Nicole Shokoples was talking about talent when she stopped mid-conversation to bring up something she’d been reading about: Orchestras.
The former vice president at Alberta Innovates had just wrapped two years with a front-row seat to the province’s tech ecosystem. She started a new role helping place women in tech and the orchestra example fit.
In the 1970s, major American symphony orchestras came under pressure over how they selected musicians. The process lacked any mechanism to ensure selection was based purely on what someone could play. To fix it, orchestras placed a physical screen between the performer and the panel. Candidates performed from behind it, identity concealed, and the panel heard only the music.
That move changed who made the cut.
Researchers Claudia Goldin and Cecilia Rouse found that the share of women in top U.S. orchestras grew from roughly 6% in the ’70s to about 25% by the mid-90s.
The talent was always there. The process just wasn’t finding it.
Technology leaders building AI functions in 2026 are living a version of the same problem. McKinsey Global Institute found that the number of workers in roles explicitly requiring AI fluency rose from about 1 million in 2023 to 7 million in 2025.
Most organizations are responding to the AI talent gap the way organizations usually do: by looking harder through the same channels, posting the same roles with the same requirements, and reaching out to the same networks.
After talking with Shokoples and digging into the research, the evidence points in one direction. The companies pulling ahead now on AI talent are not searching harder. They’re changing what the search can see, and finding people the standard process was never built to find.
Shokoples was on her final day at Alberta Innovates when we spoke, and Monday morning this week she started as chief operating officer at Toast.
Toast is a Canadian recruitment platform that connects women in tech with roles across North America while working with hiring leaders to rethink how they build teams. The company launched in 2022 and today works with more than 125 employer partners in Canada and the United States.
Shokoples and CEO April Hicke worked together years earlier at ATB Financial and Shokoples has followed the company since it launched.
“You can’t really have the talent conversation without having the conversation that we’re having at Toast,” she said in an interview from her home office in Calgary. “If you want to build really good products, experiences, and businesses that serve a diverse audience, you need a diversity of perspectives to build them.”
Toast operates on the same principle the orchestras discovered, that changing what the process can see changes who gets through.
The company removes identifying information from candidate profiles before they reach hiring managers, and clients report that the share of women in candidate pools moved from 11% to 36% in a single month as a result.
For technology leaders building AI teams, that single change in process points to a structural problem many organizations have not yet diagnosed.
The partnership gap between tech and HR leadership
A 2025 survey of 500 CIOs and 1,000 HR leaders, found that 89% of CIOs agree the HR and IT partnership is key to attracting top talent, but only 27% believe talent acquisition itself has a significant impact on business outcomes. It’s not that they think people don’t matter. It’s that they don’t see the hiring process as a lever they should own, so they hand it to HR and move on.
That is a business problem. It surfaces when the pipeline produces the wrong candidates and nobody can explain why.
Talent acquisition is a strategic lever, not an administrative one. When tech leaders treat it as the latter, handing it to HR feels logical, but HR doesn’t always have the technical fluency to know what AI roles actually require so candidates clear the filters and arrive without what the team needs.
From there, technology adoption outpaces process change, organizations get stuck in the messy middle of AI maturity, unable to scale, with pipelines producing the wrong candidates, teams struggling to execute, and AI programs falling short in ways that are hard to diagnose.
Then the finger-pointing starts.
Technology leaders see HR as a bottleneck, and HR sees technology leaders as making unrealistic demands. Each function ends up managing the consequences of decisions made by the other.
Hicke sees it consistently among the employers Toast works with.
“The conversation between the tech leader and the people leader often doesn’t happen because nobody’s sure it’s their lane,” she said.
By comparison, the organizations pulling ahead have changed that dynamic structurally.
Among AI-leading firms, the majority measure executive teams based on AI-specific metrics tied to innovation, adoption, and business impact, according to a report from Eightfold. When leaders across functions are measured on the same AI outcomes, the question of whose job talent strategy is tends to answer itself.
Your job description might be filtering out the people you need
The requirements written into a job posting determine who applies. Who applies determines who gets hired. Who gets hired determines what gets built.
HR brings the process to reach a broader candidate pool, and the technology leader brings the authority to rewrite what the role actually requires. Either function working alone on the talent question leaves the other half unsolved.
