Building the future workforce with AI

This conversation about AI and jobs is happening everywhere right now. Most of it isn't very useful. It oscillates between breathless hype and existential panic, with very little space in between for the kind of honest, grounded thinking that leaders actually need to make decisions this quarter.

So we sat down with Adriana Zubiri — twenty-plus years inside Canada's largest banks across retail and capital markets, now an advisor helping organizations navigate exactly this — and asked her to separate what's real from what's noise.

Her answer is more nuanced, and more useful, than what's been dominating the discourse. AI will absorb tasks before it absorbs roles. The timing depends on your data maturity, your governance posture, and your risk appetite. And the slowdown in adoption everyone's noticing right now isn't a grace period — it's a window. What you do with it will define the next decade of your organization.

Ask "will AI replace my job?" and you'll get either reassurance or doom, depending on who's selling. Ask Adriana, and you'll get something more useful: a reframe.

"It's not the profession that's at risk. It's the tasks within the profession." Basic coding gets absorbed. Senior architecture doesn't. Routine documentation goes. Judgment-heavy review work stays — and gets harder, because the volume of AI-generated artifacts that need review is exploding. The shift is graduated, not binary.

What's keeping the timing slower than the hype suggests? Three things, mostly: data maturity (most enterprises aren't where they need to be), governance posture (the regulatory frameworks are still catching up), and risk appetite (no CIO is letting AI ship mission-critical code unsupervised — yet). The gap between what's technically possible and what's organizationally acceptable is wider than it looks from the outside.

That gap is the window. And it's closing.

We asked Adriana what skills she'd prioritize if she were building a technology team today. She reframed the question before answering it.

"It's not about the technology team. You have to think broader." The five capabilities she named aren't tools or certifications — they're things machines are categorically bad at, and they apply to every function in the building.

At literacy across the whole workforce. Not just the tech team. The most important AI initiative at most companies isn't a technology project; it's a literacy program. Data fluency. No data, no AI. Bad data, bad AI. Most organizations are still figuring out the basics. Knowing knowing how to complete a task is increasingly the machine's job. Knowing why the task matters and how it connects to everything else is still very much a human job. Human-AI interaction design. The discipline of building workflows where humans, automation, and AI agents cooperate well — barely a decade old, and rapidly becoming critical. Communication with two audiences. Humans need nuance, context, relationship. AI needs precision, specificity, structure. The people who can do both fluently have an advantage that compounds.

The thread that connects all five: the business-technology divide is collapsing. It started in capital markets, where the technology was the work, but it's spreading everywhere. "AI has made it anyone's game," Adriana said. The mechanic doesn't need to be a great driver — but the driver who understands how the car works is going to take it places the others can't.

The pattern isn't tech jobs at risk versus non-tech jobs safe. It's basic-task roles at risk versus judgment roles growing. Basic coding, basic documentation, basic project coordination, basic content creation — these are the categories being absorbed. Not because the professions are disappearing, but because the tasks that justified the entry-level versions of those roles are being automated.

What grows in that environment? Anything that requires end-to-end thinking, architectural judgment, and the ability to manage systems with both human and machine components. And two areas Adriana flags as growing faster than most realize: AI governance and risk, and cybersecurity through an AI lens. Both are underprepared. Both are urgent.

AI governance and risk. Most organizations have a policy. Few have the talent to execute it. Cybersecurity through an AI lens. Every capability AI unlocks for legitimate use also arrives as a new attack surface — deepfakes, AI-generated phishing, synthetic identity. These aren't future risks. They're current ones, and the people who understand them are scarce.

Five years ago, almost nobody had a job title with "AI" in it outside of research labs. Today, the roles that will define the next decade are being invented in real time — and most organizations aren't hiring for them yet. We asked Adriana to name the ones she sees coming. She came with four.

Prompt Engineer. The translator between LLMs and the business. Needs both deep AI literacy and domain expertise — a rare combination that tends to get shared across entire organizations. AI Auditor. Not a tester. An auditor. Someone who validates that the model works correctly, fairly, and in alignment with the business outcomes it was built to produce. Adriana's instinct: don't hire from scratch. Take your best existing auditors and invest in their AI literacy. Process Designer. The most dangerous thing you can do with AI is automate a broken process faster. The new process designer doesn't just document workflows — they redesign them from the ground up for a world where humans, automation, and AI agents are all in the loop. AI-native UX. Who designs the experience between a system and an AI? Or between two AIs? It's a new discipline, and the designers who figure it out first will be invaluable.

If you're leading a team right now, the pressure to "do something" with AI is real. The question is whether you're making that decision from fear or from strategy. Adriana's framework is one of the most practical things she shared, and it starts with a number: 70/30.

Keep 70% of your workforce. Institutional knowledge — how your organization actually works, the informal networks, the hard-won context — is not on the open market. You can't hire it back once it walks. Bring in or develop 30% of genuinely new capability. The goal isn't to replace the team. It's to amplify it.

The adoption tactics that actually work don't start with the tool. They start with a use case everyone already shares — email drafting, meeting notes, something universal — and let people feel the time savings firsthand. Once someone experiences it once, the question that follows is always: what else can this do?

And the super user in your organization probably isn't who you think it is. At Adriana's bank, it was the EA — no formal training, no technical background. She figured out that the LLM could help her write award nominations, summarize documents, draft communications. She became the team's informal AI champion almost immediately. Super users aren't the most technical people. They're the people who understand their own jobs best.

We ended our conversation with Adriana on the personal question: what do you actually say to people at different stages of their careers? Her answers were specific. They were also short.

If you're in school: the tools will change. Critical thinking won't. Build the fundamentals strong, build the learning habit stronger. The degree is a starting line. If you're early in your career: you won't be replaced by AI. You might be replaced by someone who uses AI better than you do. Don't learn AI in general — learn it specifically, for your role, for the actual problems you're solving. If you're managing a team: your job is no longer assigning tasks. It's managing outputs from a workflow that includes humans, agents, and automation working together. Less task assignment. More system design. If you're an executive: the mindset shift is the whole game. Not "how do I cut costs," but "how do I amplify what this team can produce?" Different question, different culture, different organization five years from now.

We almost didn't include this one. Then we decided it was too good to leave out. Adriana recently completed a university course on AI and the Fourth Industrial Revolution, and when she maps the current AI moment onto what happened during industrialization — specifically in the British cottage industries — something clicks.

At the start, machines created new dependencies and new jobs absorbed the displaced. Then the machines accelerated. Other industries couldn't keep up. And people had nowhere to go. The lesson wasn't that machines were bad. It was that how you manage the transition determines who benefits and who doesn't. The regulatory frameworks. The worker protections. The investment in what comes next.

We're at that fork in the road right now. The outcome isn't written.