The AI Summit New York 2019
December 12 2019, New York, NY, United States
The AI Summit New York 2019
Video
AI Summit New York: Architecting Large Scale AI Adoption in Legacy Organizations
Available
In cart
Free
Free
Free
Free
Free
Free
Add to favorites
20
I like 0
I dislike 0
Available
In cart
Free
Free
Free
Free
Free
Free
  • Description
  • Transcript
  • Discussion

About speaker

  • Salahuddin Khawaja
    Managing Director - Automation / Global Risk at Bank of America Merrill Lynch

About the talk

A keynote session from the AI Summit New York 2019 by Salahuddin Khawaja, Managing Director for Automation and Global Risk at the Bank of America

Share

All right. I think we're good to go. Sit digital transformation automation Lessons From The Trenches. Let me give a good quick caveat and then let's 00:00 get into it the views that I'm expressing are my own and not of my employer. So recent a Mackenzie study talked about a massive opportunity to 00:08 automate knowledge worker tasks. It's the size of 7 trillion dollars. And if we want to understand the opportunity we have 00:18 to think about what happened in the past and the way to look into that is basically asked the question. Do we live in the worst of times? So if you 00:28

listen to the talking heads on TV You get all sorts of bad news every single day. Even the mood of fish get a breaking news alert, 00:36 but if you flip the question and you basically asked do we live in the best of times? One stat after another shows that the human 00:46 quality of life has vastly improved from you know, a child from different C2 life expectancy to 00:56 child mortality rates. And this is significant not just a few percent Point material double-digit growth over the last 50 60 years. And this 01:06

is as a result of a series of revolutions and it started with the printing press and if you look at these revolutions, you can look at it in the frame 01:16 of Economics these revolutions brought the cost of something down. And with the printing, press did it brought the cost of publishing books down. And 01:24 so we went from an oral to Electric culture. And as far as the European Renaissance and then came the Industrial Revolution and what textile mills the 01:33 cost of textiles came down. And then came the steam engine in the cost of Transportation came down and then came heavy engineering that led to Big 01:43

infrastructure projects steamships and that brought the cost of trade down. And then we had heavy industry and this basically ushered in the era of 01:52 mass production and this brought the cost of goods down. And finally we have the computer Revolution. And if you look at it from the perspective of 02:01 cost again computers brought the cost of arithmetic down. So you no longer needed people to calculate where Cannonball would fall you can actually 02:10 have a computer do it and this basically led to the software Revolution where we had about twenty odd unicorns or 1 billion-dollar startups in the 02:19

first half of this decade, but subsequently what you see is we accelerated the pace of change we've had over 200 unicorns or 02:28 the last five years, and that number is just increasing. So this is a concept that Mark and receive calls software eating the world. 02:38 And now if you connect this software with AI Revolution, these are two revolutions that are happening than currently unlike the prior of elution, but 02:48 there's a lot more that's happening. You've got robotics biotech blockchain 3D printing Quantum Computing solar energy. And that's the time we live in 02:57

everything happened concurrently. Let's talk a bit about Trends and then we'll kind of get into how we can take advantage of all of this. 03:06 So if something gets ten times cheaper, it will get adopted. So here's an example of the hard drive from IBM in the 1950s. This 03:15 is a 5 megabyte hard drive now 32 GB size of my thumb. This is a phone from the 90s and this is what we now 03:25 carry in our pockets everyday. Here's another example of a lighter lighter is what enables the driverless aspect of a driverless 03:35

car initially this cost $75,000 to make so if you're the inventor working at a car company and you go to the CFO and your average car cost 03:45 is $25,000 and you tell us you and let's all the car 400,000. He left you are the room but the interesting thing is 03:55 that would be one of the most short-sighted decisions that the CFO could make because the cost of a lidar today is $500. And this is just in a matter 04:05 of a few years to the cost significantly comes down and it will get adopted. The other interesting trend is that in just less than 10 years, right or 04:15

actually 10 years ago. We had just one identity get a physical identity and that was it. Ever since just in 10 years we have now have a digital 04:25 identity. And that's part of why you see all of the turmoil that surround us. We're suddenly dealing with the two identities and this is also 04:33 connected to another Trend around power structures. So if you think about the 6,000 years of human recorded history communication always 04:42 happened in a broadcast mechanism weather was the speed of the horse a book radio or TV with the internet, right? All of those power structures 04:52

