BC Business
AI is changing how we work, learn and apply for jobs. But how are institutions managing the ever-evolving tech?
Reshum Zubair is hunting for a job, and she knows she faces competition. “You really have to bring your A-game, and you have to apply first to be noticed,” the UBC Sauder School of Business student explains. She’s studying business analytics to add new levels to the data-driven portfolio she had built in her home country of Pakistan while working for international companies like Levi Strauss & Co. She graduates in August with her master’s degree.
Unfortunately, Zubair’s perfectionism slows her down. “I really take too much time to apply to jobs,” she confesses. “And because most of my time is taken away by customizing my application, by the time I send mine out, there are already thousands of applicants.”
Zubair is solving her problem using ChatGPT, OpenAI’s wildly popular artificial intelligence platform, launched at the end of November 2022. She started by compiling a document with all her work experience and products, and then training ChatGPT to customize job applications for her. Now she tells her AI assistant: “I’m going to give you a list of jobs that I’m going to apply to, and I’m going to need you to extract the most relevant lines and the most relevant projects I’ve done to put into a one-page CV.”
Her quickness in adopting the new technology puts her ahead of most Canadian businesses. Just 6.1 percent of companies in the country made use of any AI tools in producing goods or services during the 2023-24 12-month period that Statistics Canada looked at.
But companies and investors are flooding money into an AI race that will revolutionize how the world conducts business. Global private investment in the sector lingered in the low single-digit billions between 2020 and 2022. Market intelligence firm IDC expects 2024’s total to ring in at US$235 billion and nearly triple from there to some US$630 billion in 2028.
Business schools across B.C.—and the province’s premier technical institute, BCIT—have always prepared students to adapt to rapid change, and to drive some of that change themselves. But many of the billions of dollars of AI investment globally are pouring into research and development, not the production and delivery of tools available now. Computer scientists are still grappling with what the emergent technologies will be capable of performing, and how they will change business and society. Educators are likely to launch students into a wildly different future—one where AI’s eventual impacts are nowhere clear enough to grasp. The biggest changes are years downstream.
Artificial intelligence will, however, create opportunities for smart, skilled people who can use it to leverage their own capabilities. It’s a broad category of technologies, with many different approaches and potential use cases. Its applications may be developmentally nascent, but people like Zubair who know how to put them to good use are already producing transformative effects for themselves and their businesses. These platforms will grow ever more powerful and ubiquitous. Students learning to harness AI today will be taking their skills to the workstations, managerial offices and C-suites of tomorrow.
UVic Gustavson School of Business assistant professor Andrew Park is an academic with roots in tech. He started his career as a software developer and founded and sold a successful Seattle-based health technology startup. He’s interested in the technical development of the different artificial intelligence models, but in his current incarnation as a member of UVic’s business school, he’s also investigating what their development means to organizations and the public.
Most of the public, he says, is only aware of the tip of AI’s iceberg. “So you’ve got your large language models like ChatGPT that everyone knows about, and then your image models, but then there’s a whole constellation of different AI models that try to achieve different things,” he explains.
Large language models, or LLMs, are trained on enormous amounts of language data to converse in human-like ways. You can ask ChatGPT: “Give me a seven-day itinerary for the Amalfi Coast in April,” and it will suggest how to spend a week in Positano, Capri and Sorrento. But that wizardry and the publicity surrounding it overshadows the business applications of LLMs like ChatGPT and competitors like Google Gemini or Anthropic’s Claude.
Park points out that LLMs can work on any kind of written text, including software code. “Claude has become so sophisticated in software development now that you can just give it some natural language instruction and say, ‘Build me a skeleton of a very basic Twitter competitor,’ and it’ll give you two to three hundred lines of code. And then you can go in there and start tweaking it as you see fit.”
Indeed, Alphabet CEO Sundar Pichai announced in his company’s third-quarter 2024 earnings call that “more than a quarter of all new code at Google is generated by AI, then reviewed and accepted by engineers.”
“People don’t realize that these LLMs are very specifically disrupting professions and categories of professions,” Park emphasizes. “People aren’t just using them to, for example, help them with a homework assignment or write marketing material or a strategy plan.”
Software engineer Wilson Scott is focusing on his own skills while LLMs move into his industry. He left his law practice in 2021 to earn a computer systems technology diploma at BCIT, enrolling in its artificial intelligence and machine learning option. He graduated in 2023 and works as a software engineer at Vancouver-based 3D-scanning company Polyga Inc.
