Full transcript
0:05uh we're jumping right into the next
0:07topic um uh which is why AI projects
0:11fail The Missing Link of a strong data
0:14strategy our speakers Barbara perovich
0:18and Linda St please come to the
0:26stage we have you have the
0:32hi everyone Welcome to our presentation
0:34I will first start with a very quick
0:36introduction of who we are and our
0:38company and then we'll dive straight
0:40into the topic of today's presentation
0:42which is why so many AI projects fail
0:45and the missing link of a strong data
0:47strategy so my name is Linda St and I'm
0:50here today with my colleague barara
0:52perovich and we've both been working in
0:54the field of math and AI for a very long
0:56time and after a while we figured out
0:59that unfortunately data cannot solve all
1:02our problems and today we work mostly
1:05with data strategy and culture Which is
1:07far messier than any of the data sets
1:09that I have ever seen oh and by the way
1:12we often get called Ghostbusters but
1:14that's fine because we love to solve a
1:16good
1:17problem now barara and I have been
1:19working together for over 10 years and
1:22about five years ago we decided to start
1:25a company and our company is called Data
1:27coose but since we're from the
1:29Netherlands we call it data
1:31quaza in our company we provide training
1:34and strategy for diverse governmental
1:37energy companies financial companies
1:39educational uh and many more sectors and
1:43in the past five years we've actually
1:45helped to fine-tune and Define many data
1:51strategies and actually we have quite
1:54diverse customers as you can see we do
1:57some commercial customers like G TMobile
2:00but our sweet spots are actually
2:02customers who are not driving profit our
2:05sweet spot our governmental organization
2:08institution that really bring value to
2:10the society like universities and think
2:14organization that want to change the
2:16world with quite less budgets than the
2:18all other commercial and usually what
2:21they also have in common is the fact uh
2:25that they have failed miserably with
2:28quite a lot of data and project they
2:31have done it quite a lot of times they
2:33have spent quite a lot of money that
2:35they usually don't have and when we come
2:38in they are tired of guessing they are
2:41tired of trying the new thing the new
2:43Hypes they want strategic advice and
2:46they want answers they usually call us
2:49after the big fours are there so when we
2:52do and then you know I'm a geek I'm
2:54really are there any Geeks in in this
2:57room so if I ask you what what's the
3:00answer to the all of the questions that
3:02they are there what's the universal
3:05answer I'm hitchhiker's guide and the
3:08answer to everything is 42 exactly so I
3:12would love to say to them it's 42 but
3:15unfortunately in 2024 the answer to all
3:19of our data and AI issue is not 42 we
3:23hear it this morning it's 80 that's the
3:26magical number and the Magical number is
3:29actually the fact that for several
3:31service we see that business we heard
3:34business does not understand business
3:36does understand the value of AI over 80%
3:40of Executives and leaders are really
3:43grasping what it means to change your
3:45business with AI over stunning 87% of
3:51the organization are actually putting
3:54data and as their top investment
3:57priority all of their money is going
3:59down there and still We are failing 80%
4:02of the time a projects are failing which
4:05is twice the number of failing of IT
4:08project so and there are quite a lot of
4:11reasons and quina will tell you
4:14something about those
4:16reasons yes of course we all want to
4:19know why so many AI projects fail and
4:21actually ranty Research Institute has
4:24done quite an extensive research into
4:26this and they came up with five leading
4:29causes root causes of AI project failure
4:33the first being that often our
4:35stakeholders or leaders don't know the
4:38actual problem that they want to solve
4:40with AI or they don't communicate which
4:42problem that they want to solve and I
4:45have seen a lot of AI models that have
4:47been trained to solve a completely
4:49different problem that than that we want
4:52to solve or have been trained on the
4:54wrong
4:55mattresses the second being is that we
4:59simply don't have the data if the data
5:02is not there to train our AI model for
5:05our organization we cannot use AI or if
5:08the data is not good enough we cannot
5:10use it the third being and this is a
