Embed code
Note: this content has been automatically generated.
00:00:00
also thanks to digits with one for having me it's another to um be able
00:00:05
to present here what we do in the area of artificial intelligence at cisco
00:00:11
so my plan for this talk is to tell you first of all
00:00:16
why does come is interested in artificial intelligence um i guess
00:00:21
it's not that kind of thing you'd expect from um
00:00:24
classical togo company um second how um
00:00:29
um what is our procedure for for bringing artificial intelligence or customers
00:00:33
and the third um name some examples of products that we've there we are
00:00:40
developing and that we've developed um recently that to use artificial intelligence
00:00:47
so let's get started
00:00:50
it's
00:00:52
so just ionisation it's um one of the things that
00:00:57
i think power is um artificial intelligence and it's
00:01:01
nothing you know it started probably with the uh
00:01:04
first microchips and has seen a really
00:01:08
huge progress since then so there was the first wave when the smart phones
00:01:13
arrive which um brought a huge change
00:01:17
there is a currently um where at the top of
00:01:22
the um coyote and rejoice asian waves so
00:01:26
um it's very easy to have computing
00:01:29
resources um they're very very cheap
00:01:32
very easy to set up in the cloud and so on
00:01:35
and um we're right at the beginning of the
00:01:39
third wave which is the artificial intelligence wave
00:01:44
uh which means that machines will be learning and make or
00:01:48
are already starting to learn and make attachments decisions
00:01:51
um am i think huge quantities of data that
00:01:55
um previously was impossible to dig into
00:01:58
um and i guess there also be a
00:02:01
lot of robotics breakthroughs um and other
00:02:06
areas and we don't know yet what the fourth we will be but we are very excited find out
00:02:12
'kay so digits lies they shouldn't and all these
00:02:17
waves have changed a lot about our business
00:02:23
so one of the things that has come made a lot of money one on where
00:02:28
um or were text messages and this is
00:02:31
something that has disappeared practically overnight
00:02:36
so companies like what tap um are i don't know
00:02:40
any other of your favourite messaging apps have
00:02:43
taken over and no one uses s. m. s. anymore or at least very few people
00:02:49
no everything is turning into software softer is eating the world
00:02:54
and some famous people agree with this among which market reason maybe you've heard of him
00:03:01
um so says come has the ambition to follow
00:03:05
this um this trend and um of
00:03:09
course it's also kind of necessary in order to to stay alive as a company
00:03:15
buses commons turning from a classical togo company into um um softer company
00:03:21
um and i guess this starts to give a first explanation of why we're interested in a i
00:03:28
oh here i'm showing um and events that is organised basis come and maybe some of you
00:03:35
know 'cause it's um become relatively begins with concerts comes after a happen last month
00:03:43
along with business also customer um behaviour changes so people expect
00:03:49
different things from us than they did before higher expectations
00:03:54
for example um like i said the advent of smart
00:03:57
phones and um orgy networks and so i'm made
00:04:02
uh the mobile that data volume increase dramatically
00:04:06
um people are watching lots of video even him on there
00:04:10
are mostly on their smart phones um people are using
00:04:14
catchup t. v. so no longer live t. v. but
00:04:19
um recording functions and net flicks and so on
00:04:25
and very very few people um are using the
00:04:29
still the fixed a phone network so
00:04:32
again another very clear signal that we need to change and we believe a i
00:04:41
the change that we need to incorporate and here we uh
00:04:45
i highly agree with a ah it's apples video
00:04:49
game cock was set that a. i has to be horizontal so we need to put a i into
00:04:55
basically everything that we built because people expect software to
00:04:59
behave intelligently you can no longer build a
00:05:02
that i'm stupid product and get away with it and this brings us also to our
00:05:09
core belief in the um yeah in machine learning robots was gone which is
00:05:13
um that in the future our product will be stupid without a high
00:05:18
and therefore worthless so we obviously we can tell the right so hopefully
00:05:25
this gives a good motivation of y. says compares about a i
00:05:30
and now let me tell you how we plan to bring a i to our customers so
00:05:36
maybe just as a side remark i i'm in the
00:05:39
enterprise customer um departments are talking about um
00:05:43
other companies not not private customers alright so um
00:05:53
what we do is build digital enterprise solution
00:05:55
and we have lots of resources already to build on um
00:06:00
there are so this department the different resolutions consist
00:06:05
of seven hundred specialist employees we're building um
00:06:10
enable or as we call them some sort of basic softer building blocks that week and then
00:06:15
put together based on customer needs we have um
00:06:20
an advantage citation change h. m. m.