are being broken because we now own communication. So if you think about you know, Emperor's Kings dictators generals that top down power 05:02 structure CEO's post. All of that is breaking down as we speak because we now own communication 05:11 The other interesting thing is that as we adopt Technologies problem spaces shift. So when the camera was first invented, the problem was hate should 05:22 I take this picture now? It's become the filtering problem. I take too many pictures which ones are good. So as you all adopt technology AI 05:31

the problem space is going to shift you have to keep that in mind. and I Bank of America the way we think about intelligent automation is that 05:40 knowledge workers spend way too much time on manual processing and less time on the more value-added side to things so we built an automation system 05:50 that actually replaces the manual processing and this frees up people time to do some of the more smarter work and that's kind 05:59 of the journey my team and me we're on If you think about what an average knowledge worker does they get information they 06:09

perform logic and they reach a conclusion. So we've built a system that does exactly that our system gathers information our system 06:19 performs the logic that humans perform today in our system reaches the conclusion. And what this allows us to do is it allows us 06:29 or actually allows the machine to do the heavy lifting and we can actually detect a normal. He's at a much faster clip. Example we use as a branch in 06:39 Miami suddenly is opening up more accounts are typically does. So we detected an anomaly. Let's go investigate it. Maybe they were running a marketing 06:49

campaign. And maybe it was all legitimate and maybe it wasn't but letting the machine detect these anomalies that humans can't because of the number 06:56 of Records, right? We can actually move up the value chain. And the other thing is that we don't look at automation from a singularity perspective. 07:05 We don't think is just automation. We're connecting the dots and we're looking at it from a full-fledged platform perspective. So we're connecting it 07:15 2-1 information is going to be readily available to people were connecting it to workflow. So if you think about the FedEx model, you know exactly 07:23

where your package is. So for the task that we automate we want to understand where the automation is worth stock worth moving. What's the manual 07:31 aspect and then it's also kind of connected to alert notification just like Instagram and Linkedin when you wake up in the morning, you get alerts and 07:39 notifications it guides you the same thing with our automation system automation task Grand last night go look at it the task failed go look at it. 07:48 And then it's connected lastly to search. It's a Google like search capability that empowers users themselves to go search for information. 07:57

What this allows us to do is analogous to in the world of driverless cars the job of an Uber drivers going to ship from behind the wheel 08:08 to behind a computer screen. So this control room setting is what we're aspiring to kind of get to. Did you think about 08:18 our various business units from consumer banking wealth management Investment Banking trading RCF organization and our back office. What we're trying 08:28 to do with automation. What is basically connecting the dots? Enabling all of these control points to kind of talk to each other that don't talk to 08:38

each other. So now let's go back to economics. How do you want to think about AI is very 08:47 technical subject. But what a I basically does is it brings the cost of prediction down if you think about a doctor in the first ten years of their 08:57 practice, they typically see 10,000 patients. And they get better and better at evaluating subsequent patients. What what what does a I do it takes 09:06 millions of patient records and it can help predict. What ails you So what's fueling the AI Revolution and that's where this analogy works and other 09:15

aspects that doesn't work. But data is what's fueling this to all those millions of Records. That's how you can train AI models. 09:24 So this is where we spend a lot of time on the data lifecycle side to things and it really starts with having a business understanding of the data. 09:35 And from there. You can get into Data Discovery and sourcing And from there you do data governance, 09:43 and this is something that's really important to us because it's directly connected to our firm strategy of responsible growth. We're not going to 09:52

move fast and we're not going to break. Thanks soap for us. Basically we are going to actually get approval to use the right sets of data from the 10:00 right places. And then you actually can make the connections and once you make the connections to data, then you can prepare it and once you prepare 10:09 it that's when all the inside start coming out and this is where a lot of the stuff that you've heard throughout the very sessions today. It 10:18 starts with exploratory data analysis. And then you get into feature engineering you get into training the models and then you get into sharing 10:28

