AI is raising demand for builders, not erasing them
In February, a Citadel Securities analysis using Indeed data showed software-engineer job postings rising while overall job postings stayed weaker.
That split does not mean AI is creating jobs across the whole economy. However, one of the clearest fears around large language models may be somewhat overblown. The current narrative is that companies will need fewer skilled builders as the tools improve, but this has not shown up in this part of the labor market.
The sharpest conclusion is narrower and stronger. AI is increasing the value of people who design systems, test outputs, fix failures, and own results, while putting more pressure on roles built around repeatable processes such as formatting, scheduling, and throughput.
In the crypto industry, exchanges, wallet teams, data providers, staking firms, and protocol developers can use AI to write code faster, review documents faster, and automate support tasks. They still need people who know what a secure product looks like, what a broken workflow looks like, and what can go wrong in production.
Labor data points in the same direction. A January 2026 report found tech job postings rose 13% month over month, even as tech industry employment fell by about 20,155. Companies appear willing to cut in some places while still hiring for scarce technical capacity.
Longer-term projections also do not fit the simple replacement narrative. Federal projections show software developers, quality assurance analysts, and testers growing 15% from 2024 to 2034, with about 129,200 openings each year.
The same federal forecast projects 6% growth in project management specialist jobs over that span, with roughly 78,200 openings a year. Those numbers do not say every developer or manager wins. Firms still expect to need large numbers of people who can ship products, coordinate teams, manage budgets, and own delivery. And that aligns with what the current AI tools are actually used for.
A January 2026 index found that computer and mathematical tasks still accounted for about a third of Claude.ai conversations and nearly half of first-party API traffic in November 2025.
The single most common task was modifying software to correct errors, at 6% of usage. In other words, one of the most visible uses of AI is not replacing software work. It is speeding up software maintenance, debugging, and iteration.
That same workflow logic reaches beyond code
For illustration or graphic design, the evidence is thinner, but the mechanism looks similar.
When a company uses AI to generate concepts, draft a visual identity, or expand a design system, it still needs a person who can judge composition, coherence, brand fit, and finish.
AI can widen the output of a skilled designer. It does not remove the need for someone who knows what good looks like and can reject what does not.
For crypto firms, that applies to product art, marketing assets, exchange interfaces, wallet flows, dashboards, campaign creative, and brand systems.
A designer using AI can move faster across variations, mockups, and production tasks. The value shifts toward direction, editing, taste, and final approval.
The value shifts toward architecture, verification, integration, and release judgment. AI compresses production time. It does not erase the need for expert oversight.
That is why the cleanest framing is not “AI saves jobs” or “AI kills jobs.”
The better assessment is that AI is changing the mix of work inside firms. The workers who gain the most are those who can set direction, judge quality, test claims, and take responsibility when a model fails.
The workers at higher risk are those whose output can be measured as a sequence of rules and handed off to a cheaper human-plus-software workflow.
Verified signalWhat the number saysForward readSoftware-engineer postings rose while overall postings stayed weakerA February 2026 analysis found developer demand strengthening relative to the broader marketFirms still need builders even as they automate other workTech job postings rose 13% month over monthA January 2026 report showed higher hiring intent despite payroll weaknessCompanies may be reorganizing teams rather than retreating from hiring altogetherGenerative-AI work adoption reached 37.4%A 2025 survey showed broader workplace useDiffusion is real, but still gradual enough to argue against sudden mass replacementAI time savings equaled 1.6% of all work hoursThe same survey estimated labor productivity may have risen by up to 1.3% since ChatGPT launchedProductivity gains are starting to show up before broad labor destruction doesOffice and admin support rose to 13% of API trafficA January 2026 index showed more automation in email, documents, CRM, and schedulingRoutine support work faces more direct substitution pressureHighly exposed young-worker employment fell from 16.4% to 15.5%A January 2026 paper found early weakness at the entry point to AI-exposed jobsThe main risk may be a weaker career ladder, not immediate mass layoffs
AI use is spreading, but the pressure is uneven
Adoption data supports change rather than panic. A late-2025 survey found generative-AI use among adults ages 18 to 64 rose from 44.6% in August 2024 to 54.6% in August 2025.
Work use rose from 33.3% to 37.4% over the same period. The share of work hours spent using generative AI moved from 4.1% in November 2024 to 5.7% in August 2025. Those numbers show real diffusion. They do not show a labor market already hollowed out by automation.
The same survey estimated AI time savings equal to 1.6% of all work hours and said labor productivity may have risen by up to 1.3% since ChatGPT’s release. It also found that industries with one percentage point higher AI-related time savings saw 2.7 percentage points higher productivity growth relative to prepandemic trend, while noting that the relationship was not necessarily causal.
Productivity can rise before headcount falls. In many firms, the first move is not elimination. It is asking the same team to produce more.
That pattern fits what crypto firms have been doing for years, even before this AI cycle.
Teams stay lean. Work moves into software where it can. Functions with clear rules get automated first. What changes with LLMs is the range of tasks software can now touch: internal search, policy drafting, coding assistance, support triage, fraud review, and document handling.
