How to future proof your career without guessing which skills last
I can only afford to make one good decision right now, and every article tells me a different skill is the safe one.
Short answer
You cannot forecast a decade of technology, and neither can the US Bureau of Labor Statistics, which says so in its own evaluation documents. The things that hold up are legal gates, physical constraints and occupations people rarely leave. Pick the credential that opens a gate somebody else is legally required to respect, and ignore any list titled the safe skills of 2035.
There is no verified list of the skills that will be safe in ten years. The nearest thing to an authority on this, the US Bureau of Labor Statistics, projected that occupations would grow 6.5 percent on average between 2014 and 2024, and the actual figure came in at 12.9 percent. What survives forecasting error is structure, not technology. Work protected by a licence, a physical task or a legal requirement changes slowly and in public, with notice. Work that is mostly a task inside somebody else's software changes without warning. This page uses the August 2026 projections release and the separations data underneath it to show which is which, and what a certification can honestly do about it.
Two different lists, and the gap between them is the whole point
BLS publishes fastest growing occupations and occupations with the most job growth as separate tables, and the separation is not a formatting decision. Nurse practitioners top the fastest growing list at 41.0 percent for 2025 to 2035. Solar photovoltaic installers come second at 36.5 percent. Both figures are real. Both are also nearly useless for deciding what to do on their own, because a percentage is a ratio and the denominator matters more than the ratio does. Solar photovoltaic installers are projected to produce about 4,200 job openings a year across the entire United States. Nurse practitioners produce about 29,400. Registered nurses, an occupation projected to grow only 5.6 percent, produce about 180,800 openings a year. If you are choosing where to point two years of part time study, the third number is worth more than the first two combined. Fast growth in a small occupation means a few thousand hires nationally, spread across states, competing against people who already do adjacent work. This is the mechanism behind a common and expensive mistake: reading a growth percentage as a hiring signal. Growth rate answers the question of whether an occupation is expanding. Annual openings answers the question of whether anyone is likely to hire you. They are different questions and the same table gives you both.
Almost every opening you will ever apply to comes from someone leaving
The BLS occupational separations and openings table carries the number that reframes this whole subject. Across all occupations, projected average annual openings for 2025 to 2035 are 17,461,800. Of those, 16,870,000 come from separations, meaning 7,362,400 people leaving the labour force each year and 9,507,600 moving to a different occupation. Subtract and you find that net growth accounts for 591,800 openings a year, roughly 3 percent of the total. That makes about 97 percent of hiring in the American economy replacement hiring. Follow that through and the advice you have been reading inverts. An occupation can be shrinking and still be one of the largest sources of jobs in the country. Cashiers are projected to lose 6.5 percent of their employment by 2035 and still generate about 521,300 openings a year. General office clerks are projected to decline 6.0 percent and still generate about 249,000. Those are not good careers to aim at, for reasons covered below, but they demonstrate that decline and hiring coexist. The practical version: when someone tells you an occupation is dying, ask them for the annual openings figure. If they do not have it, they are talking about a trend line, not about whether you can get a job.
The durability signal is the exit rate, not the growth rate
The same table publishes an occupational transfer rate, which is the share of workers who leave that occupation each year for different work. This is the closest thing to a measured durability signal in public data, and it separates occupations far more sharply than growth rates do. Across all occupations the rate is 5.5 percent a year. Family medicine physicians transfer out at 0.4 percent. Dentists, 0.4 percent. Pharmacists, 1.1 percent. Lawyers, 1.2 percent. Registered nurses, 2.0 percent. Every one of those is an occupation you cannot enter without passing a state-controlled examination. Now the counterexample, which matters more than the pattern. Hairdressers, hairstylists and cosmetologists transfer out at 6.1 percent a year, above the all-occupation average, and BLS states that all states require them to be licensed. A licence alone does not produce stickiness. Paralegals and legal assistants, who work inside the legal system without holding the licence, transfer out at 7.1 percent. Read those four numbers together and the rule emerges: what holds is a licence that is expensive to obtain, legally required for work that only licence holders may perform, and attached to earnings high enough that leaving is costly. Cosmetology licensing gates entry without gating the valuable work, so people leave anyway.