00:06:23
tell you about more in the next slide we have tools um
00:06:28
and we are used to working in um in um
00:06:33
in a pretty high temple so really fast really i doubt that's on
00:06:40
oh here is our unique selling proposition at the digital and resolutions um first of all
00:06:48
over time we have been able to build and ten
00:06:50
competence that means we have infrastructure we're able to
00:06:54
a store and process data out both firms is
00:06:57
coming from our other business customers we um
00:07:02
um we're competent in software development in different platforms and so on
00:07:07
and what enabled us to do this was our a very comprehensive
00:07:13
three hundred sixty degrees portfolio including both standard based solutions
00:07:17
those things that i don't know product stuff to products from other
00:07:21
companies that we provide consultants for um but also customisable
00:07:26
so where we talk to the customary find out they have a problem
00:07:31
it cannot be solved by the standardised solution so we need to
00:07:35
um sit together and build something as specifically for that customer
00:07:41
and finally all of this would so certainly not be successful
00:07:44
without um i tried methodology and we have this as
00:07:47
well most importantly um we believe in the co creation
00:07:52
process so like i said we we sit together
00:07:55
or with our customers with around and fries customers and we find out what problems they have in order to
00:08:01
to find the best solution for them instead of just
00:08:04
selling um at the standard thing that we have
00:08:09
and this is um also uh i guess a
00:08:13
key component in a human centre design philosophy
00:08:19
uh where you try to find out what users want and
00:08:22
um uh how to best um fill their needs
00:08:27
we also work according to agile methodology i guess everyone does today so is it even worth mentioning but anyway
00:08:35
um so we plan to you um integrate a i into this
00:08:42
whole um overview that we already have in dealing with customers
00:08:49
and hope that will be successful and it doesn't look so you'll see when i present the product
00:08:57
right the first of all what exactly do we understand with a high
00:09:01
and i start with an example from the moment they're born into this
00:09:05
world children start learning things about it for example a here
00:09:11
a chocolate what is struck what hoping that the kids but their hands in a day this married on their face they
00:09:18
tasted and so on to the the experience directly with all their senses then finally they learn what chocolate is
00:09:24
and a i is is very similar in the sense that um
00:09:29
it enables machines to gradually learn things about the world
00:09:35
that we humans know so this is them a more generic definition of a i
00:09:41
of course there are also um i think nowadays mostly uh use cases of
00:09:47
right now our artificial intelligence where you have a very specific use case let's say uh you
00:09:53
write an email and you write a time in dayton there um and the the
00:09:59
program is able to extract isn't put in a calendar
00:10:02
is already an example of narrow artificial intelligence
00:10:08
but artificial intelligence is an umbrella term and is just means that computers act
00:10:13
intelligently similar to how we humans i'm
00:10:17
acting different situations and it includes
00:10:20
um various either um feel it's
00:10:24
like um expert systems
00:10:27
so these are i guess in the seventies and eighties what the the
00:10:32
state of the art um of artificial intelligence at that time where
00:10:36
um experts were writing rules for computers to behave in a certain way based on their knowledge
00:10:42
um obviously the more a recent um
00:10:48
important area artificial intelligence is machine learning and
00:10:51
particularly deep learning where computers can um
00:10:57
become intelligent through statistics on large quantities of data
00:11:03
and
00:11:06
oh this has uh in order to to do something like this
00:11:11
and i uh lots of things are needed and they're
00:11:16
different areas in which you can apply it
00:11:21
there's first of all perceptions so obviously a a i.
