VOL 01Spark AI · Advisory Intelligence

What AI actually does to your career

A conversation with someone who's been inside the enterprise machine — and has the context to separate the real from the noise.

FeaturingAdriana ZubiriSenior Technology Executive · Spark Advisor

9 minute read

27 video clips

6 sections + bonus

In conversation

There's a version of the AI conversation that lives in fear. Another that lives in hype. Adriana Zubiri has spent three decades at the intersection of enterprise technology and human capital — across the retail and capital markets floors of major Canadian banks, on AI panels at the University of Toronto, and in countless conversations with the leaders and teams navigating this shift in real time.

What she'll tell you is that neither version — the fear nor the hype — is particularly useful. The more honest conversation is about evolution. It's about the gap between what AI can actually do today and what people think it can do. It's about a junior developer who can't learn by doing because AI is doing it. It's about an executive assistant who significantly improved her productivity and redefined her role — not because she's technical, but because she understood her own work well enough to see what the machine could take off her plate.

That's the conversation Adriana sat down to have with us. And it starts, as most good conversations do, with the question everyone's afraid to answer honestly.

01 Section One

The hype, the fear, and what's actually happening

Yes, AI will replace some jobs. That's almost settled. But the mechanism, the timing, and the nuance of which jobs, in which ways, and when — that's where the popular conversation falls apart.

Adriana's answer to "is AI replacing jobs?" isn't a yes or no. It's a framework. The jobs most at risk aren't entire professions. They're the tasks within those professions — the routine, the repetitive, the cognitive work that doesn't require judgment. Entire roles don't vanish overnight. They hollow out from the inside.

But here's the thing almost nobody is talking about: the data isn't ready. The governance frameworks aren't there. The regulatory structures don't exist yet. For regulated industries especially — banking, insurance, healthcare — these aren't minor obstacles. They're significant hurdles to wholesale AI deployment.

That's not a reason to slow down your own learning. It's exactly the opposite.

The slowdown in replacement isn't a grace period. It's a window. The question is whether you're using it.

Adriana Zubiri

Nowhere is this tension sharper than in software development. A year ago, developers' jobs looked completely different. Today, senior developers are orchestrating AI agents that produce code, reviewing pull requests written in plain English, validating outputs that a system built in seconds. Their experience is more valuable than ever. But the junior developer waiting behind them? That's where the real, uncomfortable question lives.

How do you become a senior developer if you never had to do the work that made you one? How do you validate AI output if you've never written the code yourself? The calculator analogy holds: you can use one without knowing arithmetic, but you're flying blind on the fundamentals. And in enterprise code at scale, blind spots compound.

02 Section Two

The skills that actually matter now

When we asked Adriana what skills she'd prioritize if building a technology team today, she did something interesting: she reframed the question.

"It's not about the technology team. You have to think broader."

What follows is one of the clearest frameworks we've heard for thinking about human capital in the age of AI. It's not a list of certifications. It's not a list of tools. It's a set of capabilities that, once you hear them, feel obvious — because they're precisely what machines are worst at.

AI is bringing to the table something entirely new — the ability for everyone, without being technical, to interact with these tools. They just need to know English. Or Spanish. Or whatever language they speak.

Adriana Zubiri

The five pillars Adriana identifies are worth committing to memory: AI literacy across the entire workforce. Data fluency — no data, no AI; bad data, bad AI. System thinking, meaning the ability to understand end-to-end workflows rather than isolated tasks. Human-AI interaction design, which is about understanding where the handoffs happen. And communication — not just with other humans, but with precision when prompting the machines.

Then there's the convergence question. Is the line between "tech team" and "business team" disappearing? Adriana's answer is nuanced. She draws on her own experience moving from retail banking into capital markets, where the business side was deeply technical — because the technology directly drove their decisions. That model, she says, is spreading.

"If the business isn't savvy about what technology can do, they're not going to know what they need. They'll ask for faster horses, not knowing there's an automobile."