What a licence does that a certificate does not
BLS draws the distinction plainly in its Current Population Survey documentation. A certification is issued by a non-governmental body and conveys that a person has knowledge or skill. A licence is awarded by a government agency and conveys legal authority to work in an occupation. That difference is the entire durability argument. A certification tells an employer something. A licence tells an employer they may not hire the alternative. In the 2025 annual averages, 24.0 percent of employed people held a certification or licence, and 21.6 percent held a licence specifically. Among the unemployed, 11.8 percent held either and 9.9 percent held a licence. Read that carefully, because the direction of causation is not established by it. Licensed occupations tend to be more stable, and stable occupations produce fewer unemployed people, so the gap partly reflects which occupations people are in rather than what the licence did for them. The gap is real. The inference that buying a credential moves you across it is not supported by that table. If you are weighing a credential against a degree, the page on choosing a certification instead of a degree covers that comparison directly, and the Security+ page covers whether one certification alone produces interviews.
Legal gates move too, and that is the good news
The objection to building a plan around licensure is obvious: rules change. They do. The useful part is how they change. In May 2025 the American Institute of CPAs and the National Association of State Boards of Accountancy approved an amendment to the Uniform Accountancy Act adding a third route to CPA licensure, a bachelor's degree with an accounting concentration plus two years of experience plus passing the Uniform CPA Examination. That sat alongside the two existing routes. As of the announcement, 14 states had already formally enacted legislation or adopted rules of their own. Compare the notice period on that to the notice period on a software product being deprecated. NASBA and the AICPA published the change as an exposure draft on 27 February 2025 with a 60 day comment period that closed on 3 May, the two boards approved it on 14 May, it went into a model act, and then each state legislature had to act separately. Anybody mid-study could read the draft before the vote. A regulated occupation is slower to enter and slower to disappear, and both halves of that sentence are the product you are buying. How much warning you personally get depends on your state, and this page does not put a number on it. What it does say is that every step of that sequence was published before it took effect, which is not true of waking up to find the tool you specialised in has been replaced inside a product update.
How accurate are ten year forecasts, according to the forecaster
BLS evaluates its own past projections and publishes the results, which is more than most sources of career advice do. For the 2014 to 2024 decade, BLS projected an average occupational growth rate of 6.5 percent. The actual average was 12.9 percent. BLS correctly called the direction, growth or decline, for 86 percent of occupational groups. On detailed occupations, BLS compares itself against a naive model that assumes each occupation keeps the same share of total employment. BLS scored better on 475 detailed occupations. The naive model scored better on 345. On roughly two in five detailed occupations, assuming nothing changes beat the professional forecast. BLS also prints a warning box in the projections news release stating that the precision of its estimates does not account for the inherent uncertainty of predicting long-term labour market changes, and advising readers to focus on direction and relative size rather than precise values. Take that seriously. It means the correct use of the 2025 to 2035 projections is to rank occupations against each other, not to believe any single number. It also means that any article naming the specific safe skills of 2035 is doing something the agency that produces the underlying data explicitly declines to do.
The most repeated statistic in this field has no source
You have seen the claim that 65 percent of children entering primary school today will work in jobs that do not yet exist. It appears constantly, usually credited to the World Economic Forum. The WEF article that popularised it says some studies suggest the figure, and names no study. Cathy Davidson, the academic most often cited as its origin, published an account of tracking it down. She first read it in Jim Carroll's book Ready, Set, Done from 2007, traced his citation to an Australian website, tried to contact the study authors and could not reach them, and found the source page and the body that hosted it had both gone. She stopped using the number around 2012. The BBC World Service programme that interviewed her ran its own rough comparison of current and historical job titles and estimated that around a third of jobs today are new. That is a different figure produced by a different method, and it is offered as a rough calculation rather than a finding. Treat any career statistic you cannot trace to a named publisher and a live URL the same way. Widely repeated is not the same as verified, and in this field the two are often inversely related.
The renewal bill is part of the purchase price
A durable credential that you stop paying for is not durable. Every major certification body runs a maintenance cycle, and the costs are published, so price them before you buy rather than after. CompTIA certifications in its continuing education programme expire three years from the date you earn them and must be renewed before that date. Security+ V7 requires 50 continuing education units, which you can accumulate through approved activities or clear in one step with a qualifying higher certification, and CompTIA charges a continuing education fee on top unless you renew by earning a qualifying higher-level CompTIA certification or by passing CertMaster CE. ISACA requires CISA holders to earn and report a minimum of 20 continuing professional education hours every year, at least 120 across each three-year reporting period, and to pay an annual maintenance fee of 45 US dollars for members or 85 for non-members. Failure to comply results in revocation. AWS and PMI publish their own renewal terms on their certification pages, and the terms differ by credential, so read the one you are considering rather than a summary. Two consequences follow. First, a credential you buy while unemployed carries a recurring obligation that arrives whether or not the credential got you hired. Second, the credentials with the heaviest renewal requirements are usually the ones attached to regulated work, which is the same property that makes them durable. You are buying an obligation as well as a qualification.
