This webinar, hosted by Peel Newcomer Strategy Group with support of Digital Transformation committee of the Executive Council of Peel-Halton Settlement Partnerships, is designed to share knowledge and discuss the potential impact of Artificial Intelligence. This is the first session in a 4-part webinar series exploring the transformative potential of AI in the Newcomer-serving sector, with a special focus on how AI can enhance our services and operations.
Artificial Intelligence (AI) is transforming industries worldwide, and the newcomer-serving sector is no exception. It can unlock new opportunities for organizations and staff to delivery services more efficiently. However, the role that AI could potentially play in the newcomer-serving sector is still largely unknown. At the same time, the potential dangers of using AI, including privacy concerns and security risks, cannot be overlooked.
Speakers:
About Darcy's presentation:
With funding from IRCC earmarked to investigate ways in which AI could be leveraged to support settlement services, SEC embarked on a journey to create an AI resource to address bottlenecks and ensure quality information in their foreign credential recognition services. Darcy will share about the journey from conception to where they are today with a ready-to-implement service, focusing on learnings, challenges,and future direction. The second part of his presentation will focus on innovative ways SEC is learning to utilize AI to strengthen their internal functions.
Marco's presentation slides:
Darcy's presentation slides:
Machine-Generated Transcript
What follows is an AI-generated transcript of our conversation using Otter.ai. It may contain errors and odd sentence breaks and is not a substitute for listening to the audio.
Marco Campana 0:00
So before I set the context, I just want to do a quick land acknowledgement. So I live. I'm going to talk about myself specifically, and I encourage all of you, obviously, to be aware of where you are. I live and work and raise my children in the territory of many nations, including the Haudenosaunee, the Wendat and the Mississaugas of the credit. The land that I'm on is also part of the dish with one spoon territory, which is a treaty between the Haudenosaunee Confederacy, the Anishinabek and allied nations to share and care for this land, its waters and all of the biodiversity. Most land acknowledgements kind of end there. And what I want to try to do is suggest to people that there's more, there's deeper we can go as a sector. And the Truth and Reconciliation Commission of Canada actually provided two specific calls to action related to newcomers, so related to our sector. Specifically, the first was adding language to the Oath of Citizenship, acknowledging aboriginal and treaty rights of First Nations Inuit and Metis, Metis people, and that has been done. The second is to revise the information kit that newcomers get and the citizenship test to reflect a more inclusive history of the diverse Aboriginal Peoples of Canada, including information about treaties and history of residential schools. And that is still a work in progress by the government, but the sector has actually been active on this. So I'm going to throw a link into the chat where you can find useful materials if you're not already using some of these to within your own work and with your work with newcomers to understand the kinds of conversations you can have internally, as well as with newcomers in in your organizations. So before I dive in to to introduce Darcy, I am going to spend a few minutes sort of setting the stage. And there's a lot of us here today. So I've created a slido Q and A where you can post your questions, because if we try to do it in the to do in the chat, it's just going to kind of get completely out of control. So what I've what I've decided to do is create a little slido, and I'll share my screen now actually
Marco Campana 1:54
see here we go, and I'll start the presentation. So that link should take you to to the to to slido directly. Hopefully you're, you're seeing the right screen, yes, but you can also, if you have your smartphone, you can use the QR code that's in the bottom left corner of this slide to get to slido. So if you have any trouble in getting there during the the presentation, just let me know. And we can, we can definitely figure that out. I'll also add, in case people prefer to go to the slido website itself. So you can go to slido.com and enter this code. So what are we talking about today? Today is session one of a of a of a four part session looking at using AI in our service delivery. And you will get all of these slides from from pnsg after. But I want to just give you a sense. You haven't registered for these. Today, we're talking about AI to improve the efficiency of services, and Darcy McCallum is going to be our main speaker. On August 8, we're going to have Dan kenshaw from food, sorry, the furniture bank, who's going to talk about their journey learning about AI to using it in their marketing and other in other parts and using it intentionally. On the fifth, we'll be looking at the role and implications of generative AI and large language models and supporting newcomers. So this is actually some research that's that's being done at the National Research Council in Canada, and that will be really interesting. And then on the on September 12, the fourth is a local experiences of using AI, where you'll hear from a couple of additional colleagues that achieve an access Alliance about how they're implementing AI in their work. So I encourage you to sign up for all of them. All of these will be recorded. Everything will be shared, which means, oh, yeah, we are recording. Okay, perfect. Just to make sure, so you'll be able to access this information at any time and share it with your colleagues. But I do encourage you to visit the entire series. So what's the reality of our situation? Like with many and this is not just our sector, this is the nonprofit sector. This is every sector. The reality is you're already using you are AI in your work. And if you're a manager and trying to figure this out, your staff are using AI in some way in their work. And if you're frontline, you know, you're using AI in some way. You're playing with it. You're using it in many ways. What we need to do, and what we need to talk about is how we make that thoughtful and incremental, so it's not exponential. We need to implement this technology in a meaningful way, and that means creating acceptable use policies, even if they're draft policies, figuring out how we can have we can protect privacy, we can ensure work quality. We can disclose when we're using AI in our work with internally with clients as well, and we need to figure out how we can share it within organizations and across organizations, and provide training about how to be using AI in our work, and generally, when we talk about AI, we hear generative AI, right? So chatgpt. But the scale and the breadth of AI is much wider than that. So if you're using Microsoft 365 you're hearing about copilot, you're probably starting to implement it in there. If you're using your phone, you're seeing auto, auto correct. And there are pieces of AI in all of our work. So we need to also make sure we understand AI in a much broader context in our work, and also understand the advantages and the disadvantages to that. So I've highlighted the words that I think are important here, that the advantages are, they can they can make your work more interesting. They can automate things that you do regularly, free you up to do the kind of work that you really want to be doing in your in your organization. It can help you provide personalized help for you, so you have a personal assistant if no one's available to bounce ideas off of increasingly machine language translation, although with within reason, and also being able to use it for applications, for grant applications, and ultimately to start enhancing the client experience. And you'll hear from from Amy and her presentation in September. But one of the things that they've access has figured out after implementing their virtual Employment Resource assistant for a few years, is that they actually saved 7400 staff hours per year. That's the equivalent of four full time positions, if you break it down, that's a lot of sort of automated scut time. There were people aren't doing repetitive tasks over and over again. I mean, 74 hours reallocated to actually serving their clients. That's a major advantage, if you can implement this, implement it well, but the disadvantage, or the considerations, are that this requires significant investment, so access was able to do that with a million dollar investment, as well as time from Accenture, for example, and their technical expertise. But there's also concerns about why, when I talk about incremental that we don't move quickly into this because of the bias in AI. We all know about the quote, unquote hallucinations, and what people talk about, they're not hallucinations, it's misinformation. It means that the AI wasn't programmed properly. So we need to start, we need to stop humanizing AI and look at it from the perspective that this was actually poorly programmed. So if there's bias built in, which there always will be.
