CGI’s From AI to ROI podcast series features expert discussions on how AI drives change across organizations and how to achieve trusted outcomes. In this episode, host Helen Fang is joined by CGI experts Dr. Diane Gutiw and Andy Donaher to explore the topic of AI for good across three dimensions: environmental, social, and governance. Diane and Andy share insights and real-world examples of how AI helps to address societal challenges, improve organizational performance, and enhance well-being. This episode is part 1 of 2.

For the latest information on CGI’s responsible use of AI principles and practices, read our 2024 Responsible AI Report, or visit Responsible AI on cgi.com.

Key takeaways from the episode

1. AI for good focuses on solving real-world problems, not just on applying AI technology.

AI for good begins with identifying challenges and opportunities for both clients and communities, and then defining what information is needed and which technology to apply to address those challenges and opportunities. AI in many cases helps solve the problems that have been identified and accelerate positive outcomes.

2. AI can address challenges that were too expensive or complex to address in the past.

AI provides fast access to information, enabling us to make better decisions. This, in turn, increases business opportunities and reduces risk if the right guardrails are in place to avoid misuse and errors.

3. AI has been used to enhance fire safety and mitigate housing risks.

CGI collaborated with Mustimuhw Information Systems (MIS) to assess fire risks in First Nations communities using AI, data analytics, and digital twins. In Canada, a First Nations person is 10.2 times more likely to die in a fire than a non-First Nations person. By creating 3D digital models of structures and using AI and computer vision techniques, the CGI project aims to improve risk mitigation, insurance, and safety. Similar AI-driven approaches have been used in the UK to assess government housing humidity and temperature levels for the purpose of identifying potential insulation or window replacement issues.

4. Applying technologies like AI or quantum computing helps drive supply chain optimization.

Applying AI to supply chains, such as transportation and trade routes, reduces costs and carbon emissions. AI-driven efficiency improvements such as recalibrating vessel speeds in canals also reduce costs and benefit the environment. As technologies like quantum computing advance, both supply chains and scheduling can be optimized in ways that were not possible before.

5. There are practical AI best practices that enhance daily work beyond efficiency and cost savings.

These best practices include reusing existing AI models instead of building models from scratch, optimizing processing time, and leveraging research-backed efficiency principles. All of this improves sustainability and reduces costs. AI knowledge management is another example. In addition to driving efficiency, it reduces user stress by minimizing manual information searches and enabling an increased focus on higher-value work.

Discover how AI helps enterprises and government organizations achieve tangible and trusted outcomes.

Also visit our AI home page for insights, resources, and news on AI-driven strategies.

 

Read the transcript

Introductions
AI for good: What does it mean?

Helen Fang

Thanks so much, Andy and Diane. Today, we're going to start off our 2025 for this podcast by discussing AI for good, as it's a topic that's not only important to us as individuals and as a company, but also to our clients and to the communities that we work in. We are also about to release our 2024 ESG report, and this is going to be a new dedicated section that we've included this year as part of that report. Diane, Andy, anything to add on the report part?

Diane Gutiw

Yeah, I think it's really good timing. As you say, the ESG Responsible Use of AI report is coming out and it gave us an opportunity to look at what we've been doing in investing in AI with our clients, to be able to really focus rather than on just what can we use the technology for and shift that to what are some of the AI for good opportunities, how can we use this to really make a difference for organizations and opportunities as well as societally. So, you know, what we're going to do today is focus this conversation on each of the three ESG components of AI for good, environmental, societal and governance. So, I think it's going be an interesting conversation. Lots of examples we're going to share on where we're seeing organizations really getting benefit from the tools when they put those guardrails in place.

Andy Donaher

I think one of the things that we don't do enough is understanding how when we're implementing some of these solutions, the externalities associated with those for good. For example, when we're implementing some type of knowledge management support to help people get information more quickly.

Well, there's a knock-on effect from that around helping to decrease people's stress so that they can do value-added work and saving them time and helping them to remove those unnecessary tasks. And I think through today's conversation, I think we're going to realize how there's more good that we can do by using these tools that are already things that we're doing today and we can just look to enhance and accelerate. Just wanted to communicate that. Thanks.

Diane Gutiw

Yeah, that's a brilliant point, Andy. AI is really just a tool, and it's an amazing new tool we have in our toolbox. But we have been working with machine learning for many years and have some really great best practices that we're now able to do more. You know, the real difference, as you're saying, is we now have a tool that we can do some of those things that were complex or expensive to be able to find patterns and information.

Helen Fang

Yeah, thanks so much both of you for those really important points. I think we've heard also that a lot of the conversation around AI has been very focused on, for example, efficiency gains. And sometimes we're not measuring as much the good that it's doing for people, for their wellbeing, some of these other aspects.

So, to start off, I wanted to hear a little bit more about what does AI for good on a broader level mean to you? Diane, do you want to start off with that topic?