Decision rules that survive being wrong
Since the forecast will be wrong, choose things that pay off under several futures. Rule one: prefer the occupation with the larger annual openings figure over the one with the larger growth percentage, and look both up in the BLS separations and openings table before deciding. Rule two: check the occupational transfer rate. Under about 2 percent a year usually means a legal gate and high switching costs. Above the 5.5 percent average means people leave routinely, which also means they enter routinely, so entry is easier and the position is less protected. Rule three: ask what would have to be true for this credential to become worthless, then ask who would have to announce it. If the answer is a state legislature or a licensing board, you get years of warning. If the answer is a vendor product manager, you get a changelog. Rule four: do not buy anything until a named person tells you the credential is what is blocking them from hiring you or promoting you. If you are trying to move up where you already work, the page on getting a raise with a certification covers what your employer actually responds to. If you are twenty years into one industry and starting over, the pages on being laid off after 20 years and on changing careers at 50 deal with the specific frictions there. One caveat on all four rules. The single skill BLS ranks first for all occupations combined, and first in every education category except occupations needing no formal credential, is adaptability. That is not a thing you can buy, sit an exam in, or put on a resume line, which is exactly why it does not appear in anybody's list of safe skills. None of these problems is solved by picking the correct future technology, because nobody can pick it.
What this does not fix
This page does not name the safe skills of 2035 and does not claim any credential is future proof. It gives no forecast of its own beyond the published BLS projections, which BLS itself says should be read for direction rather than precise values. The occupational transfer rates used here as a durability signal are BLS projected annual averages for 2025 to 2035, not observed outcomes, so they inherit the same forecasting error the page warns about. It also makes no claim that buying a certification causes the employment gap between credential holders and non-holders in the CPS data, and it puts no number on how much warning a licensing rule change would give you, because that varies by state and nobody publishes it.
Where an exam fits
- NCLEX-RN, if you are already enrolled in or have completed a board-approved nursing programme, since nothing on this page gets you past that prerequisite
- CPA (Certified Public Accountant), if you hold or are finishing an accounting concentration and your state board has enacted one of the pathways, which you should confirm on the board's own site first
- CISA (Certified Information Systems Auditor), if you already have the information systems audit or control experience ISACA requires, or a documented waiver for it, and you have budgeted the annual maintenance fee
- CompTIA Security+, if a specific job posting or contract clause names it and you can afford the three-year renewal cycle and its fee
- Certified in Cybersecurity (CC), if you want to test whether security work interests you before spending real money, and you accept it will not clear a hiring gate on its own
- AWS Solutions Architect Associate, if the systems you already work on run on AWS and your employer pays for renewals
- PMP (Project Management Professional), if you already have the documented project hours PMI requires, because the credential is unavailable to you without them
Common questions
What is the single most future proof career?
There is not one, and any source naming one is guessing. The measurable proxy for durability is the occupational transfer rate, and by that measure physicians, dentists, pharmacists, lawyers and registered nurses hold up best, all at or under 2.0 percent a year against a 5.5 percent average. Every one of those requires years of training and a state examination, so the answer is expensive rather than clever.
Should I chase the fastest growing occupations list?
Not by itself, because growth percentage says nothing about hiring volume. Solar photovoltaic installers grow 36.5 percent over the decade and produce roughly 4,200 openings a year nationally, while registered nurses grow 5.6 percent and produce roughly 180,800. Look up the annual openings column in the BLS separations and openings table before you commit study time.
Is it pointless to enter a declining occupation?
Not automatically, because 16,870,000 of the projected 17,461,800 annual openings come from people leaving rather than from growth. Cashiers are projected to decline 6.5 percent and still generate about 521,300 openings a year. Decline is a reason to expect worse pay progression and more competition over time, not proof that nobody is hiring.
Does a certification protect me from being replaced?
A certification conveys knowledge; a licence conveys legal authority to do work others may not do, which is the property that protects an occupation. BLS makes exactly that distinction in its Current Population Survey documentation. If the credential you are considering does not create a legal barrier, its protection depends entirely on employers continuing to value it.
How reliable are the BLS ten year projections?
BLS called direction correctly for 86 percent of occupational groups over 2014-24 but underestimated average growth badly, projecting 6.5 percent against an actual 12.9 percent. On detailed occupations, a naive model assuming constant employment shares outperformed BLS on 345 occupations to BLS's 475. Use the projections to rank occupations, not to trust any single figure.