Scott’s company doesn’t stop him from asking Claude or other LLMs to write or augment his code, but he makes a concerted effort not to use them. He wants to use the practice time. “This is a second career for me, but I am a junior,” he explains. “I’m learning about every facet of being a software engineer and I’d like to be doing that firsthand.”
He thinks he might eventually use AI tools to augment his work and automate some tasks, but only after he’s built a deeper understanding of his profession.
Scott describes how his BCIT education helped him understand how the artificial intelligence models work, and to see their limitations: “In terms of AI coming for our jobs? I’m not concerned about that, at least not yet.”
Researcher Aaron Hunter, who is the Mastercard Chair in Digital Trust at BCIT, argues that powerful, productive AI tools will change the nature of digital work—but he’s optimistic for employment in the sector.
“Maybe there you don’t need as many of the software developers doing the same tasks that they were doing before,” he explains. “But, overall, we’re still predicting growth in the industry, because all of these AI tools still need people to work on them, manage them and develop them.”
Hunter says BCIT works with industry partners like Fujitsu, Amazon and local employers to ensure its graduates have skills useful to the sector. The school surveyed and consulted with those businesses as part of its preparation to launch its new master of science in applied computing degree in September.
The program won’t produce programmers, but rather what Hunter calls technical leaders. Businesses told him they needed mid-level professionals who can manage tech projects, not just perform the basic tasks. “It’s somebody who’s supposed to come in and lead the development of a product or lead the development of a team,” he elaborates. “They need to be able to direct and organize software development.”
AI tools aren’t just changing technology development—they’re also changing business operations. Shauna Begley heads BCIT’s business information technology management diploma program. It’s been offering an artificial intelligence management option since fall 2020. She says the diploma teaches students technical and analytical knowledge, along with business strategy.
Companies want graduates from programs like hers who can, for example, use data scraping tools to gauge consumer sentiment of their products. More importantly, they want to put that information to good use. Begley says organizations are looking for people who can answer a key question: “How do we leverage this in a way that is going to add value in our business?”
AI began offering powerful business analytics tools years before LLMs burst into public consciousness. Zubair was digging into her toolkit as soon as she joined the workforce in Pakistan after her undergraduate degree. Her employer in 2020, Lotte Akhtar Beverages Ltd., was trying to figure out why Milkis, the popular Korean soda drink it introduced to the local market, was struggling to catch on among Pakistanis. She scraped data from Facebook, the main platform her compatriots used for reviews, and ran an analysis using a branch of artificial intelligence called natural language processing, or NLP.
“I extracted high-frequency and high-relevance words,” Zubair says. “I discovered that people were using ‘milk’ a lot, and they were comparing the taste to yogurt.”
Milkis is sparkling and delivers a cream soda feel to your mouth. But Pakistanis were expecting something similar to a regional drink called doodh soda, which people make by mixing milk with equal parts Sprite or 7Up.
Zubair’s team investigated and found the white-and-blue packaging was confusing consumers into thinking Milkis was a milk-based drink. When they changed the colours, it changed people’s feedback on the taste. “It was finally moving away from milk-based to soda-based—like fizzy and young,” Zubair reports.
Chunhua Wu, assistant professor and division chair at Zubair’s Master of Business Analytics program at Sauder, says the 2012 deep learning revolution in the field of computer vision gave marketers new ways of looking at data. That was the year, among other breakthroughs, an experimental Google computer network taught itself to recognize cat images by watching YouTube videos. No humans taught or hinted to the network what a cat was, or what one looked like.
That ability—to learn and extract information or patterns from raw data without extensive human intervention—differentiates deep learning from traditional machine-learning techniques. It makes training machines vastly more scalable, because computers can now comb through massive, unstructured data sets without requiring people to, say, manually label millions of cat pictures. LLMs are deep learning models trained on astronomical amounts of language data—about 5 trillion words, in the case of ChatGPT.
Wu says marketing researchers like him started adopting deep learning models around 2013. “It’s incorporated in, for example, understanding consumer preferences toward brands, brand positioning or their social media strategies,” he explains. “How are brands communicating their positioning or delivering their message, and how this is perceived by the users or consumers?”
He describes AI as a multiplier of business analysts’ skills. The emergence of LLMs can also empower them to focus more on analysis, and less on writing software. He says about 40 percent of his MBAN students arrive with some coding experience. But he’s exploring how LLMs can accelerate that side of their work.