Marco Campana 7:03
We need to be aware of the of how the tools we're using might have that bias, and what bias we might build into the tools that we build. We need to understand privacy so that if you're sharing information about clients, to try to try to get information back, what are you sharing into that AI, and how will it use it? Does it use it to train its AI is that information now sitting somewhere and and I know when I say that out loud to you think, of course, we wouldn't do that, but copying and pasting is really easy sometimes, and we forget about extracting things from the pieces that we're using. It also has the potential to replace human interaction, right? And that's the fear, that's the anxiety. And some people I follow talk about the AI sandwich, which is human first, then AI then human last, which means AI is just a tool in the middle to help you, but ultimately, your policies, your ethical considerations, and your wisdom and knowledge are what the final decisions are based on, not taking decisions based on AI, but going through human filters. So AI has to be implemented strategically. You'll hear this a lot in both what I talk what I talk about and what Darcy talks about, complementing our expertise, enhancing, but not replacing, and making sure that we continue to have that personal touch right and so we also are always evaluating, always figuring out how AI, as it as it involves, aligns with the work that we're doing in our Our values. So where can AI help? It can help you leverage huge pots of data. We all collect tremendous amount of information about clients on a daily basis. We don't do a ton of that other than put it up to eye care and let the government kind of play with it. But AI can help us to collect and analyze those metrics and provide us with with information to better understand the people you're serving, which in turn, allows you to then start to personalize some of your services, offering new pathways. Once you understand your clients fully, you can, you can start to cater to them in more meaningful ways. And AI can help you to do that as well or automation or other pieces of technology. And the idea that that while we are using virtual service delivery, while we are using AI and using other technology tools. We're not losing the in person and the human connection, right? We're trying to create this seamless offering of opportunities and choices for clients so that they can access this information, they can access our services and programs in ways that are meaningful for them. And again, AI can help do that if you're personalizing experiences for clients who want to access using technology, it can free you up to provide more of those in person services, or those more specific services for people. Remember that 7400 hours again and again right now, we know that's that people are using technology, using AI for in client facing ways. They're leveraging technology through digital messaging, through chat bots, through online appointment scheduling, virtual event platforms, etc, behind the scenes and a lot of what Darcy is going to tell you about today is how AI can help as an internal tool and to create some internal efficiencies to to support the work that staff do in. Terms in in order to then provide better service to their clients, basically. And the reality is, is that this, this is an acceleration since the pandemic in particular, and certainly since November 2023 when chat GPT kind of debuted, and the tech world kind of lost its mind. But the reality is, is that it's only one of many technologies that you need to be juggling and considering when you're starting to look at how you offer technology in your service delivery. So yes, AI is something that should be on your radar, but it should be in conjunction with many other technologies also on your radar, so that you also are figuring out how to use those ethically and effectively in your service delivery. So what can that mean? It can mean, once we understand people better, or we can use the tools, we can create more accessible online experiences, whether it's closed captions being generated automatically multiple languages. I mean, you see the videos of the same person speaking, but it's in a different language that actually comes out of their mouth. AI is doing that, for example, how well is it doing it? We're not sure, but it is doing it, and that enhances the ability for you to provide information to different audiences, even if you don't speak their languages. But what does it mean once they come in and interact with you one on one? So the ethics of some of doing some of those things in the same way that, if you're helping clients to generate AI driven cover letters or resumes, can they actually talk in the in the words that are, that are in that that resume or that cover letter, can they support the language level that they're showing in in text when they actually interact with an employer or someone in the community? And what are the ethics around that? Integrating services, managing information, having more responsive chats your you can be around so Vera with with access employment, is a 24/7 service mechanism, right? It is never, it never goes to sleep. It can provide information to newcomers when you're sleeping. So you can provide services even when you're asleep. The question is, what level of service do you want to provide, and what can and what makes sense to provide in those time zones? But it, but it offers you, it offers your clients more access, and it offers something that can be personalized, because once you connect your your technology, your AI, with your client management systems, it can pull information out and say, Oh, I see, once they've logged in, I see that you're an engineer. You're an electric engineer. Would you like information about X, Y and Z? So it's not just giving generic information, right? And in the same way, it can be a support to you. So I've heard of an organization that is using AI, for example, in the background so that if, as they're entering a needs assessment or intake for their client, if the if the system notices that that person is paying more, a more a higher percentage of their income for rent, for example, the AI will spit out it looks like they're paying, they're paying, they're paying 50% of their income for rent. Would you like to have a conversation? Have a conversation with them about alternative housing options? So it's an assistant for that person who may not notice that that percentage may not notice that math, for example, so it allows the staff to then have that conversation with their clients. So ultimately, what we're moving towards is what some of us have called omni channel service delivery. You'll hear more about that from Achieve, for example, because they're moving in that way. But the idea is a No Wrong Door approach, right? It's not a new idea, but with technology, it can become it can scale in a different way. So we have more interactive content. Ai allows us to create information and content more easily from existing pieces. So we can take this video eventually and turn it into a transcript, which then turns into multiple blog posts, which then turns into social media messages. And all of that can be driven through through AI really quickly, again, monitored and mediated by humans. But the the possibility for creating engaging content from existing pieces that we have is is easier, in the same way with outreach and data driven insights. The more you understand about the people you're serving, the better you can reach them and tailor the messages for those people so that it's meaningful to them in the information overload that they exist in, and that information overload is real. So where can we where can we interact with clients, all along the way. For most of us, it's after arrival in Canada, right? It's even sometimes after the Permanent Resident Card. But what role could aI have from the beginning of a newcomer seeking information during the application process? At all these touch points, there's an opportunity for information and orientation, for AI and other technologies to play a role, to support people. It's a it's a huge overview of it's a hugely difficult and anxiety ridden process for a newcomer. So where are the right places where technology can help us check in and and in terms of post arrival, one of the things we know is that there's a massive amount of information overload. So what can technology do to help them navigate all of these different systems? Right if they if they never step foot in your door? Are there ways for us to get them this kind of information? And the reality is, we're doing some of that now. So for example, orientation to Ontario is a program run out of costly where newcomers can go physically to 35 locations across the province to get the same information right. So the baseline information about all of these challenges. Is, is the same. They've got, they've got multilingual workbooks and webinars and sessions. So a person could go in person, they could attend a webinar. They could download an introduction for settlement. They could interact with with a 24/7 chat bot. They could download a smartphone app. So increasingly, we're looking at this as that kind of omni channel, no wrong door system where a newcomer can get in person supports and they can get that same information online in multiple ways. So whether it's a live webinar, a webinar recording, a downloadable PDF, the chat bot itself, or interacting via by an app, for example, so where AI can fit into that. And what we need to look at is we're starting to look at how we evolve as a digital settlement agency with the newcomer in the middle. And I won't spend a lot of time here. There's, I've got deeper presentations on this, but the idea is that we need to create baselines in our service delivery. We need to create baselines in our competencies, in our professional development and how we share knowledge. We need to understand what are the frameworks for data, in the case of AI, that will help us to create good data, to then create good AI without a ton of bias that has good information that comes out of it. We need to make sure that everything we do is aligned with our sector values of inclusion, equity, anti oppression, for example, and we need to make sure that the newcomers are at the center of that and that we're advocating as the sector for the right instruments and investments from our funders. Again, Vera is is a $1 million investment right from from the private sector to help get that up and running. The impact and the outcomes are fairly positive. But what, what does that mean for the rest of our organizations? When we're looking at how we we figure out how to invest in this and what's the role that our funders can play. So when you're looking at a really simple roadmap to get there, right, the first is to develop the vision of what you want to do. So after you start understanding what AI can do, how can it transform the way you do your work? How can it either be an internal tool, a client facing tool, or a wider community tool? Then you need to figure out a roadmap in terms of getting there and figure out who you're going to work with. So as Darcy will explain to you, technology partners are centered. Central to this. We need to be working with the developers to help figure this out with us, but we also need to develop with our subject matter experts, which is you, the staff, the people in the organization, and then with the communities to figure out what's the best way for us to meet these needs and to move forward. So there is a roadmap that that is that again, we