Is the claim that 65 percent of future jobs do not exist yet true?
It is unsourced. The World Economic Forum article that spread it attributes the figure to some studies without naming any. Cathy Davidson, whose 2011 book is usually credited, has written that she took it from a 2007 book by Jim Carroll, traced it to an Australian website that later went offline, could not reach the authors, and stopped using it around 2012.
What does a certification cost after I pass?
More than the exam fee, and the terms are published in advance. CompTIA Security+ V7 needs 50 continuing education units per three-year cycle. ISACA requires CISA holders to report 20 CPE hours a year and 120 over three years plus an annual fee of 45 US dollars for members or 85 for non-members, with revocation for non-compliance. Price the renewal before you buy.
If technology is unpredictable, what should I actually do?
Prefer high annual openings over high growth percentage, prefer occupations with transfer rates under about 2 percent, and prefer credentials that only a legislature or licensing board can devalue, because those changes arrive with years of notice. Then wait until a named employer tells you the credential is what is blocking you before you spend the money.
Sources
- 01Total US employment is projected to grow 3.5 percent from 2025 to 2035, adding 5.9 million jobs, slower than the 10.9 percent recorded over 2015-25; office and administrative support occupations are projected to decline 4.0 percent and shed 752,100 jobs, the largest loss of any major occupational group 3.5 percent growth; 5.9 million jobs; office and admin support -752,100
U.S. Bureau of Labor Statistics, Employment Projections 2025-2035 news release (USDL-26-1422) checked 2026-08-28 - 02Fast percentage growth and hiring volume are different things: nurse practitioners are projected to grow 41.0 percent with about 29,400 annual openings, solar photovoltaic installers 36.5 percent with about 4,200 annual openings, and registered nurses only 5.6 percent with about 180,800 annual openings Solar PV installers 4,200 annual openings vs registered nurses 180,800
U.S. Bureau of Labor Statistics, Occupational separations and openings, 2025-35 checked 2026-08-28 - 03Declining occupations still generate large numbers of openings: cashiers are projected to decline 6.5 percent and still produce about 521,300 annual openings, and general office clerks to decline 6.0 percent and still produce about 249,000 Cashiers 521,300 annual openings while declining 6.5 percent
U.S. Bureau of Labor Statistics, Occupational separations and openings, 2025-35 checked 2026-08-28 - 04Projected average annual occupational openings for 2025-35 total 17,461,800, of which 16,870,000 come from separations (7,362,400 labour force exits plus 9,507,600 occupational transfers), leaving 591,800 a year from net employment growth 16,870,000 of 17,461,800 annual openings come from separations
U.S. Bureau of Labor Statistics, Occupational separations and openings, 2025-35 checked 2026-08-28 - 05Annual occupational transfer rates vary sharply by licensure: all occupations 5.5 percent, family medicine physicians 0.4, dentists 0.4, pharmacists 1.1, lawyers 1.2, registered nurses 2.0, hairdressers and cosmetologists 6.1, paralegals and legal assistants 7.1 All occupations 5.5 percent; lawyers 1.2 percent; paralegals 7.1 percent
U.S. Bureau of Labor Statistics, Occupational separations and openings, 2025-35 checked 2026-08-28 - 06All US states require barbers, hairstylists and cosmetologists to be licensed, and candidates typically must graduate from a state-approved program and pass an exam All states require a licence
U.S. Bureau of Labor Statistics, Occupational Outlook Handbook: Barbers, Hairstylists, and Cosmetologists checked 2026-08-28 - 07In 2025 annual averages, 24.0 percent of employed people held a certification or licence and 21.6 percent held a licence, against 11.8 percent and 9.9 percent of the unemployed; BLS defines a certification as issued by a non-governmental body and a licence as awarded by a government agency conveying legal authority to work Employed 24.0 percent vs unemployed 11.8 percent
U.S. Bureau of Labor Statistics, Current Population Survey Table 49, 2025 annual averages checked 2026-08-28 - 08In its own evaluation, BLS projected an average occupational growth rate of 6.5 percent for 2014-24 against an actual 12.9 percent, called growth or decline correctly for 86 percent of occupational groups, and on detailed occupations beat a naive constant-share model on 475 occupations while the naive model scored better on 345 Projected 6.5 percent, actual 12.9 percent; naive model better on 345 of 820
U.S. Bureau of Labor Statistics, Occupational Projections Evaluation: 2014-24 checked 2026-08-28 - 09In the BLS skills data, adaptability is the top-ranked skill for all occupations combined and for every category of typical entry-level education from no formal credential through doctoral or professional degree, except no-formal-credential occupations where it ranks second behind detail oriented Adaptability ranked first overall
