Duration 28:34
16+
Play
Video

How to build a successful AI practice in Insurance and Government organization? By Hudson Mahboubi

Hudson Mahboubi
Sr. Manager Of Data Science at Workplace Safety & Insurance Board
  • Video
  • Table of contents
  • Video
Request Q&A
Video
How to build a successful AI practice in Insurance and Government organization? By Hudson Mahboubi
Available
In cart
Free
Free
Free
Free
Free
Free
Add to favorites
27
I like 0
I dislike 0
Available
In cart
Free
Free
Free
Free
Free
Free
  • Description
  • Transcript
  • Discussion

About speaker

Hudson Mahboubi
Sr. Manager Of Data Science at Workplace Safety & Insurance Board

Hudson is a leader in Digital Transformation and Insight Driven Organizations. He currently leads the center of excellence in AI and Machine Learning at WSIB. He obtained his PhD degree in Computer Engineering form University of Waterloo, and worked in the Big Data and Data Science domain for the past 7 years.

View the profile

About the talk

The effective deployment of AI in the enterprise will require building well-rounded teams that include people from a wide range of backgrounds and skill sets, including non-technical roles. In This session we discuss how to build a successful AI team in a Government entity or an Insurance company.

Share

Good afternoon, everybody. Thanks for having me here. It's a pleasure to talk to you about about our chat. What you're going to talk about today is how to do the taxes with a I practice in an organization environment. So I just talked with the offline. So what you going to rain here today in this presentation will be mainly around up who we are what we do in wsib. And what is your ETA? I and how everybody is using it and benefits the public sector which would be considerations of AI in government and public-sector how to build an AI strategy

and how we using and building that are strategy at wsip. So to start with who we are and what we do, our organization is named wsib or workplace safety insurance board of Ontario. What do BRB are the government entity does for employer to provide no-fault Collectibles Insurance to access industrial specific health and safety information. We hope workers for their loss of earning Healthcare expenditure due to the different injuries and accidents happen at them when the work time and work in the employer's gets

help from us to get back to work then such an injury happened. So we are Workers Compensation Board and we helped went Arians as a monopoly in the profits of entering Canada be getting back to work been such injuries happens the other various small team around 20 people here in wsib call Advanced analytics their focus. All is on more predictive modeling. Because science and general AI practices in order to just show you how big is our customer base and let you know how we do our work. I can say like, you know, there are around

320,000 businesses that are covered by our insurance Port we get something round 210 to 211000 planes a year. Average. We serve something around 4.8 million workers in Ontario and have a bunch of information on how much we pay for the health care how much you pay for the loss of earnings and patient and all others across the time. I've been mentioned earlier by government rules. The are the Monopoly in terms of providing data broker compensation in Provence. So pretty much we have a very widespread let you

know range of customers and clients across 16 different type of Industries. Are there bunch of numbers as you can see, how much do you pay? And what are the big drivers of cost or type of injuries we deal with on a day-to-day business? Getting to what is the yard and who is using it very different definitions of AI or machine learning girl in the world that can hold a lot of them are widely accepted. I think the one that we embrace your wsib and I believe on a my favorites quotes from Andrew end that defines machine learning as making sure the

computers can do work without the explicit to program an Innovative improve the quality of life and helping human do better work in that one as a government agency helping people getting back to work. We want to make sure everything's Knology or every consideration that you make here. Make sure that we are improving the quality of life of saiki know our citizens and I can do a human being To go over that's how a i is being used These Days Inn industrial. We look at the global pattern of AI as you can see and I think I might be

kind of like me, you understood. I don't even the truth is across different County than April different geographical locations in North America and Europe are leading in terms of quitting AI into production and using it as an official agents to do the jobs on a daily basis the other continents and places are speeding up and try to get on board and I cannot build their like a note strategies and products in the next 12 to 24 months. And I do overtime for life cycle of those data products. They moved from