Diane Gutiw

Yeah, to focus it a little bit more both internally and for our clients, AI for good means that we're starting with a problem. We're starting with something that needs a solution. And when you focus on what is it you're trying to solve, you're able then to refine the information that's needed and refine the technology needed to be able to help solve that problem. And as Andy and I was just mentioning, AI is a tool that provides constantly evolving capabilities that we’re able to do things much quicker that we were challenged to do before because of the manual effort that was required. So, AI for good, when we look at the lens of what we're doing internally for our clients and societally, is really focused on a way that we can get quick information to be able to make more precise and better decisions to increase our opportunities and reduce our risk.

AI for good is making sure that we have guardrails in place so that we're avoiding the risks that may come with the use of a new technology, particularly a technology that ingests data and is able to provide outputs based on that data, really making sure that we're focusing on what those risks are and mitigating them so that the tool is used for its intended use. We're avoiding misuse. We're avoiding errors. And, we're really bringing value to people in what these tools are able to do. That's really where I see when we look at it as an organization that's looking both internally and for our clients at what AI for good means.

Andy, I don't know what's your take on it with a slightly different lens?

Andy Donaher

We see it very, very similarly. People often don't realize how they're using it for good. It's not a one-way street. So, if you're thinking about it, for example, from an environmental impact perspective, yes, measuring Scope 1, 2, and 3 and making sure you have a path to net zero etc. is extremely important. And as you're doing that, when you're looking at, for example, supply chain optimization. So, for using AI in its modern form to help with optimizing transportation routes, optimizing vessels, optimizing trade routes and intermodal is all helping to decrease carbon footprint. It's helping to improve your bottom line. It's helping to improve scheduling optimization for your drivers and your conductors and your pilots.

Helping people understand how these benefits to business can also support a multi-pronged sustainability approach. I think, for me, that's the biggest thing that we're seeing: an evolution in the view of sustainability and how it benefits both society and performance of organizations.

Opportunities in AI for good: Client examples and trends

Helen Fang

I totally agree. I hear a lot of considerations around the sustainability topic. But I think, oftentimes, the discussion is focused on risk, which is also quite important, but maybe a little bit less on the opportunities that we see there. How have you seen clients think about this topic and are there any commonalities or key themes and priorities between the different regions and industries?

Diane Gutiw

You know, what we're seeing and what we heard from CGI's Voice of the Client is that a lot of organizations are exploring these tools, whether it's the general purpose chat interface tools to be able to get quicker information, whether it's RAG models, which are allowing them to inquire against their own documents and data, all the way through to much more complex integrated systems – so lots of pilots going on. We've seen a slow uptake in those pilots moving into production across all industries and all jurisdictions. I think the world is kind of working at the same pace of moving those things into operationalizing.

Where I think we're really going to see that shift is technologies like agentic AI, which I'm sure a lot of people are hearing in the news and a lot of our technical resources that we're reading, is really focused on fixing the gaps that made some of the concerns for adoption and uptake in the initial releases of the large language models filling in, providing autonomous agents to check the quality and to validate the quality and measure the quality as well as to look at streamlining processes across complex tasks to be able to really get value out of the tools. The other thing I think we're going to see and we're already seeing is measuring what the value is. So, at the beginning of initiating a project, let's say it's with a contact center. Starting at the beginning, our intent is to reduce the time to resolving issues, to be able to increase our capacity to be able to address more issues quickly.

And if at the beginning when you're stating what is it we're using this tool for, you identify what's the benefit I intend to get from it, making sure you're actually measuring. Did you get that benefit and if not, what needs to be adjusted to make sure you're getting the benefit? So, that's what I see, a lot of early adoption and exploration for pilots and a slow move into implementation. But I think, the floodgates for that are about to open with solutions that are solving some of the questions people have to move it into production.

Andy, with the work that you've been doing with clients and large organizations, does that resonate or what are you seeing in that space?

Andy Donaher

Yeah, for example, there's one our Canadian team did, and our team in the UK did something very similar in terms of using AI with digital twinning for societal impacts and the determinants of health.

There's an organization here in Western Canada called Mustimuhw Information Systems (MIS). It's a First Nations consulting company owned by the Cowichan Nation who provides information services and platforms for nations across the country. I believe they have over 300 nations that they support. As they're building out the Indigenous digital health ecosystem, they approached us and asked to see if we wanted to collaborate with them in building out the Indigenous Digital Health Ecosystem, IDHE.

And so, one of the things that they brought to us was that (I didn't know this. It was a great learning for me.) in Canada, a First Nations person is 10.2 times more likely to die in a fire than a non-First Nations person. And an Inuit child is 13.6 times more likely to die in a fire than a non-First Nations child. And so, it was very impactful to us when we were thinking, “Okay. How can we use AI data analytics to help support this?” And so, one of the challenges around it is being able to gather the data.

You know, distributing a number of fire marshals around a country as vast and broad as Canada can be a challenge. And so, when we stepped back and looked at it with MIS, we came up with the idea of creating an app for the phone that as you walk into a structure, you can take a variety of photographs of the structure and then you can build a 3D digital model of that in real time.

