The BA Job That AI Is Killing — and the One It's Creating

Ivan Ureña-Valdes was in the middle of interviewing a job candidate when the email arrived. It was from Jack Dorsey. In the time it took Ivan to read it, four years at Block ended. He had survived three previous rounds of layoffs — not despite his performance, but because of it. This time, Dorsey's letter to shareholders named AI as the explicit reason for cutting 4,000 people, nearly half the company's workforce. Ivan had to stop the interview and tell the candidate he'd just been laid off and couldn't finish their feedback. "I'm the sole provider for my family," he said later. "It was tough."

If you work as a Business Analyst — or in any role where your value is turning information into decisions — that story probably landed somewhere specific in your chest. Because the question underneath it isn't really about Ivan. It's about whether what happened to him is a preview of your future or a warning you can still act on.

Here is the honest answer: both, depending on what you do next. The same month Ivan lost his job, a BA named Ahmed Squalli Houssaini sat down with 48 hours to prepare for a meeting with wealth management executives he'd never met, in a domain he'd never worked in — NAV calculations, wash-sale rules, sub-advised accounts. He faced the same wall of fear Ivan hit. But he came out the other side of it with his career intact and accelerating.

These two stories aren't opposites. They're a map. Before you can figure out which path is yours, you need to understand something most AI coverage gets wrong: the threat isn't AI replacing your job. It's AI replacing specific tasks inside your job — and the line between safe and exposed is a lot more precise than anyone has told you.

AI Has a Frontier — and It Runs Right Through Your Job Description

Think of AI capabilities as a landscape with a jagged edge. On one side: tasks where AI makes you measurably faster and better. On the other: tasks where using AI actually makes your output worse. For Business Analysts, that line runs directly through the middle of the job description.

A 2026 peer-reviewed randomized controlled trial published in Organization Science tested exactly this with 758 management consultants. For tasks inside AI's capabilities — drafting structured outputs, synthesizing information, generating first-cut analysis — participants using AI were more than 25% faster and produced work rated more than 30% higher in quality. Those are real, material gains. But for tasks outside the frontier, the results flipped. Participants who used AI were 19 percentage points less likely to produce correct solutions than those working without it. They over-relied on the tool precisely where it couldn't help them.

The IIBA's 2025 Global State of Business Analysis Report captures the same split from the other direction: 74% of BAs say AI positively impacts their careers, but the same report flags communication, critical thinking, and strategic judgment as more critical than ever — precisely because AI is handling the work that required neither.

So what does the frontier look like in practice? Tasks moving to AI this year include first-draft user stories generated from meeting transcripts, meeting summaries and action items, basic data pulls and standard report formatting, and status update emails. These are real and the shift is already underway.

Tasks becoming more valuable are different in kind, not just degree. They include stakeholder negotiation when the CTO and CFO disagree on scope. Catching regulatory edge cases the model didn't know to look for. Deciding which of four competing priorities gets built when the budget only covers two. Reading the room. Knowing when a feature should be killed despite three months of scoping. These are judgment tasks, and AI applied to them carelessly produces worse output, not better.

You lack something? Learn it. Use it. Fast.
by Ahmed Squalli Houssaini, Senior Technical Business Analyst

This is where Ahmed's story pays off. He walked into Wednesday's meeting and asked a question about fee-billing reconciliation logic on sub-advised accounts that made the lead architect pause. "Good catch," the architect said. Ahmed had known what the term meant for 19 hours. He'd used AI to compress months of domain onboarding into an afternoon — not to replace his judgment, but to get him close enough to the frontier that his judgment could operate. His developer later wrote: "When Ahmed writes a spec, the data types match the schema, the error codes are real, and the edge cases actually reflect how the system behaves."

The frontier applies beyond formal BA roles. If your job involves turning information into decisions for other people — whether you're in HR, operations, marketing, or financial planning — the same divide exists. The automatable side is wherever you produce structured artifacts from known information. The elevated side is wherever you navigate human ambiguity that no training data can fully model.

What's Actually Driving the "AI Layoffs" — and What Isn't

Understanding the frontier helps explain Ahmed's success. It doesn't fully explain Ivan's loss — because Ivan knew AI was automating his work, watched it happen in real time, and still lost his job. To understand that, you need to look honestly at what's behind the wave of layoffs being attributed to AI right now.

Gartner tracked 1.4 million layoffs in 2025 and found that less than 1% were directly attributable to AI productivity gains. What's actually driving most "AI layoff" announcements is a combination of pandemic-era overhiring corrections, margin pressure, and investor appetite for AI-forward narratives — a phenomenon with its own name: AI-washing.