00:11:25
systems must interact with humans so we they
00:11:28
need to be able to um perceive festival speech
00:11:32
vision and so on there is um
00:11:36
natural language processing so just because they and they are able to perceive the speech
00:11:40
doesn't mean they understand what the person saying says what natural language processing dies
00:11:45
um there's reasoning there's planning and of course that somewhere up there's there there's alter products
00:11:54
and machine learning which i view as a subset of artificial intelligence is also
00:12:00
huge so in order to achieve all these things that i've shown here
00:12:04
there are lots of different methods there's um unsupervised learning there's
00:12:09
um supervised learning semi supervised learning neural
00:12:12
networks um him support vector machines
00:12:16
clustering and so on there's also reinforcement learning a bit it's
00:12:20
um let's related to the rest of the the the tools that
00:12:26
we have at our disposal in terms of methods that say
00:12:34
but either ingredients are just as important so first of all
00:12:40
processing power is very very important
00:12:45
data obviously because computers learn from data as
00:12:48
i've said we statistical methods and algorithms
00:12:52
so i think there is my why a eyes seeing such a huge a surgeon
00:12:58
in popularity is because in all of these area there has been tremendous
00:13:04
uh progress and i'm past couple of years
00:13:09
so you pressing power already a bit longer than that we've we've um
00:13:14
um for quite awhile gone from computers looking like this to having um
00:13:21
to having a enormous computational power in our pockets or in our briefcases
00:13:28
um data as i've said uh with a smart phones getting popular
00:13:33
the amount of data generated by people has also increased exponentially
00:13:39
and finally algorithms so i guess where the the latest thing greedy and just to
00:13:46
um become due to arrive into a very advanced state
00:13:52
um there's tons of low that you also hear about i guess after the stock
00:13:58
um microsoft is also opening um various of their
00:14:03
a. i. algorithms to the public face but
00:14:06
the same i was recently heard about by to a releasing
00:14:10
something in the area so there's lots going on
00:14:16
and this means that people like us we're we're as i
00:14:19
said we're a softer slash local company we we
00:14:23
don't have the ambition to compete with will or face
00:14:26
bookends one but me i'm a wants to
00:14:32
to uh um enable our customers to benefit
00:14:36
from um from all these things
00:14:40
as soon as possible and to do so in a way that suits them best
00:14:49
obviously there is a there are also lots of misconceptions with all the hype surrounding yeah
00:14:54
i so um machine learning is expensive it's a real time activity everything can be
00:15:00
done with with people earning these are all the kinds of things that we have to
00:15:05
sort of fight against um when we start working with people on a i
00:15:11
is the expectations are sometimes quite um misaligned with reality
00:15:19
and so let me show you some examples of what we do
00:15:24
some products that we've built um in the recent future in the
00:15:29
recent past sorry or that we're building at the moment
00:15:35
our products are based loosely on this um framework
00:15:41
that i'm showing here so we have um
00:15:45
obviously have data you have lots of data from system itself
00:15:49
we can uh also take advantage of of public data
00:15:53
at that anyone's disposal and we may may also get customer data
00:15:57
depending on the customer um we have a perception liar
00:16:01
where um first of all um there is speech text um then there's
00:16:07
perception in the sense of compute of of the teaching computer is what the
00:16:11
data represents and what they should do with it so we have a
00:16:15
um model training um component
00:16:19
and then finally we have a bunch of micro services um in the
00:16:24
area of natural language i think an understanding for the moment um
00:16:31
uh but extending to other areas as well so we have named entity recognition topic
00:16:35
recognition sentiment analysis summarisation a trend detection and
00:16:40
so on and so we um
00:16:45
have the services uh as part of an a. p. i. and the goal is to use this tape yeah