collaboration and you repeat this over and over again and you visualize along the way and we do something called Model risk. And again, it's tied to 10:35 we're not moving fast. We're not going to break things are models are approved internally by independent parties. And their validated month-over-month 10:44 year-over-year. And the other thing that we find fascinating and again, it's tied to our firm strategy of responsible growth. Sometimes 10:53 AI models can go off the rails like the Microsoft model that became racist. So we we really want to believe in responsible Ai and 11:03

for that light has to be shown into the AI box and there's some fascinating research that's coming out of MIT that actually shed light into what is 11:13 making those decisions. If you connect all these dots around automation will we see in front of us is the 7 trillion dollar opportunity of 11:22 automating non productive knowledge work. So fundamentally, how do you take advantage of all of this? 11:31 It's at the heart of it is an imagination problem. Could you have imagined 10 years ago that you're going to take your phone out of your pocket hit a 11:41

button in a car is going to be waiting outside. So in your respective areas, what are you imagining? What are you thinking about? That's all I'm going 11:49 to talk about five ideas. If you can Implement these 5 ideas you can take advantage of this revolution. 11:59 It starts with an a big organization like ours and I'm sure a lot of you are in big companies the nervous system attacks when you're trying to 12:09 innovate there people that are just set in the old way and when you try to implement change the nervous system attacks, so the key here is 12:17

basically innovate on the outside. So what match.com CEO did here in New York, he cleared out of floor. He heard a bunch of brand new people 12:26 and he said I want you to kill Match.com and that group of people, you know, what they came up with they came up with Tinder. And when match.com 12:36 went public 60% of match.com valuation was based on Tinder. Today they were actually innovating themselves were out in debating 12:45 themselves. So simply speaking. How can you be 10% different how individual you're going to be 10% different your group your company. So when the 12:55

competition comes sure they come for your tail because you're out innovating yourself. Another angle to think about is having 13:05 a massive transformational purpose, right if you think about like the Ted conference is Right ideas worth spreading Google organizing the world's 13:15 information lofty aspirational goals that people can rally behind and in the area that we're implementing automation system is an area called process 13:25 testing and our lofty goal is test everything everyday. And this is something that's near and dear to 13:34

my heart. I I drink a child Kool-Aid and then we'll talk more about this where as product owners. You don't keep it simple. Keep it focused 13:44 avoid product dress figure out what your true north is align it to the customer and focus on that. 13:52 And part of that focuses design thinking right where you're iterating through through a series of four steps right discover, Define develop deliver 14:02 keep doing this over and over and over again and you can experiment your way through product development. A b testing along the way customer feedback 14:11

along the way. And then what's interesting is diffusions minimal a world economic Forum talks about how only like 25 to 29% 14:20 of knowledge worker tasks have been automated and in the next five years will just get to 40% is a long way to go here. This isn't a hundred meter 14:30 dash, right? And what was interesting is if you think about it from a platform perspective, right did the lightbulb come first or the electricity grid 14:39 did the plane come first or the airport? So you have to build a use case and then the platform comes and if you build a platform that's when you 14:49

really really are you can scale and you can get your flywheel to spend. If you're looking at accelerate your data science Journey the idea is how 14:58 can you leverage to us? And so internally at Bank of America? We have a stack called The Phoenix stack and we have a series of best-in-class tools 15:08 that enable us to accelerate this journey and it's fundamentally based off of the Duke but there is a series of other tools that we use. 15:17 And here's another interesting angle. And and that's one of collaboration for those that are designers year. You'll be familiar with tools like eggs 15:29

or and sketch which were the standard to us a year ago. Another tool came out of nowhere call figma and fig my built collaboration right into the tool 15:36 to all the back-and-forth the designers go through all of that was built into the tool itself. So if you think about email and chat there 15:46 just papering over cracks in your process. So if you build collaboration into the tool set then you you're really allowing people to 15:55 connect right there. And then and that's another aspect the platform that were thinking about. And what's interesting is this would have been heresy a 16:05

few years ago very much like a bank moving to the cloud right but it's it's opening up a lot right? Brian Moynihan talked about cloud computing that 16:13 Bank of America's doing we have 70,000 servers. We're down to less than 10000. We've got a private Cloud. We've got a partnership with Microsoft 16:21 Azure, right? So things are changing a lot Goldman Sachs open up their proprietary trading platform. It's a page right out of silicon Valley's book 16:28 write two more and more. This is going to happen and how to weed in terms of the platform. We built. How are we going to open it up internally within 16:38

