Where Actuaries Actually Go When They Leave (And What It Really Takes)

Jason Chong was doing standard insurance-pricing work — reserving, dashboards, the usual — when he noticed that every problem he found genuinely interesting involved scale, complexity, and a little ambiguity. He wasn't unhappy. He just kept gravitating toward the edge of his job description. In June 2026, he published a first-person account of how that pull eventually landed him in AI engineering at Quantium, deploying machine-learning applications across Australia, New Zealand, and Asia. He describes the turning point bluntly: "Working on one of the company's first AI applications, with no prior experience in the space, was the moment everything shifted."

If you work as an actuary and you've been watching the AI conversation from a distance — half curious, half anxious — Chong's story is worth taking seriously. Not because it's a fairytale. He's specific about what transferred and what didn't. His actuarial training gave him a useful question: what are we assuming here, and where could this model fail? It did not give him production data pipelines, cloud infrastructure, or the organizational fluency to align three regions on a single system. He built those separately, on real projects, over time. And he's clear about the organizational complexity that surprised him: different data standards, different stakeholder expectations across countries, as much communication and relationship-building as engineering.

That distinction — between what your background already gives you and what you'll need to build — is the most honest frame for evaluating any actuarial career pivot right now. The question isn't whether a transition is possible. Chong's dated, first-person account settles that. The question is which specific capability you still need to demonstrate.

But Chong's is one account. Before deciding what it means for your situation, you need to know which destinations keep showing up across multiple actuaries — and which skills show up on every bridge, regardless of the destination chosen.

Three Destination Families, Each With an Entry Condition Exams Don't Satisfy

Three families account for most documented actuarial pivots: insurance analytics and data science, risk and model governance, and product or consulting. Each has a specific entry condition that actuarial credentials alone don't meet.

The closest technical bridge is analytics and data science. An Institute and Faculty of Actuaries member survey found that 41% of actuarial respondents considered their current work more than half similar to data analysis, and 30% said the same of data science. That's genuine overlap, not wishful thinking. But the same survey surfaces the gap directly: respondents raised programming, machine-learning tooling, and data-architecture knowledge as outstanding development needs. The overlap is real. It is not sufficient.

The data-science market is also large enough to absorb actuarial candidates who can demonstrate relevant output. The Bureau of Labor Statistics projects 35% employment growth for data scientists between 2025 and 2035, with roughly 24,800 annual openings — compared to about 1,500 for actuaries. The destination is crowded, but it's not closed.

The product and consulting path looks different. Justin Yiu, ASA, spent four years in actuarial and analytics roles before moving into product management — eventually becoming head of product, insurance, at Grab. His account names precisely what the actuarial years didn't provide: customer research, negotiation, ruthless prioritization, storytelling. The missing skills were commercial and human, not more advanced mathematics.

You have one career and people should own it. Actuaries tend to be very conservative and may find the company owning their career instead of owning it themselves.
— Jason Bickler, FSA, Co-Head of Individual Markets, Global Atlantic

For you, the practical question isn't which destination looks best on paper. It's which missing capability you're willing to close. Analytics and data science requires demonstrated code and model output. Risk and governance requires evidence of challenging a consequential model. Product and consulting requires customer work and a record of influencing a decision — not a fancier spreadsheet.

What Actually Transfers — and Where People Go Wrong

Knowing which destination to aim for is only useful if you know what actually bridges the gap. The research on this is more specific, and more encouraging, than most career advice suggests.

Chong's experience is instructive here. His actuarial training gave him one durable question — what are we assuming here, and where could this model fail? — that production AI teams need and rarely find. What it didn't give him: data pipelines, cloud infrastructure, the cross-regional coordination that made a six-month deployment actually work. He built those on real projects, on the job, before the credential was relevant to his new title.

The SOA's 2025 member survey on AI reinforces why the upskilling gap is real but not enormous. Among actuarial respondents, 63.4% spend under five hours per week learning about or using AI. That's not a canyon. That's a credible training budget — six to twelve months of focused project work, applied to systems your organization is already using.