00:16:51
in order to build you their room highly customised solutions
00:16:56
are really sitting down with the customers and
00:16:58
i'm finding out what they want and building it for that we also
00:17:02
have sort of a more standardised set of products where we
00:17:08
uh that we use for example to showcase the different micro services
00:17:12
so for example this customer feedback is something that everyone is interested in
00:17:17
and um we do have a a product where we gather
00:17:23
um data for example first this come from social media
00:17:28
and analysts it by using these uh different services
00:17:34
and finally we also have um the the plan
00:17:37
of integrating with either a a softer products
00:17:42
that can highly benefit from from artificial intelligence example service now um
00:17:49
which is the customer desk software a. c. p. sales force
00:17:53
and so on and you'll see an moment why i think these
00:17:57
are appropriate um standards after products to be looking at
00:18:04
so our first um
00:18:08
interesting book um product that we've still using using or artificial
00:18:13
intelligence it's them is the customer desk um product
00:18:18
it helps service agents give faster and better service and how does
00:18:23
it do this by using a i obviously called mom all
00:18:29
and we um provide basically
00:18:34
realtime recommendations um
00:18:37
of solutions to the service agents to solve the customer's problem
00:18:42
well basically the agents um and has a custom on the phone
00:18:46
that explains the problem um the agent right sin incident ticket
00:18:52
and that in the background um our artificial intelligence service is looking
00:18:57
for is similar to get that is already been sought
00:19:00
and or for several similar tickets let's say and presents um
00:19:04
a list of um candidate solutions to the agent
00:19:10
and the agent can and uh choose the best solution for the particular case and in this way solve
00:19:17
the problem a lot faster than if they had to go and look through this manually um
00:19:23
either well the customers on the phone which would i guess would be quite um anoint
00:19:27
for the customer or at a later time and then get back to the custom
00:19:32
oh this is what the um the interface looks like him in this this
00:19:40
has been integrated in this or is this after of space gone
00:19:44
basically it's the this is the part where we we enter the
00:19:47
to the new ticket and this is where the magic happens
00:19:51
so showing a plastic it's with pro with solutions and highlighting what
00:19:58
um the main topic of this the current ticket is
00:20:03
and um there's also functionality that um looks
00:20:09
at the various incidents coming and and
00:20:13
notifies if there is something that looks like a major incidents so it's a lot
00:20:18
of people calling within a very short period of time about exactly the
00:20:22
same thing this means it's probably um the underlying causes probably similar and it's
00:20:27
the major incident to look at um at at all the incidence together
00:20:34
though this um oh is something like a site that
00:20:38
has been integrated into our hours just come
00:20:41
service desk a softer but it's definitely a reasonable
00:20:46
and and probably advantageous to integrated also in
00:20:51
in other service tax stuff for like service now there's mentioning previous
00:20:57
the second product that it developed is i'm also not
00:21:05
and in this product our goal is to use artificial intelligence for customer experience
00:21:11
and also to to showcase our various micro services
00:21:16
so
00:21:18
the question or this the problem that this product solves is
00:21:24
um i think customer feedback or uh providing an analysis of customer beat
00:21:28
feedback to the right people in order to improve customer experience
00:21:36
and i guess here you have a screen shot of an earlier version of the project of the product
00:21:43
it's basically a product where we um it with some of the
00:21:47
the two biggest functionalities are sentiment analysis where we look
00:21:52
at at a piece of feedback and say is this negative feedback with this positive feedback and obviously this of
00:21:59
it's a very a um and sought after functionality in
00:22:04
in terms of customer feedback or reviews in general
00:22:09
and the second sat functionality is being able