And for anyone who grew up after 2000? This won't even be a conversation worth having. They arrived already fluent.

03 Section Three

Who's exposed — and who becomes indispensable

Not all roles are created equal in the AI age. The pattern Adriana describes isn't "technical roles at risk." It's task-specific roles at risk. Basic coding. Basic documentation. Basic content creation. Basic project management. The keyword in all of those is basic.

The roles that become more valuable (not less) in an AI-integrated environment tend to share a common trait: they require architecture, orchestration, and judgment across the whole system. The developer who can design and manage AI agent workflows isn't the same as the developer who writes individual functions.

And then there's the area almost nobody is discussing with enough urgency: AI governance and risk. In regulated industries, the question isn't just "can we deploy AI?" It's "can we explain to a regulator what the model is doing, why it made a particular decision, and whether its outputs are biased?" Right now, the answer for many organizations is: not really.

If a regulator came to a bank today and asked, "How do you know your model isn't biased?" — you need to be able to answer that question. This is something that needs to be on every executive's mind.

Adriana Zubiri

That governance gap creates a talent gap. And that talent gap creates opportunity for the people who build those skills now, before the regulations arrive. Because the regulations are coming.

Cybersecurity is the other category accelerating under AI pressure. Deepfakes, AI-generated phishing, synthetic identity fraud — every new capability AI brings to legitimate use also arrives as a new surface for attack. The people who understand security through an AI lens are not the same as the people who understood it in 2018.

04 Section Four

The roles being born right now

We asked Adriana to get out her crystal ball. She declined — and then gave us one of the most concrete answers we've heard from anyone on this topic.

The roles emerging aren't science fiction. They're logical extensions of work that already exists, reshaped by new tools and new complexity.

The Prompt Engineer is the translator between language models and the business. This person needs deep understanding of how LLMs think and process — and enough business domain knowledge to shape prompts that actually solve real problems. Adriana's seen this play out: the people genuinely good at it tend to be deployed across entire organizations, helping every business unit develop the right prompts for their workflows.

The AI Auditor goes further than testing. A tester validates that something works. An auditor validates that it works correctly, fairly, and in alignment with the business outcomes it was designed to produce. Adriana's instinct: take existing auditors with deep business process knowledge and elevate their AI literacy. The institutional knowledge of what good looks like is not teachable from scratch.

The Process Designer is perhaps the most underappreciated role on the list. Today's process designers document and optimize human workflows. Tomorrow's will need to understand where AI agents, automation, and human judgment each belong — and when to scrap the entire process and rebuild it for what AI can actually do. Automating a broken process just breaks faster.

You don't want to automate a bad process at AI speed. You need to rethink the process first — and that requires someone who can see the whole system.

Adriana Zubiri

And then there's the evolution of User Experience design — the role that surprised us most. Before AI, UX meant designing the experience between a human and an interface. Now, the interaction surface has multiplied: human to AI, AI to system, system to outcome, AI to AI. Who is responsible for designing the experience of each of those handoffs? It's a new discipline — and it lives at the intersection of the creative, the technical, and the philosophical.

05 Section Five

The leader's actual playbook

If you're leading a team right now, the pressure to "do something" with AI is real. The question is whether you're making decisions from fear or from strategy.

Adriana's framework is simple and worth writing down: think 70/30. Keep 70% of your workforce — because institutional knowledge is not available on the open market, and any leader who treats it as interchangeable with new hires will pay for that mistake. Bring in 30% new, or reskill adjacent. Inject the capabilities you genuinely can't grow internally.

A smart leader does three things: elevate AI literacy across the entire team, not just the technical seats. Reskill adjacent roles where you can extend rather than replace. And inject new talent — not to replace the team, but to accelerate the learning of the people already there.