prototypes into the products and the more and more time goes by Real Steel like in the more by display the fuse of AI in any industry including the public sector, so Daughter consideration is that can do about the forecast like, you know, but what's Cockney analytical like, you know back in 2019. What would be the adoption of AI in across different Industries in across the bow and as you can see like, you know, most of the companies and entities who are going to use that

you're focused more and guarding like robotic process Automation and more on conversational system like cat swats or they're going to use different types of systems are regarding like an image of pattern recognition or using the computer vision like a use cases of the hot topics like, you know, like driving and more sophisticated stuff are getting some game over the next five years. But this till I came to dominate use case for the puppet applications a I would be around Automation and conversational chat box. To make it to the public sector

a government entity can use AI or use any type of Technology because up like all unconscious biases and consideration regarding public sector. So here we want to focus. That's okay. How is the use and what are considerations to make so talking about the opportunities in the AI for the public sector, I think that opportunities for AI in public sector is not that different from the one that offer to any other industry, but it's my start with more smaller in baby steps switch the stars from more automation of the routine processes or repeating garlic is sitting up there because it's past

focusing at to improve the delivery of the service at the air is getting to the citizen and obviously just can be obtained by increasing the efficiency of operation the day-to-day business to make sure like, you know, something is going to read across all levels of governments like, you know from provincial to Federal to all this level available and I think the integration of the systems can improve tourism or digital government is they like in the quality of work and service delivered to the city But definitely it has its own

unique challenges and some of the challenges I think are everywhere but the emphasis on how to focus on which island is an obstacle to remove with the other day, but different from public sector to private sector. I thought we had a ton of skills. Obviously if I was going to present this a presentation to you guys, maybe two years ago. I will talk a lot about data but I think over time there is no he didn't think for any leader in any organization including like the public sector. How much value is Big Data and data has and I think

the quality of becomes more attainable and feasible in there like an organization, but it can we listen fascitis hot important it is like, you know To have the right data. I think the biggest gas and challenges for to be in most of the public sector will be on the area skill details kids right now are play battle against a I definitely the right Talent is in short supply and then usually organizations you try to use some incentives to attract high-caliber talents, like, you know, we'd better

compensation some flexibility and some perks that they offer to make their job more appealing about organization to be tighter rules on hiring and limited budget that government may suffer little bit to attracting the right Talent into there a I shop and that would be something that I think it's a stomach really exist as a challenging to move on from Movie from Dad's going to effective use of data like, you know, obviously organization we should have like, you know, if ability to

understand how to access and manage data and if they cannot understand how to manage and go over the data, they cannot take advantage of a I feel 4 degrees of Bilal like electrology volume and all those at The Death of Marat Concepts relates to the big size of data. I think it was an organization Boston private and public right now understand how the need for Duty that how to create an handled. It was what I think was missing piece for public sector organization is that you know, no understand.

What are they to ask if they have some rudimentary like an understanding of everyday tasks, but when they want to Publix to maintain any of these equations equal to how many data elements they have. Different databases or for the same kind of questions birthday can find answers. They are falling short and I think this is the main concern would be like, you know moving to what having a proper data governance considering the all the history of date of over years it over the Legacy systems having a property that governance is a Governor's job as you would be a big roadblock. And I think it's

something that needs to be considered as a big challenge moving forward to use of AI because I mentioned earlier if you don't know your data assets, I think getting to the AI is far as Ridge. The other challenges I think it's very unique for most of public-sector. The organization is is like an existing procurement mechanic says so the inability for governments to avoid Bender roadblock roadblock you back in a vendor vendor locking happens is like, you know, if you treat your outcome of your AI application is an IP and it was delivered by a vendor to you guys. I can know

next time that you need to recalibrate a reuse the saying retrieving the same model for solving different problem you bump into some like, you know legal issues Abby the IP and you cannot bring a new vendor. I'm bored and there are lots of the gestation like in the limit of a African how to bring like Anaconda party on board to help. We should I think something that can be work around and I can always come maneuver around with doing it due diligence and understand he requirements of a I might be different from the traditional. Contracts can be something that they can