U.S. Bureau of Labor Statistics, Top skills by typical education needed for entry, 2025 checked 2026-08-28 - 10CompTIA Security+ V7 requires 50 continuing education units for renewal, and CompTIA publishes a list of qualifying certifications from other bodies that each count for the full 50 50 CEUs to renew Security+ V7
CompTIA, CompTIA Security+ V7 - 50 CEUs Required for Certification Renewal checked 2026-08-28 - 11CompTIA certifications in the continuing education programme expire three years from the date they are earned and must be renewed before that date; continuing education fees are required for renewal unless the holder renews by earning a qualifying higher-level CompTIA certification or by passing CertMaster CE Three-year renewal cycle plus a CE fee
CompTIA, How to Renew Your CompTIA Certification checked 2026-08-28 - 12Maintaining ISACA's CISA requires a minimum of 20 continuing professional education hours annually and 120 over a three-year period, plus an annual maintenance fee of 45 US dollars for members or 85 for non-members, with revocation for non-compliance 20 CPE annually, 120 over three years, 45 or 85 US dollars a year
ISACA, Maintain CISA Certification checked 2026-08-28 - 13On 14 May 2025 the AICPA and NASBA approved a third model pathway to CPA licensure, a baccalaureate degree with an accounting concentration plus two years of experience plus passing the Uniform CPA Examination, and stated that 14 states had already enacted comparable changes Third licensure pathway; 14 states enacted as of May 2025
AICPA and NASBA joint announcement checked 2026-08-28 - 14BLS prints an interpretation box in the projections news release stating that the precision of its projected employment levels and changes does not account for the inherent uncertainty of predicting long-term labour market changes, and that focusing on the direction and relative size of projected changes rather than the precise values may yield similar insights BLS advises reading direction and relative size, not precise values
U.S. Bureau of Labor Statistics, Employment Projections 2025-2035 news release (USDL-26-1422), Interpreting the Employment Projections box checked 2026-08-28 - 15NASBA and the AICPA published the proposed Uniform Accountancy Act changes as a public exposure draft on 27 February 2025, including the third licensure pathway of a bachelor's degree plus two years of experience plus passing the CPA Examination, with comments due 3 May 2025 after a 60-day exposure period Exposure draft published 27 February 2025, comments due 3 May 2025
NASBA, 2025 Proposed Uniform Accountancy Act Changes checked 2026-08-28 - 16The World Economic Forum article that popularised the figure says only that some studies suggest 65 percent of children currently entering primary school will have jobs that do not yet exist, and names no study 65 percent attributed to unnamed studies
World Economic Forum, We may have less than 5 years to change how we learn, earn and care, Saadia Zahidi, 4 January 2017 checked 2026-08-28 - 17Cathy Davidson, whose 2011 book Now You See It is usually credited with the 65 percent figure, writes that it did not originate with her, that she first read it in Jim Carroll's 2007 book Ready, Set, Done, that she traced his citation to an Australian website, that she could not reach the study authors, that by then the citation was down and the Innovation Council of Australia had been closed, and that she stopped using the figure around 2012; she also reports that the BBC researchers compared current and historical job titles and found that maybe one third of jobs today are new Traced to a dead source; BBC rough count put new jobs nearer one third
Cathy Davidson, 65% of Future Jobs Haven't Been Invented Yet? Cathy Davidson Responds to Cathy Davidson and the BBC, HASTAC, 31 May 2017 checked 2026-08-28 - 18The BBC World Service statistics programme More or Less ran an episode titled Have 65% of Future Jobs Not Yet Been Invented?, first broadcast 26 May 2017, tracing the statistic and interviewing Cathy Davidson among others BBC More or Less episode, first broadcast 26 May 2017
BBC Radio 4, More or Less: Have 65% of Future Jobs Not Yet Been Invented? checked 2026-08-28
Related
I do not have a degree and cannot afford one, and I cannot tell which certifications are real qualifications and which are expensive decoration.
I passed Security+, I have sent more than a hundred applications, and I have not had one interview.
I am about to give up months of evenings to a certification and I want to know whether my employer will actually pay me more for it.
My position was abolished after two decades, the industry is not coming back, and I cannot work out whether to retrain or just keep applying.