00:22:12
to recognise in the customer feedback entities
00:22:17
uh named entity recognition searches um names of people this
00:22:22
might be the customer already age and that they
00:22:24
talk to if they are uh referring to a customer interaction um it might be an address
00:22:30
or a and r. o. another company a phone number or a date anything
00:22:37
um and this helps pack the um the the feedbacks
00:22:42
um according to what they're about and then
00:22:47
providing an overview of a set of feedbacks through these um keywords that we extract
00:22:55
it's also useful when um when we went to a sort of
00:23:02
democracy tight feedback and to provide access to customer feedback
00:23:05
to everyone at the company which is what we've done and says come with this product in that case um
00:23:12
what needs to be then is uh in order to comply
00:23:17
with data protection laws is an amazing customer information so
00:23:22
so we also do this wasn't with named entity recognition so in order for
00:23:25
for every employee of says come to be able to look at
00:23:30
customer feedback and internalise it and find out what they yeah at their
00:23:35
individual level can do better in order to improve the customer experience
00:23:40
we need to make this data anonymous we cannot go around and and um
00:23:44
mm open it's a incident tickets from the hotline to everyone
00:23:52
oh this is worked out pretty well i'm in the the
00:23:55
product has been quite successful instead of says come
00:24:01
and now the plan is widths sonar to make
00:24:04
it more widely available and perhaps um also
00:24:09
tell it to other companies that might be interested in that is what it looks like
00:24:13
we can select language um obviously we're very
00:24:17
interested to have um great performance on
00:24:20
swiss languages a german french italian maybe not amend it actually they are
00:24:29
um so for this uh which it is something that we're still working on gathering data for this
00:24:34
languages and i'm coming up with trained models that
00:24:39
work best on this particular case um
00:24:44
i'm back to sonar we can filter um according to sentiment
00:24:48
um we can filter according to the data source so
00:24:52
um we're using for example peter and face book from social meet yeah but we also have internal sources
00:24:58
like i said the hotline is one example but we could also
00:25:01
think of uh for example um online online chat um
00:25:06
software that step we have one that's just come website where you can go into live with an agent
00:25:13
and then down here you see sort of summaries of what's the the
00:25:18
feedback um i says in the time frame that you selected
00:25:24
and then there is another visualisation is based on the analysis that we're doing
00:25:33
we also have topics again based on the time frame that we select you can extract
00:25:38
um a different topics the the most important
00:25:41
topics that the feedbacks um are mentioning
00:25:47
and finally is see the uh actual raw data as though the feedbacks of the people
00:25:53
and you can go all look more into that if that's what you're interested in
00:25:59
though this um sums up the products that that we have so far
00:26:05
um for the future so we have customer experiences we have customer support
00:26:13
we we have a um a product in progress for human
00:26:17
resources where you we can an allies um employee surveys
00:26:22
and and come up with summaries a lot faster than than the current
00:26:26
uh way of doing it manually um we have a um
00:26:32
product planned for uh the legal domain um as
00:26:37
a a big company as a big service
00:26:39
company they're obviously lots of contracts with other
00:26:43
companies and very often this number is
00:26:46
so high that it's very uh tedious to to an lies to conduct manually and have a good
00:26:51
overview of this so this again one very good use case artificial intelligence and so on
00:27:01
and just to finalise i would like to say i'm
00:27:06
against his come we're uh we're uh an
00:27:09
former tell co oh it turning into a service company so we're not starting
00:27:14
bad big with artificial intelligence but on the other hand everyone started small
00:27:20
well all of these big sed a big currently very big companies have started i'm

Share this talk: 


Conference program