should be doing this with any technology and any needs assessment and any needs in the community. So don't forget that you already know how to do some of this stuff. We just need to apply it now to technology, and we need to apply it with our sector value. So I like to remind people that 24 years ago, the sector came together and created a set of core values that are completely applicable to technology today, even though, when you look at again, the link is here and you'll be able to click through it when you get this PDF, when you click through and you look at this, none technology isn't mentioned at all, because it's 24 years ago. But when we talk about access, we need to make our services access, accessible, technology wise, but we also need to figure out how to create access to technology for newcomers, which is part of inclusion. When we look at collaboration, we need to be collaborative within our organizations. We need to be collaborative in the sector, and again, with these outside partners, like digital digital industry developers, for example, we need to be we need to respect the individual, and how better to respect them than to understand them fully and to provide them with information that is actually meaningful to their specific situation, and we need to be reliable. So if we're using technology tools like AI, how do we ensure that what we're what we're putting out is, is unbiased, is accurate, is up to date, and and and is relevant for the people that we're putting it out there. So one of the things I like to say is that we need to know backwards a little bit in order to move forwards. Let's remind ourselves that the values that our sector has been created on are actually completely relevant to creating ethical and responsible use of technology as well, including AI, and that we need to build this in to that roadmap that I mentioned when you're looking at your vision and your roadmap, so be strategic, and let's not forget about the ethics. Let's not forget about our core values. So moving forward, can we empower communities with accessible, personalized services? The answer is yes. Can we strengthen our impact and have more data driven service and outreach? The answer is absolutely yes, and ensure that we support newcomers in their settlement journey, putting them their needs at the center. The answer to all of these questions is yes, but it requires intentional strategy to do that, and some and some really strong approaches moving forward. So I'm going to stop and hand it over to Darcy, as I mentioned. Put it. I'll put it in the the chat again. You can ask your questions at slido for folks who are who are new. We're using slido to to to help with our with our Q and A because there's a ton of you here today, but I want to hand it over to Darcy, and I'm not going to read Darcy's Darcy's bio, because Darcy and I go way back. Right? So Darcy and I go back to when he was the coordinator for Swiss settlement workers in schools, many, many moons ago. He and I are veterans in this sector, and what I will and He is currently. Is it the CEO or the ED? He's the head of social enterprise Canada, which helps does many things, including helps run the Welcome Center in York Region for newcomers. But what you need to understand about Darcy is that he loves this sector. He understands this sector deeply. He is incredibly committed to this sector, and he wants our work and all of our work to be relevant and center newcomers in it. Darcy is also like me, not a technologist, although he is tech literate. So when he talks about the work that he's doing, it comes from the perspective of, how can I make this technology work for my staff. How can we make it work for the work we're doing? How can we make it how can we ensure that it's working for our organization, but also ultimately, work to help make that settlement journey faster, easier, more inclusive for newcomers. So you can read, you can go to you should connect with Darcy on LinkedIn. You should read about his all of his background. He's got more M's in front of his name or after his name than I do, and that's great. But what he is is someone who has been working on this and approaching this technology in a practical, meaningful, ethical, responsible way, and that's what he's going to talk to you about today. So I'll hand it off to you. Darcy, thanks,
Darcy MacCallum 21:16
Marco, that's very, very generous of you. Thank you so much. I'm just going to get my PowerPoint slide up and going, and as we said, we haven't done the tech run through, so we'll see how this goes. Really nice to be with you all. Just, you know, greetings from Newmarket. That's where I'm at right now. Although I am actually a resident of Mississauga, my claim to fame within the settlement sector is I've worked in Peel Region. I've worked in Durham Region. I've worked in Toronto. And now the joy of working in York and I live next door to halt, we do our shopping at Canadian Tire, my wife's favorite store in Oakville.
Darcy MacCallum 21:56
Yeah, just a little I like to be a little bit personal. So just so you know, father of three amazing daughters, Married for 34 years to my wonderful wife. My eldest is an incredible reader and blogger. My youngest is heading to Japan today, and my middle daughter is getting her doctorate in clinical psychology next week. So we're heading to La we're pretty excited about that. All right, I want to preface, as Marco's already said, It
Darcy MacCallum 22:23
neither of us really are what you would call like the you know, the tech gurus we are. We're novices. And I just want to share with you our learning journey and invite you to engage with with us, both in our struggles as well as our successes, because you'll see their struggles along the way. Just a little bit about sec, you know, we all want to know who they are. You know, what's interesting since becoming CEO here about 20 months ago, our focus has really been on how we do work, not necessarily what we do. So those values at the end, about practicing kindness, so that we we treat everybody in staff and anyone that we engage with a deep sense of their humanity. That's our first priority. We find solutions together so nobody's sitting in a corner dreaming up their own ideas. We look for ways to do better, which is why we're playing this whole AI game, and we build trust. Seek to do things that actually, foster and build trust. If we're doing those four things, then whatever we do makes sense, if you will. Here's the list of services I know. That's how most of us understand social service agencies. As Marcos said, we run the Welcome Center in Newmarket, and we have a lot of family and childcare services. We have a little bit of service in Peel region. So, yeah, I'm still connected. And we have an organized, a business side of our organization that does translation into interpretation and communication centers. So I want to talk about two things today. We've engaged AI, two main ways, one is by creating a large language model system to support our foreign credential recognition services. We call it the smart path, AI. And then the second thing that we are doing and playing with is leveraging chat, GPT, their bots to streamline access to vital information within our organization, and so I'll explain both of them to you, and I got one of our directors to actually put a little video together so he can demo the Bots for you. It's pretty cool. All right, so we started. How did this get started? You know, it's already been noted that AI is incredibly expensive, and I walk into this organization as the new CEO. I overlap with the previous one. I've known her for many years, and she said, Hey, I got a gift for you. IRCC has agreed to invest in us doing a proof of concept on how AI can be used in settlement. And I'm like, holy cow, I'm not ready for this. I don't know the first thing about AI. And. Um, and as I stepped in, I became clear that the project was something that she had, but she's retiring, and we had to really, you know, rev it up within our organization. And our uncertainty centered on three questions, you know, where do we focus the investment, where, you know, what is our expertise and and what can AI actually accomplish? Now, to be honest, I think we should have started with the third question, but we didn't really get to that till we were quite a ways in. In fact, the second question we didn't even quite figure out right from the start, but let me just sort of walk through them with you. So our first question, where to focus the investment? My predecessor thought, well, let's invest it in our communication center, you know that place where we do interpretation, translation and communications, so we coordinate appointments for the seven welcome centers in York and Durham we do in live interpretations for any services in those sites. And she thought that we could leverage AI there to support our our staff who are engaging clients to, you know, make sure that they're, you know, providing up to date information that there. Maybe we can reduce training time, because there's so much information and settlement that you have to learn and, you know, maybe provide some initial answers to clients. But, you know, as we thought about it, we began to realize that if we were going to try to use AI in this way, it would require us to work quite closely with all of our partners. And it's not that we don't want to collaborate. We love collaboration, but we came to find out that that might be a bit tricky with with a first step into AI. You know, there's multiple stakeholders create multiple and very complex processes for moving into a proof of concept. And then, you know, settlement, to be honest. I mean, it's a really broad sector. Any of you who are settlement workers here, you know what I'm talking about, because you're covering so many areas of need that people bring up. And then there's many others that are already working in this area, maybe not with AI, or maybe you're dabbling in it. You got 211, you got settlement.org you got peace geeks, you got a whole bunch of others. And our expertise is not, if you will, in broad settlement, but it's actually in foreign credential recognition. So we said, that's our expertise. That's that's where we need to focus, AI, it's our largest team. We have over 20 years of experience with this, and we have a dedicated data set that we've developed, action action plan templates for 162
Darcy MacCallum 27:31