almost the organization can overcome. Environment, I think it had its unique challenges are there was a reason to study done in by Deloitte in Canadian market and they study I think it was back in December 2019 and they found it right now in Canada. There are some things around 800 AI companies and off of those like, you know more than most of the Mack and Warren majority of them are less than 50 employees, very small companies and more than 50% of those days that exists even five years ago. So the number of the players in this

game of AI under people who helped you to provide AI environment is changing everyday is growing and for organizations like government to be on top of those players in the game and choosing the right at provider for the AI Market would be a very challenging please. It's not the same for 4w, go to cloud computing for the cloud computing, you know, the key player in the market like, you know, I'm pretty much like you know, when you get to AWS Google, I like it on Microsoft and Alibaba compare that you covered like eighty-five 86% of the offering so pretty much you have

limited number of the big vendors with longer footprint in the industry and more trusted leaders that let you know traditionally conservative culture of governments can rely on and Trust by Phillip yesterday. I become a little bit challenging which one is the right technology and which one and even knowing about all the players in the game would be very tricky and time-consuming. And a daughter, what would be the Legacy picture of this the Legacy culturalism think the first thing that comes to everybody's mind when he talked about

the government obviously everybody if you want to adopt and bring a new technology there will be challenges the challenges of change management and a bringing something to an population which are not early adopters by Nature got their culture is maybe late majority in the sense of accepting to change. This would be obvious the very challenging the private sector. There are like, you know, that's a bit strong cultures regarding this experiment encourage employees to enter a try new things and go beyond and like in the regulation, but I

get to the government the culture and legal legalities and the constraints around that difficult to elevate in a comparable manners to the A private sector, but it doesn't mean that it cannot start possible. Obviously. It has its own different challenges in Vegas ticket. And the last but not the least. I think it would be the AI ethics when it gets to the public sector ethics play much more sensitive role in how technology is replacing human and doing some work or assisting in those which I will discuss for AI takes in more

details in comics. So what are the ethical consideration of AI for the public entity or a government? I think you have to start with regulatory in the governance. Right? You have to say that. Okay, how did they allow their the development of a application for a public good to help all I can deliver the Mandate of the public nonprofit organization. And what's the moral status of using AI machines are robots that can go to make the decisions and who would be liable for his decision and all those implications. And so

unhealthy are using all the rules and transparency when we Implement AI algorithms to prevent any biases or any unwanted outcome. So that's one of the biggest consideration for any ethical AI development in public sector. The next would be the morality, right? So do we have a right that we stop using the same robot to be created yesterday. Can we destroy that? So are they belong to the organization or they become a part of the public bill who owns them who can decide Mindy finished your life and what if that

buy unwanted the the algorithm with its own belief or its own state of learning kit installed and trade some biases that you cannot even notice until very late. So do not like all those morality considerations. That should be considered when choosing an AI inside. The next would be the Safety and Security so One of the things that they get to safety security, he's a hobby make sure that the outcomes of like the product of a I directly or indirectly are not

harmful or not against the public good. How do we make sure that if something like something bad happens we're going out, cause some damage how we can recover and how can recover from that situation so irresponsible plan or like, you know rescue strategy even be get some security harm by unwanted design an outcome of the day I application. The next would be the legitimacy and Nanda position. So I think this is very important because then it gets the government everybody is asking his

decision made by this algorithm legitimate. Is the application is being used for the right purpose or unwanted the or even like, you know unconsciously you being triggering Some Like You No, Agenda. It's my party and I might be some unwanted and an unconscious bias. So how do we know that the training data is legitimate how we know that the quality control data is right how we know the data collected difference between apply for every decision. If any private-sector company makes my penis gets to the like in the public

sectors cost of fashion mistakes and starch pickles with a huge burden on citizens and taxpayers and under like a party song party. Like, you know, the political party is running the government making a huge impact and I think they stopped as it's okay if I'm going to do it you have to do is try if there is any risk for us who can be blamed for and then If there is somebody to blame for usually there will be resistance even going February. So the other piece I would be that like, you know regarding like in Social economy RV