Then there's the failure mode, which Rob Stone, FSA, documents better than any abstract warning could. Stone built a successful 25-year actuarial consulting career before leaving a principal role to become a financial representative. By month six, he'd hit strong new-representative metrics. A client told him, "you actually care about us." Early validation by any measure.

Actuaries are trained to be professionally wrong in a useful way. We build models, we know they are simplifications of reality and we are rigorous about quantifying how wrong they might be and in which direction.
— Jason Chong, Senior Data Scientist, Quantium

By year two, he'd learned something the early success didn't reveal: the job required repeatedly chasing company-dictated performance metrics, every year, to validate his contract. He returned to a VP-level actuarial role. Not because he failed. Because he discovered what the work actually required on an ordinary difficult week — and it wasn't what he wanted.

Stone's case is the most useful timeline data point in this entire conversation. The question isn't whether you can do the job in month six. It's whether you can tolerate what the job requires every week in year two. Before targeting a destination, write two descriptions of the role: its most exciting day, and its most ordinary difficult week. The bridge-building exercise — the project, the code, the customer conversation — doubles as a test of whether you actually want the destination, not just the title.

The Numbers, With the Caveats Attached

With a destination in mind and the gap named, the remaining question is practical: what do the numbers actually say about pay, and which credential — if any — is worth the time?

The BLS provides a same-period comparison from May 2025. The actuarial median was $130,000. Data scientist came in at $120,230 — about $9,770 lower, or 7.5% less at the population median. Financial manager sat at $166,570. Software developer at $135,980. Management analyst at $101,860.

These figures are all-worker occupational medians from the same reference period. A financial manager's higher median requires five or more years of related management experience, per BLS — and that requirement isn't satisfied by an actuarial designation alone. The table is an initial screen, not a promise.

The more useful number comes from Acturhire's October 2026 salary benchmark, drawn from 5,720 actuarial postings with disclosed compensation. The median advertised actuarial base was $138,250 — but individual-contributor roles came in at $119,152, while VP-level roles reached $230,000. That spread is the insight. The comparison question isn't "actuary versus data scientist." It's "my current actuarial level versus the level I would enter the new role at." Level mismatch, not occupation mismatch, is the typical pay-cut mechanism.

The credential decision follows directly: pick the smallest training intervention that closes the specific hiring gap you face. An SOA predictive analytics module plus one reproducible project may open an insurance analytics door faster than a multi-year investment designation. A FRM or CFA makes sense only if the receiving employer names it as a requirement and your existing work already qualifies toward the experience component. Before enrolling in anything, ask a recruiter or hiring manager: which requirement does my actuarial background satisfy, which is missing, and which level would I enter at?

The national median is an initial filter. The negotiated offer, compared to staying in actuarial work at similar seniority, is the decision.

Start Before You Resign

The numbers give you a screen. What converts a credential and a salary benchmark into an actual offer is the portfolio — the specific, defensible work sample that shows a hiring manager you can already do the job, not just study for it.

Chong describes himself as "rigorous about uncertainty, curious about new technologies." The first half of that sentence — the rigour about uncertainty — is what his actuarial training gave him and what production AI teams struggle to find. The second half, the curiosity, is what he chose to act on. Only one of those two things required any change.

The actuarial habit of asking "where could this model fail?" is precisely the question that distinguishes a model a team trusts from one that looked good in testing and collapsed in production. That question travels. What doesn't automatically travel is the evidence that you've applied it to a live, consequential system — and that evidence is what a hiring manager needs to see.

Here's this week's exercise: identify one decision at your current employer where an AI or analytical output is being used without documented review of its assumptions or failure conditions. Write a one-page review — what the model assumes, where it could mislead, what a user should verify before acting on the result. That document is the first line of a portfolio. It doesn't require a new job, a new credential, or a resignation. It requires the same critical thinking you already apply, aimed at the tools your organization is now depending on.

You don't have to leave to prove you can think differently. You have to leave something on paper that shows you already do.

Recommended Tools & Resources