occupation codes. So it's it's really our resource rich piece, and the need is there. I mean, we understand your clients report how incredibly challenging it is to navigate complex layers of information. And our staff talk about the challenges that they have trying to, you know, pull these action plans together and then stay on top of the changes that are going on in the services that are available. So, you know, this is where we realized we needed to focus in, you know, for the record, though, just complete honesty here, we didn't quite have this figured out till after we had already secured our vendor to help us with the AI. We're really fortunate that they flexed with us, and I'll talk about them in a few minutes. They're They're fantastic, but they walked with us through a learning journey. And believe me, it was a journey, and that journey really focuses, especially on this question, what can ai do? And we had to wrestle through it, you know, what? What can we automate? What should we automate? And, you know, just want to say straight up, from the very beginning, we included our HR manager in this conversation because we didn't want to just be talking about technology. We wanted to talk about how technology, and AI in particular, would impact our staff and how they work, and what their experience is like working here at sec. Then we had to start wrestling with what's required to, you know, for that automation. Do we have the right data? And we got lots of data, but is it ready for this? And then the big question really was, how does generative AI even work? You know, we didn't even know about this concept of scraping. I don't know if you're aware of what scraping is. I don't even know if I can define it, but it's basically when an AI goes out there and it, it, it goes through all the information on on a particular site, and it processes that. And I straight up, I think it was about 48 hours before we put our RFP out that we knew we learned about scraping and plugged that in. It was wow, you know, talk about building the plane while you're flying. We were in way over our heads, and some of our staff was, I think, enamored with chat GPT, and so we had to dig deeper. So one of the key resources we found was this one by data. IQ was a white paper called defining a successful AI project. I want to walk through you, with you, through how we process the questions that they give. And like I said, the link is here on this on this slide. Right? The slides are going to be provided for you, so you can take a look at this white paper. We found it invaluable in really focusing us. So first the question is, who's going to benefit from the project? Now again, we, we thought about creating a client focused, you know, piece, but we we recognized as we went along that what we really need to be doing is supporting our staff, not replacing our staff. That was really important for us, that that, that what we want to create as a tool for them and maintain human engagement with people as they're coming into Canada and processing their their credentials. So you know how, how would we envision AI helping our staff, and how did they envision it? Some of them were already using chat, GPT, and so we needed to talk and see, well, where are you using it? But what ultimately pushed us forward was thinking about the time where the where those bottlenecks, where was the time that they were spending most? And it was preparing meetings for clients, getting these action plans together and then confirming the information, especially if a client was asking about a particular occupation code that wasn't as frequent of focus as others. So who's going to benefit? Number one, it was our staff, and through our staff, the clients. Okay? Number two, how would AI help us do something we cannot do or do well today. So for this, we looked a lot of what's those repetitive and predictable activities that we do? Where can we save time? And where can we help them do things that they cannot really do on their own because of capacity? Again, it's not replacing staff. I'll say that a few times, probably through this presentation. It's not a client facing AI, it's a it's it isn't a way, but it's really built for our staff to support the clients. And so the focus was improving the experience of our staff and our clients by automating the mundane pieces and pieces that our staff just couldn't keep up with because there's too much information out there. So third, why is using AI better? Again, this really digs into the consistency of knowledge. You know, we, you know, we have a staff of about 20, and those who've been doing foreign credential recognition, and they're, they're like a fine tuned engine right there, that Mercedes flying down the 401, or, I guess that's a parking lot, maybe some other highway, but they're, they're really moving well, and you bring someone on, and it's really a lot to learn. And so how could AI help us with onboarding? Again, reduce repetitive processes, and then staff can focus on most what they do best, which is client engagement, answering questions, helping bring you know clarity where there may not be some so what's the upside? That's question four. What's the upside and what's the downside, or the consequences of using AI or not using AI? In one sense, you know, one of the things we realized is that many of the clients, especially internationally trained professionals, will already be using AI. They may be going into chat, GPT or to other systems. And we know that's a fact that AI fabricates information. So if it doesn't have an answer, it can create it on its own, you know? And so we, we want to make sure that they're getting accurate information. So the other thing is that we, we also realized, you know, we're playing with a small sum of money we don't have, you know, the gazillion dollars that AI developers have. In fact, we only had about $200,000 to play with. It wasn't that much. And so it was really taking the pressure off our team and saying, let's, let's do a proof of concept. If it succeeds, we can demonstrate the potential for generative AI across the settlement sector, and maybe it can be scaled so that other orgs can benefit from it. If it doesn't work, at least we can share with people the potential and the limitations of AI. And so we we had to sort of look at it that way, and I think that helped us to be able to move forward in this project. Fifth, where does the data come from? You know, looking back, one of the most sobering moments for us was having an external party look at our data. Yeah, let's just say our data is not as consistently, you know, produced as we thought it was. Our staff don't all work in exactly the same way, and so we have to find ways to create some con, some consistency. We can have seat we are trying to, like, whoever's microphone is on, if you could turn that off. And then the other thing again, is the data was coming from dozens and dozens of websites where that we have on all of our templates. So. So that part we don't control, but what's, and I'll explain a bit how that actually works for us. This is a huge, huge benefit for us, but that's, that's the second source of data. One of the key findings we had was that, and I hinted at this earlier, we needed to minimize external dependencies for a first AI project. In other words, we need to keep it simple for AI to work the data source needs to be consistent and clean. So we had things we needed to clean up on our end, and especially cleaning our processes, as I've already mentioned. But if you're thinking of dabbling in AI, keep it really simple at first. You might have lavish dreams of where this could go, but if you stay focused and stay simple, you'll be able to create a proof of concept that you can build on, and that's what we've been able to do. So let me talk a little bit about our vendor. We did a public RFP, and we had think about a half dozen different vendors apply, and we ended up selecting an organization called LeMay. Lemay.ai they're a vendor that's recognized with the Government of Canada. They've done work with the Government of Canada with other jurisdictions across Canada. We they gave us a reference. We were able to call them was a group in BC, and just talking with them about how LeMay was able to help them take a concept and concretize it into an actionable piece. Was was phenomenal. That really sealed the deal for us. And they've been so responsive, like, if you want to think about it, a year ago, a year and a minute ago, when we started on this journey, AI could do certain things, and now it's it's manifold more what it can do, and they've allowed us to change, if you will, AI engines as we moved along. So finding that right partner is huge. And then the other thing I'd say is really important to have a project manager. I learned this while I was working at TNO. Shout out to TNO in Toronto.
Darcy MacCallum 37:00
We worked with healthcare sector, and the healthcare sector uses project managers all the time to help keeping things focused and make sure things don't fall off the side of our desk. For the rest of us, this was one project out of many things that we're doing, our project manager was able to help us stay focused so that we kept moving forward. So for us, there was a breakthrough, and the breakthrough was when we discovered that artificial intelligence is not human intelligence. And we've had some really fun philosophical arguments about this at our leadership table, but we all agree that AI cannot replace a human maybe someday, it might. There are some people that you know think it'll take over the world. We're, I'm not necessarily one of those, those people, but, you know, and we started to call our AI Arnold, as in Arnold Schwarzenegger from the Terminator. You know, I am a machine. And the whole notion is, we want people to realize that this AI tool that we're getting it, we're hoping it can terminate our wait list, rather than terminate jobs, that's what we're looking to do, is terminate the wait list and and streamline the work that we're doing. And so we thought two ways that this could actually work for us. First, it could if we could teach the AI to curate draft action plans from our templates based on some basic client information, that would save between 20 and 45 minutes per client for our staff. That's a huge help. So that's the first thing we did. So there's a client facing piece. The second thing we realized is we really need Arnold, or the smart path, as it's formally called, to do that mundane work of checking links to and all in all our templates to ensure that information is up to date and the links that we're giving clients are live. So let me get into the nitty gritty how it works. First of the action plans. So we've built a portal where a client will go in and they'll enter their information, and then the AI will take their their information. So includes tombstone data. It includes their education, background, their experience in their profession, and areas that they're particularly interested in. The AI will go into our database and create a draft action plan, and when it's important, we call it a draft action plan, because we know the AI may get it wrong, they send that to the client and then also sets up a an appointment with the client to meet with one of our credential specialists. So there's a calendaring feature built into this whole thing, and we're encouraging people to meet with our specialists, to walk through the action plan. And if there's changes that need to be made, we make those, and then we feed it back into the system so the AI can learn from that, and next time, produce an even stronger action plan as we go. So that's the action plan side. The other side is updating the templates. So we have 162 different templates, and there's dozens and dozens and dozens of of of links that are on these. And so we need AI to regular. Be scrape that concept I talked before for changes. So it'll alert us if there's a link that's no longer active, or it'll alert us to if there's a change in information. And then humans that information comes to humans to our staff, will determine if we need to make a change in the template, and if we need to alert clients to a change in a regulatory profession that maybe they're not accepting a certain type of degree anymore. I had a friend, a good friend of mine, who decided to become an ultrasound technician. He decided to get his degree down at Niagara University in New York, and while he was doing that degree, the college decided to change their policy, and they long no longer accepted that degree. And he didn't find out about that until he was already working at a place in Hamilton, and they had to step back, and he had to go through a whole bunch of steps, because this regulatory body just unilaterally changed the rules midway through the year. It's crazy.