5:13very common one is that we all want the
5:16shiny tool we want the technique
5:18technology first and then the problem so
5:20we find find a tool that we really like
5:23we want to implement it we want that
5:24specific AI model and we completely
5:27forget which problem we're actually
5:29trying to solve or even worse with
5:31trying to find the
5:34problem and then sometimes we simply
5:36don't have the infrastructure we don't
5:38have the tools or the techniques to
5:41build our
5:42Ai and the last one and this is actually
5:45quite common and as I said in the
5:47beginning of our presentation
5:49unfortunately data cannot solve all my
5:51problems and also AI cannot be solved
5:55cannot solve everything in every use
5:57case so what I see very often is that we
6:00find a problem that is simply just too
6:03big too complicated to fix with AI and
6:06then it will
6:08fail and as we said we we are real
6:12strategist we we know quite a lot about
6:15data driven and botom up approaches and
6:18how to fix we can code I mean we can do
6:21python we can do but that's not our
6:23sweet spot our sweet spot when you know
6:25the biggest mistakes are made you know
6:27when the biggest mistakes are made we
6:29hear it in presentation before it's act
6:31actually when leadership fails and
6:34Leadership driven failures that's when
6:37strategy kicks in because for example
6:40optimizing for the wrong business
6:42problem we had a um really nice Hospital
6:46which was actually one of the leading
6:47hospitals in the world who was trying to
6:50get help to help as many people as
6:52possible saying we want to optimize the
6:56problem of our operational space because
6:58in Netherlands we do have we have a
6:59crowded country so we don't have space
7:01to build so we have to optimize we are
7:04like you know and we will try to fix the
7:07every minute that there is like shipping
7:09and in shipping and out not because of
7:11the profit because of the fact that we
7:13are really trying to help so much people
7:15and this was a really nice strategy we
7:17taught it we worked with universities
7:20with startups with researchers from the
7:22hospital itself and we crack the problem
7:25you know we fixed the optimalization you
7:28can just come in and we will precisely
7:30know when you have to get out of that uh
7:32operational room and we will fix
7:34everything clean everything and the next
7:36one can go in we never implemented it
7:40did you know why because top surgeons
7:43are artists and you don't cut in artist
7:47time that's and also the reason why
7:49people were coming to this hospital was
7:51not the fact that we have the past
7:53efficiency time was the fact that we had
7:56those to surgeons all of the time we had
7:59the perfect thing solving something that
8:02we should never try to
8:04solve and then the other one we always
8:06see is using AI to solve simple problems
8:09and the one that we usually see with
8:11financial institution in work for quite
8:13a lot is fraud detection using Ai and
8:16machine learning to detect
8:18fraud when the simple no you cannot die
8:21twice to probably prevent quite a lot of
8:24misused cases in fraud so you will see
8:26that we actually use AI for if then El's
8:30questions of even something as stupid as
8:33answering your email this beautiful
8:35matte product we are setting to Sol
8:37problems that should actually to do
8:39tasks that should never be done
8:41overconfidence in AI quite a a lot of
8:44our leaders are looking for for way out
8:48and when you don't have don't have a way
8:50out then you actually put your
8:53confidence in something that it
8:55departments comes in or some other hype
8:58and think that that's going to magically
9:00solve all your problems and even if the
9:03product is good which we usually see
9:05products are quite good you try to
9:09implement something which was not
9:10optimized for your organization you
9:13think you will buy it you will put it
9:14into your organization tomorrow
9:16everything is Sol and you can go to your
9:18board meeting and say we are in in in we
9:21are now in green we are not red anymore
9:24but you don't un you underestimate time
9:26and costs but these are you know these
9:29are the easy questions if you really
9:32think about it if you really talk with
9:34business people and you explain all
9:36those things to them they will say yes
9:38we understand and we can grasp this and
9:41we can actually put aside this kind of
9:44failures and
9:46still 50% of them will also fail how
9:50many of you were at the key presentation