causing loss of jobs. Are we causing like, you know, some people losing their work and like in the moving to Mississippi cuz like, you know, we are replacing humans with the robots. Obviously, this is especially when you deny environment the most of the staff getting you tonight. I have some protection and there will be some change management and 19 Good ethical practice needs to be put in place before going framework for public sector leaders to overcome that it's simple a cure to all privacy and ethical use packages in

public sex are they are expensive devil will to make sure they're all the rightful citizen to be fully observed and do not cross any lines and watch all those red tips. The other piece would be the mixture like, you know, you have policies and guidelines for the little bit of a eye and make sure those are aligned with existing policies and guidelines and if there are parts that like, you know, they're not clear make sure you bring some transparency to the next with me that make sure any database that you use for your training does not have any stomach virus and make sure they

understand it might be there is unconscious and you have to go there and bring it to a different techniques to remove those and make sure like in a you are embracing diversity and inclusion and every step of the way. Make sure let you know I have another look look for all the things that you put in your algorithm decide to date a lot of time unconsciously like, you know are our greatest big building like why is it so hard to detect it's so might be one on one thing but I think continuous monitoring and evaluation could call to make sure if such thing is happening go and fix it and

bring transparency. I think you should have built your guest and consideration is really get to the government everybody. Look at the Black Box the more transparent you are with how you're using your technology building a greeting people. Will try you have picture by So let me go to the next one will be okay. Now we know how to do it. So how to build an AI strategy and I think a strategy I think for any strategy going back to Michael Porter's famous article in Harvard Business Review. It's like, you know what you are not choosing to do. I think I think it gives us a

framework and Define austere let you know are the red lines on decorative stand where we should not cross and when he is not working inside those boundaries we can find like, you know based on our goes like something to achieve and like how to make and use up like, you know due to started you for a good I need government entity in Canada VR. We are having some alignment. There are existing a strategy for organization. Like I said, you have to align and like, you know Candice Charger price for Action, which is talking about where we are in terms of use of data understanding

robotics. Then we are moving in the government to bars use of more autonomous systems and artificial intelligence and making sure that tomorrow's my day off like, you know and iot smart connected systems and Societies in cities. So I think it was interesting for any public entity in Canada right now. He's that's in the charter talking about they can no expectation of the AI Global GDP of 14% - 2036 music and play get big role in the GDP. And if it becomes the part of like the public private sector public sector should be with

some Liza, please continue that cart and go towards that. So, how's it going to be a strategy Avicii? Every Saturday have fillers how we do it. Is this Sunday division? Okay. I know we can go out very far, but how far really want to achieve what is our level of ambition and then we know what level of ambition in lingerie. I to go to the mission and we're saying okay, where should I focus? I need somebody important to identify which problem if you want to talk about. How do you want to use a I and I think what it brings as a side effect is your plan of

what you want to tackle shapes your requirements for technology. So if you are going to make very ambitious that you might have a different technology footprint and requirements compared to something very basic and simple and next from that would be going from the vision and liking the focus to the value proposition Soda Springs seemed like, you know, obviously but I can do good is using big chunks of data and produced inside quickly safely and do we want to Who dis to reporting a descriptive do we want to go to the predictive model? Do we want to trade products the prescriptive and get

out lines or do we want to do RPA and how we want to do it and in each of those how deep you want to go in our areas of a business and obviously for every starting to you need to have evaluation framework creating a message to make sure you are measuring Roi and also your measuring all existing biases and accuracy. So it went to get the public figure out there will be more and on accuracy and bias on by 4 Non profit, but we have to make sure it has become reasonable and long-term you have some savings on cars while you are working multiple removing you buy acetone have