Darcy MacCallum 41:02
Ah, so yes, I already did that part, so I went backwards here. So some lessons that we've learned. First thing, this has been a huge demand on our IT team. We went in excited and eager and we underestimated the learning curve and the pressure that this cause we've again, honestly, one of the struggles we're having right now is we've hit a bit of a lull in our process. We've created this, this, this AI tool, and it's ready to be put into a demo so that we can show to funders, but it's we can't put it live yet, and there's been some challenges in making sure we understand, actually, the tech that's being created for us, and we don't want to go live till we fully understand that. So it's been a bit of a learning curve, and you got to build time in for that. The other thing is that there's ongoing budget requirements. So again, we were funded to create a proof of concept. I think we've done that. If we want to go live, we got to pay for hosting. We got to pay for continuing to update it, we got to realize AI is fast moving. So there's there's cost there, and then there's cost per user, and then there's then for us again, it's also looking at how AI impacts jobs, and we realize there's going to be some new jobs that have to be created to support the AI itself. It'll actually strengthen the overall system, but you have to add some new pieces to the to the puzzle. So and then I'll just, you know, Matt Marcos already addressed some of this a bit, but just for us, how we've looked at some of the risks and challenges, I know the issue of bias and misinformation. The reason that we have human agents involved in this is specifically to address these first two issues, privacy. The our AI system is a closed AI so that the information clients are giving are actually it's staying with sec. It's not going out into the public area. It's not like chatgpt, where they're gaining all you know information from you. I talk about costs already, and some of the things we're still working on, copyright, IP, risks, and we're learning about the length of time and the compliance with AI legislation and regulation. This is such an evolving field, so there's a lot of work that has to continue to be done by us. So I've touched with the LLM. Now I want to move quickly into the secondary we've worked into, you know, leverage and strengthen our internal processes, and that's playing with bots. I'm really lucky here, and our team is really lucky. We have a guy, one of our directors, Ian Richie is a chat GPT junkie, and he had some ideas of, how can we streamline access to vital information and improve processes? Now think about it, when you come to a new organization, you get an HR manual. Our HR manual is over 100 pages long. Are you going to learn that all in like one day? Are you going to be able to remember it when urgent times come? And so is there a way that we could actually use a chat, you know, AI type feature using natural language, where people could look up things and it would refer reference it to them. I'm going to demo that for you in just a minute. He created two bots. One was an immigration eligibility lookup bot. So especially we want to create this. This was actually the initial idea for AI in the first place that my previous CEO had, and this is just creating a bot where our welcome center agents can actually just ask some quick, put some quick questions forward, and they can get the information quickly about whether a client is eligible for a particular program. I'm not going to demo that because of time, although it is something we could do. The HR manual is the one I want to show you. This again, it provides staff with quick answers to HR stuff. And rather than me talk a whole bunch of this, I want to share a video, and I'm really hoping that this works for us. If it doesn't work, then you're. Have to suffer through me talking. But here we go. Oops, that did not work, just like back here.
Ian Richie 45:09
Hi everyone. Thanks for the opportunity to talk about, are you able to hear this journey with artificial intelligence at social enterprise for Canada, we can Garcia to give you a little demo of some of the thinking and testing we have been doing to try to see how SEC can better leverage this new cutting edge technology, and how we can create opportunities for our staff to start to learn these new tools and build skills. I've worked in small and large organizations over the past 30 years, so I'm very familiar with how organized organizations function and how they work. So when I was looking for use cases to learn how to create generative virtual assistants, I started with a use case where I personally know we all struggle accessing information, the dreaded HR policy manual at sec, as with many large organizations and companies, when you start your employment, you are often handed or given access to a large policy manual and a large procedure manual. For example, our current HR policy manual is 113 pages long and isn't is either in a Word or PDF document. For many employees, especially new employees, finding specific information is sometimes a challenge. Also, many people have to locate and find the document, which is also a challenge. I also know that customized virtual assistants work best when you have a source of truth document that you can refer to and have access to. So this is where I started. Let me show you a few examples of how and where I think virtual assistants may add value inside organizations. So as you can see in my screen, I've created a new virtual bot called Lisa HR. And Lisa is our HR manager, and we have many leases in our organization. So everybody refers to Lisa as Lisa HR. So I called this little bot Lisa HR, and you can see it's really an HR orientated assistant for SEC staff. I've pre programmed four different commonly asked questions, and down below you can see is where I can either where I can type in a question or attach a document. So let's start with this canned button. I am a new SEC staff member. Where do I start? So I'm going to select this machine should think for a minute, and then it'll come up with a response. And so you can see it started providing me the information. I will let this load,
Ian Richie 48:07
so you can see that it's provided me with basic information on my orientation process. It's provided a policy, manual review, very quickly, it tells me that HR downloads is where we're going to do most of our necessary information and guidelines where that's contained. It also has some important contacts, and it has some key policies to note. Okay, so now I'm going to ask it, can you please provide me a checklist for the above, correct my spelling and hit enter I'm
certainly so now, what it's done is taken the information up above and basically put it into a checklist format. If I wanted to, I could ask it to either create this in Excel and download it, or I could simply copy and paste this from from this page into a Word document. Let's try one more example. Another common question is, I'm a new full time employee at sec. Can you tell me what benefits I am eligible for enter.
Ian Richie 49:52
So it starts by going through the source document, which is the policy manual tells me my curl for full time employees at. SEC for vacation, statutory holidays. It actually calls wellness days. It actually calls it wellness days because we've actually renamed sick days to wellness days purposefully talks about leave of absence. So basically, it's going to go through and talk about some of the key benefits that you have at SCC. So that's one example of a virtual bot that I've created for our HR department.
Darcy MacCallum 50:35
All right, we could show you more, but that's the HR bot. Our next steps with this, it before we make this public for our staff, is we're asking the bot a few more diverse questions to make sure it's giving the right answers. Make sure that we're addressing gaps we've already seen, because the AI is going into our manual and it's popping out these answers we're seeing where we're actually have some inconsistencies within our manual, garbage in, garbage out. So we're making some fixes, and then the other thing that we've realized is that we need to add references to the answers. So it's not just giving you that list of answers, but it's showing you which policy, what policy number, what page number you can find this information, so that there's a source of truth, and you know that it's not just making it up. Then there's the other side about training staff. We recognize that there is a discomfort some people have with using AI, and so we're actually in the process of pulling a team together that's going to drive this project and drive some of the other questions and things that we're wrestling with around AI. So it's gonna be made up of frontline staff, mid level and senior management, and then we will go from there where there's the process of updating the bots we got to train people on doing and then learning how to integrate this into our onboarding. So anyways, that's that's the whole thing, food for thought. So anyways, time for questions and answers and or maybe questions and hopefully some answers. Thanks so much.
Marco Campana 52:04
Awesome. Darcy. Thank you for that. Really practical, really useful, to show what happens in the back end of these kinds of processes. And I can see in the comments that that people appreciate kind of that view behind the veil, if you will. We've got, we've got 30 minutes, folks. So let's dive in. There's a few questions on the slido. I just reposted the link. That's what we're using. So if you want to come in and post a comment, post a question, I'm going to read you the questions Darcy and and they're not necessarily going to be in order, but let me ask the first one, which, which I think is, is interesting. What are the downsides of using AI and client service for settlement? Do the benefits outweigh the drawbacks? Now you have, you're not quite there yet, but you're obviously having this conversation, so I'm curious what that looks like for you.
