with the user through teams, through Slack or any other chat to clarify
25:47
the problem to get more and more information, enough information than to bring it
25:52
to the IT support agent, but also take care of using automatically other APIs.
26:00
It means say, look, I mean the user is complaining about the
26:03
device being very slow. Let me go and create the API of next thing
26:07
to get the right information about this device, integrate that in the issue description
26:11
and then provide this right. And once we have that, can we also
26:15
automatic automated the resolution. And here's the challenge, As I said, from
26:19
the technology respective, it comes back to the ability of the virtual agent to
26:23
plan. It means understand the problem, plan resolution and then execute it right.
26:29
And this I find quite exciting, and I think the next Thing,
26:33
we are a very good position because we know the business and we have the
26:37
data. So we need to build the technology that is basically taking advantage of
26:42
our business knowledge but also the all the curabilities and the data we'll bring with
26:48
next Thing in order to provide the best compiled for the It spot absolutely fascinating
26:57
discussion Byhamed. We will. We do have a last question for you.
27:00
It's that sort of traditional reality by it's corny question, but just wanted to
27:06
say you speak, You speak very, very engagingly about AI. It was
27:11
a great, great discussion. I'm disappointed that Megan couldn't be with us today
27:15
because I know she's currently knee deep in researching the history of AI, and
27:18
I think we'd have a lot of questions. You might end up kind of
27:22
listening to this and following up with you, I expect, and I was
27:25
surprised to hear Dina that you are you are you are an AI optimist.
27:30
At this point, I still think of you as like the terminated to Peck
27:33
dystopian on the panel, but they obviously just get muddled up between your love
27:37
of sci fi dystopia, so it's the mix of both, right, I
27:41
love my human side and still I love my tech side, so I need
27:47
to balance it out between to maybe being there exactly exactly. Well, I'm
27:52
gonna still be kind to all of the AI's out there. I still know
27:55
they need to grow up and mature a little bit more, but I'm not
28:00
moving away from using them every day and everything that we're doing more or less.
28:04
However, I'm urging humans. So it's two messaging, one for the
28:07
humans and one for the I. Here has to be more humans as to
28:11
just keep on learning. Dina is as polite and tries to set a good
28:15
example to a I. She tries to influence it if she thinks we should
28:19
all be influencing its benign as possible, which sounds sensible. But yeah,
28:26
just for listeners, if you missed our discussion with Pedro I Think two episodes
28:32
ago. Now he talks about his the origins of next Think in AI research
28:37
and some of the sort of larger technical questions overhanging the future of AI in
28:41
decks really really worth the listen. We had Bill Shanginger last week, absolutely
28:47
fascinating discussion where he touched upon the kind of dynamic between AI and employee experience
28:52
and what the implications of AI and a workplace might be over the next few
28:56
years. And we have some fantastic episodes, the longest theme still to come,
29:03
and I should say, you know, look, if you if you
29:04
want a really really deep deep dive or an intensely invigorating download, you should
29:11
be thinking about coming to experience being in Boston or London or both. Towards
29:17
the end of the year. We will have links in the show notes so
29:19
that you can book your ticket right away. For that just reminds me to
29:23
ask you that that corny question, Muhammad before we let you go, which
29:29
is what do you most like about being a Next Thinker? Ah, this
29:33
is an easy question. I was expecting something very nasty. It's an easy
29:37
one. I mean, you're already touched on the technology and next thing being
29:42
a data company, and I would not add on this. I think what's
29:47
Let me talk about the human aspect, which is I think for me is
29:51
the most important that at work. I mean, it's it's a great culture,
30:00
an atmosphere at Nexting. It's the good balance. I mean, we
30:04
have been working with different colleagues. It's a great mindset of working all of
30:08
us as really one team working together towards a common goal, always having working
30:17
hard, but also having fun here in Lausanne. And this is something I
30:23
really liked and I discovered when I joined Next Thing. And also the culture
30:27
of you know, being getting things done, being really impact driven, showing
30:34
things rather than telling this all mindset of builders and building things, that's something
30:40
i'd like. So yeah, a lot of fun, I mean, great
30:44
colleagues and something I enjoy well put. And you know we do have I
30:48
think we sort of product development side in particular. Love that sense of the
30:52
kind of united purpose and solving the problems that are presented to you collectively is
30:59
obviously something body we speak to you gets all that out of. Maybe I
31:02
should include vacancies in the show notes as well. I don't know, maybe
31:06
I will. You know, you'll have to check them and see if I
31:07
did or not. Just remains for us to say, absolutely fascinating discussion.
31:11
Lovely to have you on, Michhewe Muhammed, and I think your role at
31:14
Next Think is so important, so interesting. We're sure to ask you back
31:18
on if you're If you're if you'll honor us with a repeat visit in the
31:21
future. Set. I would love to come back. Please invite me.
31:23
I have a lot of things to say. Splendid, splendid, thank you
31:26
so much, Thank you everybody, and we'll see you sin. To make
31:30
sure that you never miss an episode, subscribe to the show in Apple Podcasts,
31:34
Spotify, or your favorite podcast player, And if you're listening on Apple
31:37
podcasts, make sure to leave a rating of the show. Just have the
31:41
number of stars you think the podcast deserves. If you'd like to learn more
31:45
about how Next Thing can help me improve your digital employee experience, head over
31:49
to next think dot com. Thank you so much for listening. Until next time,