Bank of America and then maybe that's the call walk face to it. And I don't know if eventually where we go from there, but short the needles going to 16:45 move on that. And then another angle is people write your true north should always be your customer and 16:52 sometimes you know, it's difficult right where who, you know your customers. And in really really being focused on on on your customer and the 17:02 experience that you offer them. How do you capitalize the crowd? You know, we had a naming contest for a system. We were building within the first 17:10

day. We had like over 200 responses. It's crazy how people can get involved but you have to be able to leverage it and you can't do it on your own so 17:18 you might as well build a team around you. And the last thing is the power diversity, right? And this is you know, not only diversity and in terms of 17:27 the traditional dimensions of in a race and gender but it's also background right like you you want to have like a bank like ours we have people from 17:36 startups joining us and they bring a completely new way of thinking and business. We're in no people were bringing in have no domain expertise in that 17:44

business so completely fresh thinking and so that's the. The power of diversity. You think about research 17:53 year-over-year for all the techies here, right and all the maybe the men who don't have here in terms of the tough projects right there. Just 18:02 so many projects fail. So many projects are delayed and that's where you think about a job right where you basically bite off what you can chew. 18:11 Ship the product shipping is a feature to right and then you get the feedback and you iterate off of that. And that's the way we approach things a 18:22

Bank of America and then here's a provocative thought right? Agile is dead. With your technology teams, if you can cover this much 18:31 ground, if you're able to empower the business users. Can you cover? More more if you think about Solutions like 18:41 Tableau like Trifecta and you think about other solutions that are on their way like out systems like uncork these are low code no cold environments 18:50 where business users Savvy business users can actually Implement technology and you actually get rid of the technology business Contract game where 19:00

the business is complaining about technology Technologies complaining about business if techies just enable to platform the business controls your own 19:09 destiny if they're make a product and it sucks. It's on them if they make a product. That's great. That's on them, too. And so those Solutions are you 19:16 know, almost there and they're going to make sure over the next few years in terms of these no code platform and it's going to be a great way to 19:26 expand your unit technology scope. Put all this together. It's parallelism right innovate from the outside don't 19:33

innovate from the wouldn't otherwise the nervous system will eat you and then think about purpose having a massive transformative purpose think about 19:43 building a platform, but first build one use case and build two words that platform. And then think about people right the customer experience 19:52 diversity. And then lastly but we just talked about was enabling and empowering the actual business users. I 20:00 just one last thing. If you think about humans and how we think in a very sequential logical way, right 20:10

year-over-year experts were asked about their forecast of solar production in year after year experts basically answered 20:20 in a very very linear way. And it is fascinating that you can see this over 5 years 10 years. They answered and exactly the same way 20:30 linearly. But in actuality the production was happening exponentially. So you think about like all the disruptions that's happening around 20:40 us. Maybe it hasn't impacted. The bank says much for reason that I could spend a completely different session on, you know, when the speed happens it 20:50

will happen really fast to keep that in mind. And this is a simple example, right? This is a trading floor from one of our computer Banks up in 20:58 Stamford 2008 ten years later or actually less than 10 years later all gone. That's how fast things move right again saw through eating the world. 21:07 Thank you so much. Appreciate it. 21:16

Cackle comments for the website

Buy this talk

Access to the talk “AI Summit New York: Architecting Large Scale AI Adoption in Legacy Organizations”
Available
In cart
Free
Free
Free
Free
Free
Free

Video

Get access to all videos “The AI Summit New York 2019”
Available
In cart
Free
Free
Free
Free
Free
Free
Ticket

Buy this video

Video

Access to the talk “AI Summit New York: Architecting Large Scale AI Adoption in Legacy Organizations”
Available
In cart
Free
Free
Free
Free
Free
Free

Conference Cast

With ConferenceCast.tv, you get access to our library of the world's best conference talks.

Conference Cast
505 conferences
19653 speakers
7164 hours of content