Darcy MacCallum 52:54
Um, like I said, clients are already using AI. So from where I sit, we need to, we need to engage it so that we can be talking in an informed manner with our with our clients. I think that's that's absolutely critical, because we, we live in a society where you can't even figure out what's true anymore. And AI is going to produce, if you go into chatgpt and you're asking it in a public forum what the rules are in Ontario, you could get all kinds of answers that may or may not be true. So by building systems that are we're controlling the source of knowledge that's going into the the AI, it helps us to be able to engage our staff, engage clients, and make sure that they're getting the right information. You know, obviously there's, there's a lot of concerns with AI, a lot of risks that we talked about. We're working through them. I know you've got Dan next session, and he's going to talk about their AI manifesto at the furniture bank. It's one of the best pieces. We're using it as a basis for how we're going to be approaching AI, what we're going to use it for, where we're not going to use it. And so we've, we're still in this journey, but that's sort of how I would answer that right now,
Marco Campana 54:12
yeah, and I think that's fair, and the idea of learning from other folks as well is really, is really useful and a good lesson that we don't have to pave this road ourselves, right? We can learn how there are other folks within the sector and outside of it. There's a good question about about just the development so who should we be speaking to about integrating AI in our work? Can an agency develop their own tools? Does the process have to be accepted by IRCC? That's a big question, and I'll dive into it as well. But, but what do you think about that?
Darcy MacCallum 54:42
Well, what's what's interesting is that my predecessor, Pat cousins, actually got IRCC to fund this. So they will there. There's an interest in the government to see how AI can support I think as a sector, we need to make sure that as community agencies, we're taking the lead. This conversation and not letting the government determine it for us. So we have to be giving solutions to the government, ideas to the government, so that they can be engaging with us as as equal partners, if you will, and trying to figure this out. You know, does IRCC need to accept it? Well, if you want them to fund it, I would suspect so there, although it's interesting, I had a conversation with ESDC and their foreign credential recognition office, because we wanted them to start getting a look at this as well. And there's a real interest in public private partnerships. And my hunch, you know, my crystal ball would tell me there's going to be changes in funding to settlement in the near future, and we might need to be looking at alternate ways to be able to fund these types of things.
Marco Campana 55:52
Interesting. I'll add my two cents as well. It says, if you think about I would say the answer also depends. If you're thinking of doing something pretty huge. You want to have that conversation with IRCC, but do you need to ask for permission? No, right? And I look at your We're funded by, but I look again at access, and I look at other places, and I look at the reality that we talked about earlier, which is, AI is being used. It's being baked into most people use Microsoft, 365 it'll be baked into that you, you, you don't need to ask for permission, but you should definitely be keeping them informed for exactly the reasons you're explaining, which is, we need to lead them. They have not let us. When it comes to technology, our funders, they're looking for for good practices, and they're looking for interesting ideas. And so I think the more we share what we're doing and how it's working and how we're approaching it, the better their eventual funding and support for it will be. So it's not so much a permission, it's in. The reality is, is that a IRCC is using AI tremendously, and is only going to be increasing their use of it, right? And so how it's being used in the sector needs to be something that helps educate them as well. We
Darcy MacCallum 57:03
need to be upfront with it too, right? I mean, like, that's one of the standard go tos now, especially for instance, in education, if you're using AI in producing a response in education, you need to, you need to not only say that you were using AI, you need to say what questions you were using, where you when, which, what system you were using, as well. I'll just say, you know, for me, what's probably been one of the most exciting things is, as we started on this journey, we've begun to rub shoulders with social entrepreneurs. And to me, it's the social entrepreneurs that are doing some of the most incredible, cutting edge stuff that's out there. I don't know if you heard of devont and their AI tool for working with resumes. With Kibbe, they're amazing geo focused AI that is working with employers, helping people find first jobs with immigrant networks. And the AI that's behind that whole partner, linking people to mentors, there's some really cool stuff out there. People are doing with social sort of enterprise models, different funding sources, and we can learn from them. But then we got to figure out, how does that fit within a funded environment? And that's where the challenge comes, I think, with working with you know, whether it's IRCC or anybody else, absolutely,
Marco Campana 58:17
there's the previous point you were making about transparency. There's a good question about that. So do you think that the client should know that the that an agency is using, AI,
Darcy MacCallum 58:28
1,000% Yes, right.
Darcy MacCallum 58:31
I mean, like if you as a settlement worker are using chat GPT to provide answers to your client, you need to let them know that you're doing that right and where you're where the information is coming from. I think it's, it's all about transparency. Remember, one of our four values is we build trust. We do not build trust. If we're hiding this from people, we need to be honest and open with them and what we're doing. And again, like many people, especially people that are internationally trained professionals, they're already using AI so they're not, they probably won't be scared by it, but they want to know why you're using it, how you're using it, and how it's going to benefit them and impact
Marco Campana 59:10
them some. Cynthia, if you could post your question into slido, that would be great, because I'm going through questions there, and we're not unmuting at this moment. That would be awesome. There's a good question about some accuracy. So can you share some hints to help get more accurate answers from Ai? In your experience so far,
Darcy MacCallum 59:34
well, it's for us, it's been limiting where the sources are that it's getting information. So like for instance, with our HR manual, with the bot, the only source of information is that manual, right? So it's not going out into outer space and looking at what's happening somewhere else or any other organization. It's right there. And you know this, this gets a little bit beyond my technical know how, in terms of of how we make sure we train AI, but that's. We want to do more testing with it, with regards to our large language model, the smart path AI, you know, again, that's, that's why we have humans involved. As it's presenting reports, for instance, and it's going and it's scraping a website, and it's telling us, oh, there's a change on this website in terms of the services that this that are offered around bridge programming, or there's a change in terms of requirements in order to get certified. Are we want to have references for that and have the human still check that out? We're not. We just know that AI can hallucinate, so you have to have a human involved in it that can check the sources of making sure the AI actually gives you references. That's one of the nice things about copilot is copilots already sort of built in that there's a, there's a, if I recall correctly, and we haven't got it going yet here, but it's, it's the there's a, there's a built in referencing, so that you know, where you're getting the the document and information from.
Marco Campana 1:01:05
Yeah, that sourcing is really, really interesting and important. And I mean, one of the things with with generative AI that I think is important, chat GPT is starting to get better at that. But there are, there are other chat bots you can engage with. So perplexity.ai, for example, is one that I use a lot, because it sources the data. That doesn't mean it's not going to hallucinate from time to time. You still need to check the sources, but it'll say, and it'll reference on a particular bullet, this is from this web page, this PDF, this document, for example. And so you you still have to have that final human confirmation to say, is this actually in there somewhere? Is this correct? And so the more you're a subject matter expert, the more you kind of, you might, kind of know, but, but I think that that that we have to keep an eye on, you know, the variety of different tools that are, that are out there, for example. And I think that's, that's something that's really
Darcy MacCallum 1:01:57
important. And it goes back to building the policies around which you're allowing your staff to use these tools, and what the responsible use of those tools are, right? So yeah, and,
Marco Campana 1:02:08
yeah, exactly the acceptable use of, what are the tools we're using? Why are we using them? How are we using them, and what, what's, what's your responsibility in terms of checking and and the final kind of the final kind of use of that data, for example. So that's helpful. That's really good. Let me just see. So it's a very specific question, and I think again, it's probably come up in your conversations because you've been so intentional with your staff. So as an individual worker within an organization, how can I leverage AI to support my settlement service work? It's a big question, but maybe look at it from the perspective of as you've been talking with your staff and even your AI geek, how are they looking at the practical applications of AI in their work?