9:52this
9:53morning did you hear the key
9:55presentation seven since yeah and how
9:58many of you have actually agreed with
10:01the fact and do you know actually how
10:03the co Kodak
10:05failed how many of you see the co Cod
10:07fail
10:08story yeah and you know what the problem
10:12with that story is it's totally not
10:15true uh the reason why people call us in
10:18because and actually the reason why they
10:20call us Ghostbusters because we go after
10:23The Ghost and let's just bust this myth
10:27because Kodak did not fail because they
10:30did not jump on the technology train in
10:33contrary um Kodak was usually people say
10:37C are blinded by their success they
10:39didn't want so they missed the digital
10:42technology but that's not true actually
10:45dig Kodak was leading and invented the
10:48first digital camera in
10:521975 but stepen Samson and he was an
10:55engineer working at Kodak it was not a
10:57nice thing it was a toaster you know
10:59know and it will took 20 seconds to take
11:01one picture but it was state-ofthe-art
11:04at the time uh people say yeah but you
11:07know what that same Steve sson said it's
11:11cute but don't tell anybody about it
11:13because we don't want to invest
11:15money so yes they found out the
11:18technology but they didn't want to
11:19invest but that's also not true because
11:23actually the company spent around five
11:26billions on research and development
11:28related not just for producing cameras
11:31but a field of digital imaging 1970s
11:35guys they invested so much money and if
11:38just to put it in into perspective this
11:40little guy uh does anybody know who that
11:43is that's Rover NASA it's actually yeah
11:48for in this little guy to bring them on
11:50Mars costed 1.5 B billion so quite less
11:55or the research and development budget
11:58of Apple in 2010 was 2.4 did you see the
12:02amount of money that they actually spent
12:04to increase this so yes they were
12:06willing like all of us now willing to
12:08spend quite a lot of money of AI but
12:11they said you know what um it was they
12:14they had the money and they wanted it
12:16but they actually were going to that
12:18scene of Pride you know they wanted to
12:20be the best they wanted actually to
12:23focus on that film quality instead of
12:25digital
12:27Simplicity and also that is also not
12:29true it was true for the first prototype
12:32you know that big toaster thing that
12:33they wanted to build but later on they
12:35produced models that are quite simple
12:37and easy to use and relatively cheap so
12:40they jumped on that
12:42wagon but they said yeah but then they
12:44missed the paring you know that they
12:47missed the real disruption and the real
12:49disruption they missed the fact that
12:52when people started using cameras which
12:55emerged on the phones people started
12:57actually sharing photos instead of
13:00printing it and even this one is not
13:03entirely true because Kodak acquired the
13:06photo sharing a site called a photo in
13:09201 they guys this is way before
13:12Facebook existed they wanted to push
13:15this do you know what the mistake that
13:17they made they focused on
13:20efficiency they actually tried not only
13:24to share but also to get people to
13:27utilize their services that they they
13:29already had in a better way because it
13:31was more
13:32efficient and they are focused on
13:35maximizing their profits and I see in
13:38almost every presentation today that
13:41that focus is creeping in we are using a
13:43technology which is kind of disruptive
13:46and we are focusing on maximizing things
13:49and profits in one process how do you
13:53think that will go and this is a common
13:57trap yes this is actually the Trap of
14:00marginal thinking and this is when an
14:03established company that seemingly does
14:05everything right that's right everything
14:07and did the right things still fails and
14:10this is because they focus on their
14:13established already existing customers
14:16and already existing
14:18Services there's actually a beautiful
14:20book written about it by Clayton
14:22Christenson that we really recommend if
14:24you would like to read it um but we also
14:26see this marginal thinking in AI if you
14:29look at these numbers you see that most
14:31of the AI projects focus on efficiency
14:35on creating better customer experience
14:39or making more uh processes more uh
14:43reliable but actually only
14:464.7% of AI projects does something new
14:50is an
14:53innovation now this is where data