accuracy. Make it to next part of the strategy you have to identify that okay, what are my skills and capabilities needed and what's my current state? What can I do? Can I recruit more full-time member members can I identify my talent and send them for apis killing? I like him teaching them in Utah skills, or I should go to Outsourcing and partnership. And the last piece is how to manage the change it change from traditional brick-and-mortar kind of Officer digital office moving from using a decision based on maybe bunch of the spreadsheet into an output of a recommender system Disney change

and change needs leadership. And I think this is very communicative property communication with all the stakeholders create think of Center of Excellence has and have some transparency with the reason for is scalability future is much needed. I'm talking about this Saturday. Let's save how we use it in a wsib. Obviously. We have a mandate envisioned our vision is he want to be inside thriven, which means we want to use insights from data for every decision. We are making an everyday in this organization rebuild some of Foundations.

We have processes. We have identified capabilities betrayed her a strategy from Gas Monkey identified the need for data governance of strategy Enterprise Knowledge Management, as some standards that was built from those we built Enterprise strategy for analytics and now we are ready to let you know our strategy into action and obviously I start our journey to become an ID or just inside driven organization. How to use it reuse it in terms of natural language processing for chalkboard text mining using featuring jean-yves for different predictive model. We

have process mining in optimization for optimal ultimate the triage process for civil cases be used extensively machine learning for different predictive models non-compliance anomaly detection. We use automated process rt8 for the process that you have repeated nature and can be done and obviously for Maybelline lots of to be used OCR a lot to convert all none like, you know machine readable format into something that can be consumed by model and how we work simple business opportunity we go and see what's the problem. Are we trying to stall the

ivy look at the date of the sea do we have enough? If you have a quality of a date then if we pass this point the next step would be that do I need to report do I need a model or do I look into the eye of the keishin and when if the answer is a model you build them? How do you train the model and send it for feedback reiteration the calibration and it becomes something that can go to production to go and deployed and make it the process and obviously as you can see like, you know, maybe 50% of things that stay on the porch in the next 50 get to the modeling and maybe less than 10% makes

it to the action deployment because this needs to take like, you know food investment and I cannot lots of change management from Ocean Park. So we believe in any AI application needs data computation power or infrastructure and his skills. And this is how we try to build a team B make sure you have lots of Dayton organization BR procuring let you know we had no infrastructure for competition in procuring more and more to scale up and we are working to build a team of talent and the team's a honest broker wsib. I think I would be that start with the business process to

tell us can we solve this problem with the date do the cheapest quality of a date and make sure the data can be used to have a good quality and you have a bunch of data scientist and it's a decision with the job of creating model fine-tuning Tomatoes converting the hypothesis and do some investigative analysis. And at the end we have a tire machine learning engineers and air signs 2wd for those complete my other than meat deep learning app. Create an application for the production. And obviously we have a leadership around at the job would be

there to make sure all these bits and pieces work towards the Mission Vision and mission of the organization should be coming out by the old by delivering on results and celebrating like, you know small bit. And I think I hear is the my last night. So I will be happy to answer any questions.

Cackle comments for the website

Buy this talk

Access to the talk “How to build a successful AI practice in Insurance and Government organization? By Hudson Mahboubi”
Available
In cart
Free
Free
Free
Free
Free
Free

Ticket

Get access to all videos “Global Artificial Intelligence Virtual Conference”
Available
In cart
Free
Free
Free
Free
Free
Free
Ticket

Similar talks

Sudip Chakrabarti
Venture Investor at Madrona
+ 4 speakers
Ian Foley
Venture Partner at Level Ventures
+ 4 speakers
Anupam Rastogi
Partner at Emergent Ventures
+ 4 speakers
Vik Sasi
Partner at Dreamers VC
+ 4 speakers
Dennis Liu
Investor at Ford
+ 4 speakers
Available
In cart
Free
Free
Free
Free
Free
Free

Buy this video

Video

Access to the talk “How to build a successful AI practice in Insurance and Government organization? By Hudson Mahboubi”
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
566 conferences
22974 speakers
8597 hours of content