Keynote
Jean-Baptiste Clion, Coordinator DevFest Switzerland
26 Nov. 2016 · 9:40 a.m.
How to convince organization to adopt a new technology
Daria Mühlethaler, Swisscom / Zürich, Switzerland
26 Nov. 2016 · 10:14 a.m.
Q&A - How to convince organization to adopt a new technology
Daria Mühlethaler, Swisscom / Zürich, Switzerland
26 Nov. 2016 · 10:38 a.m.
Animations for a better user experience
Lorica Claesson, Nordic Usability / Zürich, Switzerland
26 Nov. 2016 · 11:01 a.m.
Q&A - Animations for a better user experience
Lorica Claesson, Nordic Usability / Zürich, Switzerland
26 Nov. 2016 · 11:27 a.m.
Artificial Intelligence at Swisscom
Andreea Hossmann, Swisscom / Bern, Switzerland
26 Nov. 2016 · 1:01 p.m.
Q&A - Artificial Intelligence at Swisscom
Andreea Hossmann, Swisscom / Bern, Switzerland
26 Nov. 2016 · 1:29 p.m.
An introduction to TensorFlow
Mihaela Rosca, Google / London, England
26 Nov. 2016 · 2:01 p.m.
Q&A - An introduction to TensorFlow
Mihaela Rosca, Google
26 Nov. 2016 · 2:35 p.m.
Limbic system using Tensorflow
Gema Parreño Piqueras, Tetuan Valley / Madrid, Spain
26 Nov. 2016 · 3:31 p.m.
Q&A - Limbic system using Tensorflow
Gema Parreño Piqueras, Tetuan Valley / Madrid, Spain
26 Nov. 2016 · 4:04 p.m.
How Docker revolutionized the IT landscape
Vadim Bauer, 8gears AG / Zürich, Switzerland
26 Nov. 2016 · 4:32 p.m.
Closing Remarks
Jacques Supcik, Professeur, Filière Télécommunications, Institut iSIS, HEFr
26 Nov. 2016 · 5:11 p.m.
Rosie: clean use case framework
Jorge Barroso, Karumi / Madrid, Spain
27 Nov. 2016 · 10:05 a.m.
Q&A - Rosie: clean use case framework
Jorge Barroso, Karumi / Madrid, Spain
27 Nov. 2016 · 10:39 a.m.
The Firebase tier for your app
Matteo Bonifazi, Technogym / Cesena, Italy
27 Nov. 2016 · 10:49 a.m.
Q&A - The Firebase tier for your app
Matteo Bonifazi, Technogym / Cesena, Italy
27 Nov. 2016 · 11:32 a.m.
PERFMATTERS for Android
Hasan Hosgel, ImmobilienScout24 / Berlin, Germany
27 Nov. 2016 · 11:45 a.m.
Q&A - PERFMATTERS for Android
Hasan Hosgel, ImmobilienScout24 / Berlin, Germany
27 Nov. 2016 · 12:22 p.m.
Managing your online presence on Google Search
John Mueller, Google / Zürich, Switzerland
27 Nov. 2016 · 1:29 p.m.
Q&A - Managing your online presence on Google Search
John Mueller, Google / Zürich, Switzerland
27 Nov. 2016 · 2:02 p.m.
Design for Conversation
Henrik Vendelbo, The Digital Gap / Zurich, Switzerland
27 Nov. 2016 · 2:30 p.m.
Q&A - Design for Conversation
Henrik Vendelbo, The Digital Gap / Zurich, Switzerland
27 Nov. 2016 · 3:09 p.m.
Firebase with Angular 2 - the perfect match
Christoffer Noring, OVO Energy / London, England
27 Nov. 2016 · 4:05 p.m.
Q&A - Firebase with Angular 2 - the perfect match
Christoffer Noring, OVO Energy / London, England
27 Nov. 2016 · 4:33 p.m.
Wanna more fire? - Let's try polymerfire!
Sofiya Huts, JustAnswer / Lviv, Ukraine
27 Nov. 2016 · 5 p.m.
Q&A - Wanna more fire? - Let's try polymerfire!
Sofiya Huts, JustAnswer / Lviv, Ukraine
27 Nov. 2016 · 5:38 p.m.
Closing Remarks
Panel
27 Nov. 2016 · 5:44 p.m.

Recommended talks

Etat de l'art : Hypothèses et corollaires du Cloud computing
Edouard Bugnion, Professeur EPFL
25 Oct. 2017 · 5:14 p.m.
Une app sans map ? Vous n’y pensez pas !
Emanuele Gennai, ESRI Suisse
31 Oct. 2017 · 10:34 a.m.