14:55strategy comes in and unfortunately we
14:58cannot give you you a framework a
15:00standard data strategy framework that
15:02you can just Implement tomorrow I cannot
15:04give you an action plan that's easy to
15:06follow and you can just do however we
15:10can teach you about some of the
15:12practices that we use as strategist on
15:15our day-to-day job and that we have used
15:17in all the data strategy projects over
15:19the past few years the first being is to
15:22really focus on the job to be done look
15:25at what your customer actually wants
15:28what do they want to to experience
15:29what's their job to be done not just
15:32focus on margins but look at what they
15:34want to achieve and that's where you
15:37should adapt your strategy to the second
15:40being to establish a framework in which
15:43you can actually uh measure your
15:46long-term goals the organizational goals
15:49instead of just focusing on short-term
15:51wins and
15:54margins and the third one is that our
15:57customers even when they hire us to
15:59create a data strategy a lot of times
16:02ask us for a specific solution I want
16:05this I want self-service bi I want to
16:07Data Warehouse a new platform and we
16:10always present them with all alternative
16:13paths show them what else can be done
16:16and what the outcome then could be we
16:18always give them counter factual
16:21analysis and the last one and this is
16:23not the least one is to avoid the
16:25cheering trap actually if you want to
16:28remember something from our presentation
16:30please just just remember the touring
16:32trap and the touring trap is something
16:35that's called by some of the misbeliefs
16:38and one of the biggest misconceptions
16:41about AI especially my researchers is
16:44the fact that we think that it needs to
16:47replace tasks that human beings are
16:50doing and it needs to do it
16:52efficiently and you know this is not the
16:55case because then we are to going into
16:58that Mar thinking of efficiency and we
17:01are not creating the Strategic Advantage
17:04for organization we are just optimizing
17:07and there are ton of other ways to
17:08optimize instead of using AI which are
17:11much for better and there is also one
17:14other thing with a touring tra it's not
17:16also not beneficial for your own company
17:19but it's also hurting the society the
17:22way how we now think and approach AI
17:24it's not going to change the world and
17:26if it changes it's changing in a way
17:29that's actually causing us to go into
17:32part of inequality environmental issues
17:36using chat GPT which is three bottles of
17:38that water every time you ask a stupid
17:40question just to be you know to answer
17:44what to to automize a task we should not
17:47be doing at all but there is there is a
17:51sweet spot there is something why we're
17:53still there 30 years after the first
17:56line of code that I programmed and I I
17:59still believe that we can do it and
18:01that's the fact that we can avoid this
18:04trap by actually realizing that there is
18:08far more opportunity in automating
18:10humans to do new tasks instead of
18:13automating but already everybody is
18:16doing because you know if you really
18:18want strategic ad Advantage you have to
18:21go beyond and you have to dare to
18:24abandon maybe some stupid things that
18:26you're already doing after
18:29that's what strategy
18:31is yes strategy is about setting
18:33yourself apart from the competition it's
18:37no matter about being better at what you
18:39do it's about being different at what
18:41you do that's what this beautiful quote
18:44says guys I hope you liked our
18:46presentation I hope we inspired you or
18:48learned you something uh if you have any
18:50questions or if you would like a list of
18:52reference that we have used in this
18:54presentation please feel free to reach
18:56out and also and also this
18:59uh actually we're powered by undp and
19:03empowering uh uh women in data and Ai
19:06and we do a lot of coaching TR projects
19:09so if there are any female data experts
19:13looking for mentorship free of charge
19:15advice anything we can do just send us a
19:18shout out LinkedIn and let's conquer
19:21this world thanks guys thank you guys uh
19:25we do have time well firstly a little
19:28thank you for from us thank
19:31you um uh are there any questions for
19:35Barbara and
19:40Linda no uh well then I guess you can
19:43catch them afterwards in the hallway
19:46thank you so thank you again thank you
19:48bye