The way those requirements are written has a disproportionate effect on who applies in the first place.
Harvard Business School researchers found that a single degree requirement causes almost two-thirds of qualified workers to never apply, because they see it and don’t submit.
Organizations that remove degree requirements see modest gains, but the pipeline stays narrow if the rest of the job description stays the same. Research from the Future Skills Centre shows then a man with four years of experience will look at a job posting that requires 10 years experience and apply anyway, whereas a woman with the same background is far less likely to.
Organizations have known about this gap for years and have tried to close it through diversity training, hiring tests, and grievance systems. But it hasn’t fixed the problem.
A 2025 Harvard Business Review analysis examining decades of evidence found those interventions failed to move representation numbers and in some cases reduced them. Every one of them started too late, after the job description had already determined who would apply. The criteria had already done the filtering before any program began.
Women’s representation in technical roles in Canada has held at roughly 21% for a decade, a number that has not meaningfully moved despite years of effort.
The AI workforce is the direct product of those accumulated decisions.
Women make up just 22% of the global AI workforce, according to February 2026 analysis by Fast Company. In most organizations the job requirements that produced that number carry forward from one hiring cycle to the next without review.
When those requirements get embedded in AI recruiting tools as training data, the tools learn to find what the data shows has always been hired, and the pattern compounds.
The systems those teams build carry the same blind spots the teams do.
A March 2026 study that included interviews with AI professionals found the pattern repeated. For example, computer vision systems built without diverse perspectives consistently misread dark-skinned individuals.
“If you keep hiring the same guy, the same man from University of Waterloo, over and over and over again, you’re not going to get a different way to solve a problem,” says Hicke.
High-performing AI teams consistently hire for varied skill sets, and cross-functional teams are significantly more likely to report gains in efficiency and innovation.
What technology leaders should do to build better AI teams
Before making any changes to how they hire, a technology leader needs to ask if the requirements written into our AI roles reflect what those roles actually need, or do they just reflect what this sector has always hired for?
A LinkedIn report says organizations that shift to hiring based on skills rather than credentials can expand their qualified AI talent pool by 8.2 times. That suggests the shortage most technology leaders are experiencing is not a market problem but a requirements problem.
Hicke’s ask is to understand the current state of your own pipeline before making any changes.
Find out where candidates are dropping out. Determine whether technical assessments are filtering for capability or for familiarity. Look at what internal promotion data shows about who advances and who plateaus, she says.
“Your candidate pool should reflect the community that you’re serving. If you don’t have a full view of this, this is a time when you can’t just go off intuition. You need data to back it up.”
Shokoples joining Toast is also a reflection of that. Shokoples studied sociology and criminology, disciplines built around spotting patterns in systems and asking why interventions fail to change them.
She spent two years as a vice president at Alberta Innovates watching the same thing repeat across the province’s technology ecosystem, and it’s part of what drew her to Toast, where that instinct for structural problems is exactly what the role requires.
“The tech ecosystem can be an echo chamber. A lot of the same perspectives exist. It’s not anyone’s fault. It’s just a lot of proximity, culture, and the result of how tech has grown.”
She brought people with neurodivergent perspectives into teams specifically because they thought differently.
“We got some super creative solutions,” she said. “You get wildly different solutions when you don’t have everybody who thinks the same.”
It starts with rethinking what the role really requires, and with building a closer partnership with HR. The organizations doing this well have stopped treating talent as a people question and started treating it as a product question.
The orchestras didn’t solve their problem by training judges to be less biased or by running better onboarding programs. They changed what the process could see, and the people who had always been there started getting through.
Final shots
- Women’s representation in Canadian tech has held at 21% for a decade despite years of effort. Every intervention that failed started after the job description had already done the filtering. The specification review is the only place to get upstream of that problem, and it belongs to the technology leader.
- Organizations leading on AI have built a strong partnership between their technology and HR functions. Those in the earliest stages largely have not. The gap between them is not mainly a technology gap. It’s a talent strategy gap, and the organizations that close it first will have built something their competitors are still looking for.
- Shifting to skills-based hiring can expand a qualified AI talent pool by more than eight times. Most technology leaders experiencing a shortage are treating it as a market problem. The data suggests it’s a requirements problem, and requirements are something they can change today.