Darcy MacCallum 1:02:54
That's a really good question. I'm not even sure how to answer that. You honestly, I would encourage staff to talk to their management so that their management know they're using AI, and encourage the organization to bravely go into this area together and talk about what you're doing. Learn from each other. You know, if you're, I don't know that. I want to give any recommendations of where to start, because they it's, it's, it's, it's an area that's fraught with, you did, I
Marco Campana 1:03:31
think you did, though, which is to start the conversation, yeah, you know, I mean, in my consulting work, one of the things I often find is forget AI, if you go into an organization and you talk to the frontline workers and say, Tell me about the technology. About the technology tools that you're using, and I'll be anonymous in how I report this to management, there are generally 70% of the tools management wasn't aware that staff were using because they're using them because they need to use them in their work. They need to use them because their clients, because no other tool is giving that for them. The problem with that is that they don't, they haven't figured out acceptable tech policies, and they don't tell them, because the first answer out of most managers mouths is no, because I don't understand it. So figuring out how to have that conversation and being open about it is really important, and I think that that that really matters. Yeah,
Darcy MacCallum 1:04:16
I really encourage organizations leadership to get ahead of the game on this. I was chatting with another CEO a few months ago about our project and and I just encouraged. I said, go ask your staff who's using it, but don't, don't, don't, don't judge. Just ask to learn, you know, you know, listen for understanding first, and then maybe, and then bring people together, because together, there's a lot of wisdom, and people will actually, it's interesting. People will actually be more strict with themselves than you might be as a boss. Right as as they come together, and they come up with ideas of how it can work, there's so much wisdom out there, but find those people, create the safe space to talk about it, and then you're going to be able. Before, because if you don't create the safe space, people are going to use it without you knowing, and then you can find yourself, you know, strange things can happen here in this day and age, absolutely.
Marco Campana 1:05:09
So I'm going to do a really quick self plug. People may have heard of things like the business canvas or the lean canvas. It's the idea that you can hash out a business plan in like a table. So I created the nonprofit service canvas. And it's really the entire point of it is to allow someone to take 30 to 45 minutes throw their idea down in a way that makes it presentable to to their to their manager, to start that conversation, right? And so I'll throw that in the chat. But what it does is it asks questions like, you know, who are you targeting? What are you doing now that isn't working as well. What's your solution? How will you measure the outcome of it? In essence, you're starting to create like a mini proposal. But it'll only take a there's just a really quick script to go through to help create that. So you can say, I want to use generative AI, okay, why? What are you how are you going to evaluate the outcome and the use of it? It's it creates a better conversation with, with your manager or your or your team, even to sort of talk about, I have this idea. You may have already been doing it, but this at least puts it on paper in a way that can can have a conversation with, with with each other. So just a little plug there. I found it. I've used it with folks. It helps to open that conversation a little bit. Some more good questions. I think you've answered that one. So let's see. Oh, yeah, this is, this is your bread and butter. So how can AI, or does AI help foreign credential holder, newcomers coming to Canada, connect with the right people, points of contact, looking for a job, their first job, for example, where is AI playing a role that you've seen with clients? Well, it can
Darcy MacCallum 1:06:45
give them misinformation. It can definitely give them good information, too. For me, it's the best use of AI I've seen. Is what's going on with some of the social entrepreneurs that I know, I'm excited about what we're doing, but it's, it's, it's being able to sort through the information, like we tried going into chat GPT and asking it, you know, for action plans, for for some of the, you know, for a particular profession, and some of the information's right and some of it's wrong, some of it's helpful, some of it's not. And so, yeah, I just really think we need, there's a huge value, right? You know, I think again, there's this fear that AI will replace, like, we have an interpretation translation team. You say, Oh, people can go on Google Translate. They don't need interpreters anymore. They don't need translation. Oh, yes, you do, right? Because there's, there's nuances that go on within human interaction. That's why I say artificial intelligence is not human intelligence. So I'd be interested to know if I was talking with a client. Well, what are you using? What are you learning from it? And have that interaction. That's what we encourage our staff to do. But I don't know if I have much more to add to that.
Marco Campana 1:07:58
Marco, yeah, no, you also helped answer another question, which is will, will, will, internationally trained professionals still need settlement workers if they can get the answers from chatgpt, the answer is absolutely,
Speaker 1 1:08:08
absolutely, see, it's
Darcy MacCallum 1:08:13
what, what? What you're able to provide people is information, and then you become a sounding board as a as a a specialist in immigration and a specialist maybe in foreign credentials to help people make good decisions. We believe that this information like we're really keen on getting our tool into pre arrival and just letting other organizations doing pre rival leverage it, get people to see it, because people are making decisions with wrong information before they come to Canada, they need to know how long it's going to take to get a credential, and how much it's going to cost you, and what the barriers are that they're going to face. Because everyone, everybody in this call, I'll bet you, you've talked with dozens of people who've come as internationally trained professionals without the right information. So, yeah, we got to get it out there. And they need to interact with with professionals to be able to have be a sounding board, make those good decisions. We don't want to create dependency, but we do want to create, you know, a source of trusted engagement,
Marco Campana 1:09:19
yeah, and, I mean, I think when you when we look at this, this perspective in a sense of the role, in many cases, of settlement workers or employment counselors or just frontline staff, is to help people navigate information and other and and service systems, different systems. So the idea that information overload is lessened because of all of this AI is not true. It proliferates right? So how do you as a frontline worker, continue to manage your expertise, to help people navigate that information, to be that sounding board, to say, yes, you should do your own research and then come in and let's talk about it. Yes, you should use a resume template that's out there to craft your resume if you're able. To and then come in and sit with me and we'll, we'll finalize it. We'll do the fine tuning, the editing, make sure it's your voice, make sure. But the but the tools can lead up to that. That's that 7400 hours again, right? For those who are able, you can do a lot on your own, but you shouldn't do that final thing without a human interaction with someone who has that nuance in the same way, like language has cultural and other nuance. Employment has many nuances that that a generic AI isn't necessarily going to capture, right? Yeah, we had
Darcy MacCallum 1:10:27
quite a few conversations with our staff about whether or not we should send the draft action plan to the clients before they meet with a credential professional. We went back and forth and back and forth. And one of the things that sort of sealed the deal for us is that there are people that are already going to be out there that are really self sufficient and want to just figure things out on their own. So if we can give them more accurate information, they can go, and then they know where they can come to get support. They know where they can go. Because this, this, this, this, this group helped me, and they didn't, didn't tie a string to it. So trust that people will come to you when they have their questions, like I talked before. It's not about creating dependency, it's about creating a reliable source and partner as people journey through their settlement process,
Marco Campana 1:11:21
and there's a follow up question that I think I'm going to partly answer, and then I'll lead to you, but that that leads to the idea of mitigating risk to the organization using AI as a client service tool, or within the organization, if we bake in prompts in the AI to say, this is a good time to check in with your worker or confirm this information, right? Like, I mean, this is the thing that AI is so good at convincing us it's right all the time, until you ask it. I don't know if you've had this experience when I've said I think that's actually factually incorrect, and it's like, Oh, I'm sorry, you're right that I wrote the wrong answer. And if you don't know to challenge it, because it's so persuasive, it writes in such a human way. What if we, in our AI tools, anticipated mitigating risk, so that throughout the process of a chat, let's say, or an interaction with it with an automated bond, the person is reminded that there are human beings they can connect with. And in fact, would that potentially even lead people to who aren't accessing service to understanding that there is some, there's a human behind this that they can start accessing Right?
Darcy MacCallum 1:12:25
Exactly. That's why we built in that. When you get your draft action plan, you automatically get invited to book an appointment, right? So that it's, it's, it's cueing you that, hey, you know what? You should talk to somebody who's who's got experience with this. So it's, yeah, leveraging AI. That's our client interface. With the AI is very limited to that initial piece, and then it's all human engagement from that point on. So the AI is behind the scenes, so that's how we've sort of created it.
Marco Campana 1:12:55
Yeah, and you've addressed this next this is coming right out of this conversation, but you've talked about this in the beginning of bringing AR into the conversation. But the challenge about adopting AI is that they don't have rules or policies about AI. And Jessica, I'll send Darcy, and I can send you a whole bunch of stuff. It starts with that data IQ, those very specific questions. But there are good frameworks that are out there that we can share with folks. The problem with that is it's info overload, right? But practically speaking, Darcy, how can we help kickstart this information, or this conversation with HR or an IT department, if they have one?