My EA was one of our most productive users of the LLM from day one. She found so many things she could do with it — writing award nominations, reviewing decks, summarizing, taking notes. Nobody taught her. She understood her own job well enough to find every angle.

Adriana Zubiri

The super user insight is one of the most practical things Adriana shares. In every organization that has successfully raised AI literacy, the spread didn't happen through training decks. It happened through people who figured out what the tool could do for their specific job, and then became informal ambassadors for it.

The adoption tactics that work are not "here's an LLM, go figure out what to do with it." They're: here's one use case you all share. Let's do it together. Once people feel the time savings once, the natural follow-up question is always: what else can this do?

Nobody wants to learn a new tool. But everybody wants to be more effective. Those are different entry points, and the second one is the one that actually works.

06 Section Six

Wherever you stand right now

We ended the conversation where most good ones do: with the personal. Because ultimately, the AI transition isn't an abstract organizational challenge. It's something every person — whether they're in their first year of school or their twentieth year in management — has to navigate for themselves.

Adriana's advice shifts for each stage of a career. But there's a thread that runs through all of it: don't wait for someone to hand you a plan. The people who come out of this well aren't the ones who moved the fastest or had the most AI certifications. They're the ones who understood their own work deeply enough to know what AI could take from them, and what only they could do.

If you're still in schoolThe tools will change. Critical thinking won't. Build the foundations of your discipline first, and build the mental model that learning doesn't end at graduation.
If you're early in your careerYou won't be replaced by AI, but you may be replaced by someone who uses AI better than you do. Know the tools in your field — not in general, but specifically in your role.
If you're managing teamsYour job is no longer assigning tasks. It's managing outputs, from workflows that may include humans, agents, and automation working together. That's a different skill.
If you're an executiveThe mindset shift is the whole game. It's not "how do I cut costs with AI?" It's "how do I amplify what this team can produce?" Those questions lead to very different organizations.

★

Bonus Section

★ The Long View

AI vs. the Industrial Revolution

We saved this one for last, because Adriana is, among other things, a history buff who recently completed a university course on AI and the Fourth Industrial Revolution. And when you see the AI transformation through that particular lens, something shifts in the conversation.

The parallels are uncomfortable. At the very beginning of the Industrial Revolution, displacement was manageable. Early machines created new dependencies, which created new jobs downstream. But as mechanization accelerated and other industries failed to absorb the displaced workers — the cottage industries in Great Britain collapsed. There was no next job. Just people with nowhere to go.

The lesson isn't that the Industrial Revolution was a mistake. In the long arc, it clearly wasn't. The lesson is that how you do it matters. The machines weren't the problem. The absence of regulatory frameworks, worker protections, and social infrastructure was the problem. The factories we consider normal today (safe, regulated, humane) didn't exist at the start. They were built, over decades, through painful experience.

We haven't built those structures for AI yet. The regulatory frameworks are nascent. The worker protections are undefined. The "new industries" that absorb the displaced aren't visible yet. That doesn't mean they won't emerge. But it does mean the outcome isn't automatic.

This isn't pessimism. It's exactly the kind of long-view thinking that distinguishes a strategist from a spectator, and an advisor from a vendor.

We need to think about how we want AI to coexist across all the different jobs — and how we're going to control that. We cannot leave it laissez-faire and just see what happens.

Adriana Zubiri

— and finally —

"The people who come out of this well aren't the ones who moved the fastest. They're the ones who understood what they brought to the table — and what only they could do."

Adriana Zubiri

Spark
Advisor

About the advisor

Adriana Zubiri

Senior Technology Executive · Spark Advisor

Adriana spent more than two decades inside Canada's largest financial institutions, leading teams across retail banking and capital markets through some of the most consequential technology transitions of the last twenty years.

She now advises organizations on workforce strategy, AI integration, and the human side of transformation. She recently completed a university program on AI and the Fourth Industrial Revolution — for fun.

Found this useful?

Spread the word

share it with someone who'll get it