Ivan's account is instructive here. He watched AI compress his routine data-sourcing and querying tasks dramatically, describing it as a "whoa" moment when he realized how powerful the tools had become. That part was real. But he was also inside a company that had roughly tripled its headcount between 2019 and 2022, whose stock had fallen sharply, and whose CEO publicly acknowledged the company had "over-hired" during COVID. Both things were true simultaneously. AI accelerated the devaluation of tasks that were already becoming commodity work. The financial incentive to act on that acceleration already existed. The result was Ivan.

This is the honest picture, and it's more useful than either the doom narrative or the dismissive one. AI is genuinely automating specific BA and analyst tasks. And many "AI layoffs" are financial restructurings dressed up in AI language. Knowing the difference matters because the real threat and the real opportunity are both more specific than the headlines suggest.

The goal isn't to deploy an algorithm — it's to solve a real business problem, drive value, and align with strategic objectives.
by Sid Arya, President, IIBA Vancouver Chapter

Meanwhile, the hiring market is sending a different signal entirely. Despite overall US job postings sitting only about 6% above pre-pandemic levels as of early 2026, postings that mention AI surged more than 130% above the same baseline. Employers aren't hiring less — they're filtering harder for a specific kind of worker. PwC's 2026 Global AI Jobs Barometer documents a 56% wage premium for AI skills. The roles contracting are those built primarily around structured data processing and templated output. The roles expanding are those requiring someone to decide what the AI output means for the actual business.

For anyone reading this in a marketing, HR, or operations context: the same split applies. The honest question to ask when your company announces AI-related restructuring is whether the tasks being eliminated are ones where AI is genuinely capable today, or whether this is a cost correction that needed a reason. The answer changes what you should do next.

How Fast Can You Actually Get to the Right Side of the Line?

Here is the part that tends to get lost in coverage about upskilling: the repositioning doesn't require reinventing your career. It requires a specific, learnable competency shift that experienced knowledge workers can execute in months.

Ahmed's method, stripped to its architecture: when he didn't know wealth management, he used AI to learn it and use it in the same afternoon. He built a layered approach — extract the domain vocabulary, map the end-to-end process, pull the regulatory landscape, identify the stakeholders. Then he cross-verified AI outputs against actual system behavior using tools like Postman and SQL queries — the technical verification step AI cannot do for you, because it cannot access a company's internal systems. The combination created his "Good catch" moment: AI got him close enough to the frontier, and his judgment operated from there. His own summary of the philosophy: "You lack something? Learn it. Use it. Fast."

Sid Arya, President of the IIBA Vancouver Chapter and Director of Enterprise Analytics and AI, frames the destination clearly: "In an AI project, the Business Analyst is not just a requirements gatherer — they are a strategic partner. The goal isn't to deploy an algorithm — it's to solve a real business problem."

A Forrester Total Economic Impact study on enterprise AI deployments found technical team productivity improved up to 35% when humans used AI tools within properly governed workflows — with the key variable being human oversight of AI outputs, not AI operating autonomously. The productivity gains went to the people who stayed in the loop, not the ones who handed off to the tool entirely.

Three practical moves translate Ahmed's experience into a general framework. First, validate rather than just generate: use AI to draft, then systematically check its outputs against real sources. The failure mode documented in the frontier research is over-reliance; the skill is knowing when to push back, treating AI output the way you'd treat a junior colleague's first draft — with respect but without uncritical acceptance. Second, own the human layer: stakeholder alignment, priority trade-offs, political navigation aren't soft skills, they are the tasks that directly define your irreplaceability. Third, build the domain, not just the tool: the tool is interchangeable. Domain knowledge combined with AI's speed is the durable differentiator. Ahmed's method compresses domain acquisition; it doesn't eliminate it.

The Fork in the Road Is Right Here

Ivan and Ahmed both faced a version of the same question — whether AI would accelerate them or replace them — in the same month, in the same profession. The difference wasn't that one was more talented. It was that one had already built a habit of operating at the frontier and learning through AI rather than around it.

The BAs who are thriving right now aren't the ones who weren't disrupted. They're the ones who got curious before they got scared — and used that curiosity to move their work toward the side of the frontier that AI can't reach yet.

Three concrete moves to start this week. This week: take one recurring task — a meeting summary, a first-draft user story, a status report — and run it through an AI writing tool. Don't use the output directly. Treat it as a junior colleague's first draft: what did it get right? What did it miss that only you would know? That gap between the AI output and your correction is your current frontier, made visible.

This month: identify one technical skill adjacent to your role — reading a basic API response, writing a simple SQL query, understanding what a data schema looks like — and use AI to teach it to you in a single sitting. Not to become an engineer. To stop needing one to translate for you.

This quarter: find one project at your organization that involves AI and ask to participate — not as a technical lead, but as the person who connects the AI output to the business decision. That role doesn't have a title yet. It will.

The fork Ivan and Ahmed stood at is the same one in front of you. The difference is you can see it now.

Explore Further