Darcy MacCallum 1:13:27
Well, I think like we're using Dan's stuff from the furniture bank. I mentioned that is the his is one of, I think it's the best out there in the nonprofit right now, in terms of a clean, concise, this is why we use AI. This is how we use it. This how we don't use it. And again, like we're doing, we're pulling frontline staff together. Now we wanted this to not just be a ivory tower conversation between directors and managers. We want our frontline staff to have say and have their voice be part of the process as we determine these, you know, this pathway, and address risks and how we're going to use it.
Marco Campana 1:14:09
So I threw a link in for folks into the chat that is a link to an article that has some frameworks, but also near the bottom, it's got all of the links to furniture banks. They've created a responsible AI manifesto. They talked about how they've used AI. And you, again, I definitely encourage you to come to his session. He is definitely a proselytizer of use of AI. He thinks we're not using it enough, but he's also very focused on responsibly using it, right? So in their case, they used it in a campaign where they created images to avoid poverty porn, or Yeah, poverty porn, where instead of using real people, they used AI, generated people for their their fundraising campaign, and the implications, and in doing that, and they still work with a professional designer, right? They didn't just throw stuff in, but you'll hear more about that from him, but I agree, I think that there's, there's much to learn from how they're approaching this, but there are also other a. Frameworks that have been around, actually, for years, that we can, we can, we can play with right? Entire, entire workshop guides and, yeah, you every, every day there's a new webinar out there, right?
Darcy MacCallum 1:15:11
Yeah, I'm a believer that we have to be careful. Like you said, there's information overload. I think if we bring the people together in your organization that are already passionate about this and interested in it, you will be able to glean the wisdom you need to chart your pathway as an organization, and then talk with other organizations. You know, we'll be glad to talk with anybody about that. And I know the other presenters. I mean, everybody is presenting here I've chatted with at some point, right? And you learn from each other, and so it's to me, it's the journey is as important as real. I could just take dance stuff and just plug it onto our thing, but then we don't understand it. We haven't internalized it. So let's do that together as a team. We find solutions together. There is another value of ours, right? That's how we do it.
Marco Campana 1:15:58
That's right. So just a reminder about the realities of a technology world and digital inclusion. So what approach does your organization follow to deliver services for clients with low digital literacy? I think it's an easy answer, but just curious.
Darcy MacCallum 1:16:16
Well, from the inner for internationally trained professionals, that's not generally an issue. And then I'll be honest like I said, our settlement, our settlement piece, is very, very small. So within our welcome center system, we have lots of resources that can support people that may not have that literacy. So about 80% of the people that access our foreign credential services. Want virtual meetings, not in person meetings. That tells you something about the clientele already. So we would lean on partners. If there's things people that don't have some of that capacity. We don't have to be all things to all people at sec. We want to partner with people that can provide some of those other services,
Marco Campana 1:17:02
yeah? And, I mean, the short answer is, you'll, you'll meet with them in person, but the AI, as a staff tool, which is what you're developing, will help support that in person interaction, right? So AI is still a lever. You're still leveraging the tool, but you're leveraging it in an in person interaction, or a low fidelity technology interaction, for example, maybe it's email, for example, an email back and forth, or an in person. And so I think that for everybody to remember that you know you're not stopping delivering services for for clients who have low digital literacy, but also don't assume that they have low digital literacy. This goes back to the whole center your clients and find out right when I ask. I mean, if you, if you think about if I said here to people, what questions do you have at intake and assessment that ask and tease out the digital literacy, comfort and desires of your of your clients? Do you even ask if they have a cell phone number? Most intake forms just say phone number. Do you ask if they want to communicate with you using email, WhatsApp, WeChat, Facebook, messenger, Viber, you name it. Or are they happy with with email? Do you ask about their their access to the internet, the devices they use, and how you know? Do you know that your clients are predominantly mobile friendly and don't have, necessarily tablets, laptops or desktops. And if you're sending them things that look terrible on a small screen, you are literally not serving them properly, for example. So we need to also build in that literacy, and it's easy to do at the at the stage of intake and assessment, asking a couple of questions, imagine, and I say this to people, let's say yourself to what you serve 12,000 people a year. If you ask people, How would you prefer to that you that we communicate with you, and at the end of three months, you have 3000 points of data, and 80% of those people say, I would like you to WhatsApp message me. You have the beginnings of a digital strategy for communicating better with those clients. You still need to serve the other 20% in other ways, but it tells you where you need to invest, right? Instead of saying, I'm going to email you, and they're like, I don't use email because increasingly people don't. They're coming from places where they use digital messaging, which you as frontline workers also predominantly use. So making those connections, making them data specific. So that the argument isn't I just know it. I Intuit it. The argue is, here's 3000 points of data that tell us that we need to move in this direction, for example, and that's really important as well, to not assume client literacy or illiteracy in any meaningful way. All right, we've got three minutes left. I
Darcy MacCallum 1:19:31
love how articulate you are on this Marco. It's not like I've been saying
Marco Campana 1:19:34
that for years or anything like that. I think we've gotten to most of the questions, but I just want to make sure yes, there is definitely a learning curve. It depends on how you use it absolutely. So yeah, so in the same way that we are helping clients navigate this technology, we need to figure out how to navigate it ourselves. And I think that's part of the lesson from your presentation, Darcy, is that this takes time. Yes.
Darcy MacCallum 1:19:59
Be familiar for what we're going to implement as well, right? You can't just assume, Oh, we're going to use this, you know, and it's like anything else, it's got to be integrated into our onboarding and ongoing training.
Marco Campana 1:20:10
And I think that's an essential lesson. Is that this is a deeply communicative and consultative process internally in your organization, right? That is really important. And yes, this might be a question for another session, and it will be for achieve and access. Actually. How do you see AI working with helping with documents and applications, filling tons of personal information and repetition there, and I will say to you that there's actually a former immigration lawyer who is figuring this out. He's shifted his business entirely to an AI tool called visto AI, and you can play with co pilot, another, another, another co pilot is out there which is exclusively trained on IRCC information, and I follow him on LinkedIn, and one of the things he's saying is that he's seen immigration lawyers who, in the past, stopped serving refugee applicants, for example, because it's too it takes too long and the ROI is too low. But with a with this tool, for example, they're actually starting to pick it up again, because it takes the mundane, the repetitive stuff, and it throws it through an AI system that allows them to serve them more quickly and at scale. And that is a lesson for our sector. This is part of our sector, right? Immigration Lawyers, immigration consultants, so we can learn from what that looks like and lay Yes, the question is, you can't attend the next sessions. There will be slides and recordings available, right? Jessica, so I'm going to hand it back to Jessica. There are I'm going to say that I think we've answered all of the questions, but I will review those questions, and I'll send any to you, Darcy, and we can maybe send them out to folks afterwards as well. But I just want to say, Darcy, thank you for this really practical, insightful view of what it means to implement AI with your staff and within your organization, still a work in progress, and I think that's part of the lesson. Absolutely. Thank you. Thank you for this. Any final parting thoughts before we let Jessica jump in.
Darcy MacCallum 1:22:03
Do it with your team. Talk about it. It's all about people. Don't forget that.
Jessica Kwik 1:22:09
Wonderful. Thank you so much, Darcy, and I think I echo Marco on the practicality you shared on how to make this work and to start simple, and your advice on really working across the organization. So we'll be following up with the slides recordings and any links to share, as well as a feedback survey. So I know there may be other questions that come up, so feel free to add that to the feedback survey as well. And you know, just, I know there might be discomfort, so just want to echo Darcy's comment about wanting to support staff rather than replace them, and and just really looking at, how can we really work towards leveraging AI for our the clients that we serve. So thank you both so much. It's a wonderful opportunity to learn from both of you. So thank you both Darcy and Marco and everyone for attending. You.
Subscribe to get the latest posts sent to your email.
Please take this short survey to help improve the KM4S web site. The survey is anonymous. Thank you for your feedback! (click on the screen anywhere (or on the x in the top right corner) to remove this pop-up)