But crypto products still involve security trade-offs, operational risk, compliance judgments, user-experience decisions, incident response, and release discipline. A model can help with all of those tasks. It does not own any of them.
The same applies on the creative side inside crypto businesses. Teams can use AI image and design tools to generate options faster, test multiple directions, and build more variants for social, editorial, product, and campaign use. But speed does not settle the hard parts. Someone still has to choose which visual language fits the product, which illustration style matches the brand, which dashboard or landing page reads clearly, and which asset crosses a line on quality or trust.
In that sense, AI can make skilled creative workers more productive, just as it makes skilled developers more productive: by reducing time spent on first drafts and widening the range of outputs they can explore.
That is also why managers and senior individual contributors look more durable than the public debate assumes. Federal definitions for project management specialists still center on staffing, schedules, budgets, milestones, and risk. Those are not ornamental functions.
The work of turning a product idea into something a firm can ship, maintain, defend, and explain still requires humans to lead.
In crypto, where teams often move across jurisdictions, smart contract stacks, and shifting market conditions, that coordination burden can rise as AI lowers the cost of producing drafts and prototypes.
Even the debate inside AI usage data points to a mixed picture rather than a clean handoff from humans to models.
A September 2025 report found directive conversations rose from 27% to 39% between early 2025 and late summer 2025, suggesting users were delegating more. But a January 2026 update found augmented use had regained the lead on Claude.ai in November 2025, at 52% versus 45% for automated use. Firms are still testing where they trust the model to act on its own and where they still want a human-in-the-loop.
For the crypto sector, that line likely runs through security, treasury operations, listings, market surveillance, product launches, and brand-facing work.
AI can reduce the time spent on repetitive work inside those functions. But as the financial and reputational stakes rise, the value of judgment, review, and accountability rises too. That tends to favor experienced operators, editors, designers, and technical leads over firms hoping to run critical systems or public-facing outputs on autopilot.
The bigger labor question is who still gets a path in
The strongest warning sign is not a collapse in demand for experienced builders. The strain at the bottom of the ladder is increasing, and a January 2026 paper found lower employment only for younger workers in the most AI-exposed occupations, with the share of employment in those jobs slipping from 16.4% in November 2022 to 15.5% in September 2025.
The authors stressed that aggregate effects remained small, estimating that even if the entire decline translated into unemployment, it would explain only a 0.1 percentage-point rise in aggregate unemployment since November 2022. Still, the signal is there.
That fits the rest of the evidence. Routine office and administrative support work rose by 3 percentage points to 13% of API traffic in a January 2026 index. The categories include email management, document processing, CRM work, and scheduling.
A 2025 study also found that clerical occupations remained the highest exposure category globally, while estimating that one in four workers worldwide were in jobs with some generative-AI exposure, and only 3.3% of global employment sat in the highest exposure category. Transformation looks more common than outright replacement. But transformation is not painless when it starts by cutting junior tasks.
The same risk could extend into junior creative and junior technical roles. If entry-level work gets absorbed into AI-assisted workflows, fewer people may spend their early years doing the production tasks that once taught pacing, taste, debugging, revision, and client judgment.
In software, that may mean fewer junior coding and QA openings. In design, it may mean fewer production-heavy roles where people learned layout, systems thinking, and visual discipline by doing. Firms may gain speed in the short run and still weaken their own pipeline.
That is where the forward-looking case gets more serious. If firms use AI to shrink the volume of entry-level coding, coordination, support, research, drafting, and production work, then fewer people will get the apprenticeship that once led to senior jobs.
The short-term economics can look good. Teams stay smaller. Output rises. Margins improve. But the medium-term risk is a thinner talent pipeline.
Crypto firms, which already struggle to hire people who understand market structure, security, product, and trust under pressure, could end up competing even harder for experienced operators if they stop training enough new ones.
Global forecasts support a mixed outcome rather than a one-line verdict
A 2025 forecast projected structural labor-market change equal to 22% of today’s jobs by 2030, with 170 million jobs created and 92 million displaced, for a net gain of 78 million. The same forecast listed AI and machine learning specialists, fintech engineers, and software and application developers among the fastest-growing roles in percentage terms. But an IMF review warned that advanced economies would feel both the benefits and the disruptions sooner, and that gains could concentrate among higher-income workers and capital owners.
That leaves a cleaner conclusion than the public debate usually offers. AI is not yet showing up as a broad collapse in demand for high-skill builders. The numbers point the other way. They show stronger hiring signals for developers than for the broader market, rising use of AI inside work, measurable productivity gains, and clearer substitution pressure in administrative and clerical tasks than in expert technical roles.
The same logic also appears to apply to creative work. In both cases, AI looks more like a force multiplier for skilled workers than a substitute for them.
For crypto companies, the next step is plain. Firms can use AI to produce more drafts, ship more tests, generate more concepts, and automate more support work. They still need humans to decide what gets shipped, what stays secure, what meets policy, what fits the brand, and what breaks trust.
The near-term winners are likely to be the teams that use AI to widen the output of experienced operators without destroying their own training pipeline.
The next open question is whether companies keep hiring the people who can own outcomes while quietly cutting the people who once learned how to do so.
