What are Canadian companies actually doing in terms of projects with cloud and generative AI? How are they navigating the uncertainty of a relatively new technology? Those were the questions that were addressed by a customer panel during a briefing with Canadian media at AWS re:Invent 2023.
The panel featured a number of companies from different verticals, ranging from born in the cloud startups to large established enterprises.
Here’s some of what they had to say about the use cases they are working on, and the rationale for using generative AI versus more traditional approaches to AI such as machine learning (ML):
Dwayne Benefield, SVP connected homes & entertainment, Telus
In customer service, Telus is deploying an AI engine to improve customer service in call centres, to “take questions, understand intent and generate answers.” But beyond that, Telus is looking to AI to take a further step and help troubleshoot issues.
Benefield asked us to imagine someone “streaming Netflix and all of a sudden it goes down and you’re like, ‘what’s wrong with my WiFi?’ AI is capable of checking a wide range of potential issues and proposing actions to resolve the situation. These answers can be real time and proactive.”
He also outlined Telus’ plans for smart homes, focusing on an issue that is a sore point with everyone who has adopted more than one smart device in their home.
“Every device you buy comes with their own hub,” he stated. “That’s not very smart.” As a case in point, he noted, “I actually have 39 apps on my phone right now (to manage smart home devices).”
Benefield knows that AI can change this. “We’re leveraging AI to remove that complication and actually make it more personalized and easy to set up,” he said.
He announced that Telus is working with AWS to to build the world’s first device-agnostic smart home solution. “Now all the devices you have, or devices you plan to bring in…are controlled by a single pane of glass or a single voice system.
“Instead of having to program your morning routine in five different apps…you just say, ‘hey Jarvis, wake me up at 7 am, open the blinds and get my morning coffee’. And Jarvis will say, ‘what temperature do you like your coffee?'”
Benefield noted that Telus is deploying this in beta next month, and in consumer-facing applications starting in Q1.
Eman Nejad, head of data science, WAHI Realty
WAHI is a startup, less than a year old, but the company has already become a disrupter in the real estate market, with a “cashback” promise passing the savings back to the buyers. Their AI offering promises even further disruption.
The company is using AI to pull information from the photos supplied by the sellers. Doing this means they can go beyond the standard listing information to personalize the search for the home on a new level.
Nejad noted, “we can spot things like specific cabinets, finished basements or a specific type of backyard.” This adds “a new dimension to the industry.” These AI-enabled image searches “automate the interface, and the image (will) clearly help the home buyer to find the (most) appropriate, more personalized, and relevant sale.”
The company had originally started with standard machine learning solutions. But when generative AI came to the forefront, Nejad realized that it was far superior to standard approaches for this particular type of application. It was also much faster to implement. As a result, their new system is in beta test today.
Iran Reyes Fleitas, VP, global head of engineering, Experience.Monks (Formerly Jam3)
Experience.Monks offers highly creative, technically enabled services for a wide range of major brands around the world. For them, generative AI adds new dimensions to their offerings.
For example, they would traditionally create one ad for one region. If that was successful, they manually updated those ads for other regions, customizing them for language, locality and ethnicity. This was time and labour intensive work.
Fleitas said they are working to leverage generative AI to allow them to “create an ad on one site, and it creates and replicates that ad in all the regions that you want.” The customization covers language and even the images used. “So what used to happen in a month, now is happening in a few days.”
Another example was their work with Adidas, where, Fleitas noted, “we created a platform that could generate a sneaker in 3D based on user prompts.” Further, and this is the critical point, he noted, the AI will ensure that customizations are within the brand’s guidance. “So in the end, we’re not going to create a Nike sneaker, we’re going to create an Adidas sneaker with all of the personalization of what the users (want).”
He referred to this as “hybrid personalization,” where AI can ultimately allow customers to actively help customize products.
Everyone who has interacted with generative AI is taken by how it feels like conversing with a real human being. Experience.Monks leverages this to captivate their audiences.
The company is using AI to produce surprisingly realistic avatars to be used for customer service. These shockingly realistic avatars not only provide answers to a wide range of questions using natural language, they also detect and respond to emotional cues and sentiment indicators.
The company has gone beyond images on screen. It has used AI to animate a mechanical and comical puppet creature used in one of their programs. Usine AI, the puppet, truly comes alive, with lifelike interactions, as it follows facial and verbal cues.
Generative AI made it possible to breathe life into these devices, but also to do so quickly and within the budgets of their customers.
Jean-François (JF) Gailleur, SVP engineering, AlayaCare
AlayaCare is working with AI enabled solutions to address systemic issues of home care that have become more acute with the combination of an aging population, cost restraints and the shortage of trained staff.
AlayaCare is looking to generative AI to deal with three key issues:
– attracting and retaining people
– scheduling
– managing patient data
With a shortage of skilled workers, it’s essential that they get the maximum productivity from everyone in the field without causing huge amounts of stress that could lead to retention problems.
Any time even one person is unexpectedly unavailable for work, it can cause major disruptions and significant staff time to manage. For example, an average home care worker may have 40 clients, all with different conditions and needs. If you have 50 similar workers in an area, that’s a huge population.
So if a worker is off suddenly, the number of permutations of clients, location, special needs and times make even minor adjustments a logistical nightmare. Yet, managing this many permutations and finding the optimal balance is a task perfectly suited to generative AI solutions that can manage this type of scheduling in real time.
Yet another issue faced by healthcare workers is in dealing with the vast amount of data that needs to be collected on a patient. Sorting through that amount of data takes time that could be spent in patient care.
Once again, AI solutions can review this data and do the “heavy lifting” to create and help manage individualized care and risk profiles. Freed up from sifting through, analyzing and prioritizing records manually, caregivers can have an AI “dashboard” to help manage their work.
Gallieur was careful to note that these applications don’t take the place of trained staff and clinicians, but they free up those trained workers to spend the time where their precious skills are most needed – on patients, treatment and services.
Anne Steptoe, VP of engineering, WealthsimpleÂ
Wealth Simple already has significant experience with traditional AI solutions in a cloud based environment. As Steptoe noted, “Our company is digital native, born in the cloud, so I think that makes us more agile. Our data warehouse, our production workloads, our ML platform, are all hosted on AWS. We have built our own ML platform (on AWS) … that allows us to be really innovative and have a much faster response to our clients.”
But generative AI solutions in the cloud make the company even more responsive and agile by vastly reducing the time to delivery required by traditional ML. Steptoe noted that recently, the company was able to leverage generative AI to “put out a brand new model in two steps with one line of code.” As she noted, “it’s pretty amazing what we are doing with this technology ourselves.”
Steptoe noted that productivity gains in areas like call centre applications accelerate productivity but also allow their teams to provide the type of oversight that is “making sure they are checking, making sure that (answers) are as accurate as possible.”
Appreciating the strengths and challenges
All of the panelists acknowledged that there were both advantages and challenges to working with generative AI models. None claimed that these models were a panacea. But even in the early stages of generative AI, each company has been able to generate high degrees of productivity, increase customer satisfaction, and have extraordinarily fast time to market with their products, services and ideas.
Generative AI is not ideal for every application, but where it can be employed, its speed and capabilities allow for applications where ML would be too expensive or take too long to implement. It offers speed to delivery which is orders of magnitude faster and cheaper than past approaches. And in some unique areas, like, for example the way it can deal with visual and unstructured data, it has capabilities that are not available from traditional ML applications.
With the right use cases, and with human oversight used strategically, it is, even at this stage of development, a powerful tool for these businesses, and one that will continue to improve over time. As it does, these companies will have the added advantage of having developed the experience to leverage it and exploit its full potential.