“For example, in this course, we play around with data a lot,” Wu explains. “So I teach them, in three days, how to use software to create visualizations. And on day four, we build a comprehensive data dashboard on Airbnb marketing in Vancouver.”
That’s the traditional, piece-by-piece way of highlighting patterns in data, to try to understand it, he says. But then he takes an afternoon and tells the class: “Let’s forget about the traditional software. Now you have the data. It’s just a CSV file. Now let’s chat with these AI tools and see what insights you can get—whether you can also create a dashboard.”
Wu reports that some students were able to get code built by ChatGPT and get interactive dashboards running within 20 to 30 minutes.
So when Wu talks about AI being a multiplier, he means it can multiply a user’s proficiency or productivity. A tool might triple a low-skilled worker’s output from one project per day to three, and a high-skilled worker’s production from five to 15. Or it can multiply the quality of their work. Either way, he contends, the effect of AI on human capital is much more differentiated at higher skill levels. But, he adds: “If your original understanding of analytics, for example, is zero, then AI wouldn’t help you at all—because it’s hard.”
Researchers like Wu quickly applied earlier deep learning models to analyze data, but their usefulness in other business realms remained limited—for a time. Eventually, innovators began learning to use these analytical tools as creative tools as well.
In 2022, Nike gathered archival footage of tennis great Serena Williams from 1999 and 2017 and processed the data through machine learning to simulate 130,000 games between a 17-year-old Williams and the champion at 35 years old. The company used these simulated matches, and the videos it created of some of them, as the centrepiece of its award-winning 50th anniversary advertising campaign.
ChatGPT’s public release shortly afterward marked a jump in the power of deep learning models, and a tipping point where they exploded into mainstream usage. Models that can write and speak, like ChatGPT, and models that can produce fantastic photo-realistic images, like Stable Diffusion, are collectively known as generative AI.
Companies in all manner of sectors are employing generative AI for an ever-expanding range of tasks. Expedia Group, for example, incorporates ChatGPT into its app to help travellers plan trips.
Luana Carcano, academic director of undergraduate programs at SFU’s Beedie School of Business, describes how generative AI is starting to transform the fashion industry—a keystone of business and culture in her native Italy. “Creativity is really the heart and soul of fashion, and AI can help, because it could be a brainstorming tool for creative people,” she points out. Designers can tell computers to turn verbal descriptions or images into designs or illustrations. They can experiment with design variations quickly and cheaply.
In the fast-fashion side of the industry, where brands like H&M produce about 20,000 different styles annually, computing power will accelerate design and production, Carcano predicts: “It will help them to automatize some processes, but also to analyze the big data that they have and create patterns faster than humans can do.”
Deep learning is already helping businesses gain better insights and create new services and products. So what about the future? Royal Roads professor Mark Lokanan is creating AI models to predict it. Specifically, he’s working to create models that comb through data to detect warning signs of financial crimes like fraud. But he can use the same methods to look at many other areas.
“We are basically building models to predict any type of financial crime,” he explains. “So I use them to predict money laundering transactions, mobile fraud transactions, securities fraud transactions and accounting fraud transactions. We’re also using these same models to predict turnover in HR for companies, credit defaults, loan defaults and various other domains.”
Lokanan, an up-close observer of AI’s growing power, offers a bigger-picture prediction for the future: these machines won’t replace people in their jobs.
“You cannot replace the human investigative skills in a fraud case,” he stresses. “You cannot replace the HR expertise you need. More and more, we are seeing that we need more domain expertise. And when I say domain expertise, I mean real qualifications.”
Why?
“Remember, AI can’t operationalize itself,” Lokanan elaborates. “AI can produce for you the insights that you need to make decisions. But for a fraud case, you need the accountants or the internal auditors to look at these insights.”
Lokanan’s view sheds some light on a paradox in Statistics Canada’s 2024 report on AI adoption by Canadian businesses. Nearly 40 percent of companies surveyed that used AI reported that AI reduced tasks previously employed by employees by a “moderate or large extent.” Yet 84.9 percent of those AI-adopting companies reported no change in their head counts after adoption, and only 6.3 percent reported a reduction in the total number of employees.
Lokanan argues that technology—like, for example, spreadsheet programs or accounting software—has historically freed people from tedious tasks to focus on work that requires decision-making and judgment.
“AI is not going to take your job,” Lokanan says. “It’s probably going to automate about 20 to 30 percent of your job, but you have to adapt.”