And then, in addition to that, you're using AI and computer vision techniques to detect any sort of fire suppression, fire detection assets in the structure. So, smoke detectors, smoke alarms, fire extinguisher, these types of things. Then what you can do is not only evaluate and create a risk score, but make recommendations around the types of things that you can do to help improve the risk management of that, how to identify things that you can do to mitigate risk that improve your ability to get insurance, that improve your ability to decrease insurance, and then improve your ability to mitigate risks associated with harmful events. And then, additionally, what that's going to be able to do is then be able to look at the environmental factor. So, is a particular property decreasing in heat faster than another in the same conditions? Is there different humidity levels, etc.? And that's one of the things they did in the UK, was evaluating the humidity levels and the difference in temperatures to determine in council housing where there may or may not be issues related to insulation or window replacements. So, you can target your funding rather than a broad swath of funding. So, there's different types of things like this.

We can also talk about green IT. We, at CGI, have made great strides in green IT and it's a service we provide to our clients, where we can look to decrease the carbon emissions from our data centers. So. we're using things like motion detector lights. We're using things like renewable energy purchases to provide power for our data centers. We're doing all these types of things to get to net zero across those and are able to support our clients with that. It's also beneficial to the profitability of the organization.

Diane Gutiw

I think that was a really good overview of a lot of the different areas that we wanted to cover.

I think, you know, if we break down the ESG into those three areas that Andy just gave us a good overview on, environmental is the first one. And I know, Helen, there's a section in our Responsible Use of AI ESG report on addressing not just the compute power and resources needed to support these evolving AI technologies but also in best practices around optimizing models. And Andy, you touched on that a little bit.

So, I'm actually going to pass it back to you to talk specifically about some of what you've been doing, exploring with the team on quantum computing and other methods of green IT.

Andy Donaher

Sure, I'd love to. So, there's a couple of things that we've been working on around that. So, quantum computing, I have personal experience with quantum computing, having used it for the optimization of supply chains and the optimization of scheduling. You know, not everybody knows that it's a viable solution that's currently in production.

Organizations like D-Wave and Microsoft and others can use it now on a daily basis. And when you're looking at supply chain optimization, and you're looking at complex problems with a lot of externalities and a lot of integrated variables, it really does help to be able to take those things in.

Actually, this didn't require quantum computing, but one particular organization did an analysis of their vessels through the canal and what they determined was that the pilots, when the lock released, they would accelerate out of the lock and they would accelerate to the next one. And then, when they got to the next one, they'd have to reverse the engines to slow down to wait. And so, by doing that analysis and determining, this is what's actually happening. They were able to reprogram, recalculate the necessary speed coming out of the locks and they saved a tremendous amount of money in fuel, but also had a decrease in carbon offsets because they were managing it more efficiently, more effectively.

And so, if you start to think about the expansion of that in terms of not just in the smaller segments, but the analysis of larger segments in intermodal transportation and the ability to use quantum computing to solve NP-hard type problems or NP-complete type problems where the variables are very large and interacting. There was one particular problem we were working on, where, in trying the solution after 35 hours, we just turned the environment off because we couldn't get a solution returned. And then, with quantum computing, we were able to solve it in eight seconds and then able to optimize from there. So, being able to apply that to your supply chain, being able to optimize your costs, being able to apply that to scheduling for people, those are all opportunities that we're currently working on, things that we've delivered. The opportunity in front of us to make very, very significant impacts in the very short term, immediately is profound, to be honest.

Best practices for using AI in our day-to-day

Diane Gutiw

Yeah, I think those are some really impactful examples. Before I hand it back to you, Helen, you know, what are some of the tactical things that everybody, you know, we're not all involved in these big opportunities or don't have a say in how the hyperscalers are leveraging energy and compute power. But what can you do day-to-day to focus on environmental concerns when it comes to AI?

And some best practices that I know is, you know, simple things like reuse models rather than developing the same models over and over again. If somebody's already run a model, leverage that rather than repeating the same thing again, being careful and cautious and cognizant in your use of these tools to get answers to questions, because all of these things impact the amount of processing time. And there are some great best practices coming out of research and research principles to optimize your use of AI and machine learning and training the models. So, I think there's lots that we can take away from research.

Helen Fang

Thanks so much, Diane and Andy. Diane, to your point, I think it's really helpful to think about how we're using AI in our day-to-day and when we're doing any sort of AI development or working with the systems, that we should also think about these considerations. And to Andy's point earlier too, often also has some cost implications. And I think that's one thing that everyone's always happy to do, is to reduce their costs in different areas.

I think, before I move to the next one, that Andy also made a really great point about supply chain. And we've recently started talking at CGI also about the citizen supply chain, so this wider view on our communities and how everything is interconnected. So, I think it's really great to hear some concrete examples about how we're applying AI there and how it's helping in different areas, because oftentimes we hear about the potential. I think it's not as common actually to hear about these concrete outcomes and what's really already been implemented. I think, sometimes, we're actually much further along than maybe people who aren't as involved day-to-day in the AI topics as you realize.