Technology is bifurcating fast: AI/ML engineers, security architects, and senior infra roles are appreciating while QA, frontend production, sysadmin, and routine application work compress. Here is the structural read on where the line sits and what to do about it.
Technology is the one category where the people building AI tooling are also the people getting compressed by it. That makes the structural read inside the field both faster and stranger than the other categories we track. The work that used to define a software engineer's day, scaffolding routine UI, writing CRUD endpoints, generating tests, drafting deployment scripts, has migrated into the model in eighteen months, while the architect-tier and AI-native work has gotten dramatically more valuable. The AI Career Risk In Technology reading shows the split clearly: the seats that own systems are appreciating, the seats that produce inside someone else's system are compressing, and the line between them has moved upward by about half a band in two years.
The short version is that technology is bimodal in the same way marketing and finance are, but the speed of the curve is faster because the tooling is being built and adopted by the same population. There is no regulatory friction slowing it down. There is no licensure perimeter forcing a human signature on production code. Where you sit on the architect-versus-producer line is the read that matters more than seniority of title or years of experience.
What is actually compressing
Four concrete pieces of technology work are getting absorbed by AI tooling right now, in production, at companies that hired engineers to do this work eighteen months ago.
First, frontend production work. Component scaffolding, form layouts, dashboard pages, marketing sites, admin consoles. AI coding assistants and generative UI tools now produce a working first draft from a prompt or a Figma file in minutes. The frontend developer at the production tier, the seat that used to be the entry point into the field, is exactly the seat where the compression is most visible. The role is still there, but the deliverable inside the role has shifted from writing markup to reviewing and integrating AI output.
Second, test generation and QA scaffolding. Unit tests, integration tests, mock data, fixture builders, snapshot comparisons. AI test-writing tools now produce coverage at acceptable quality from a function signature or a feature spec. QA engineers at the routine tier, where the work is generating cases for known surfaces, are getting compressed faster than almost any other role in tech. Senior QA work, the kind that designs test strategy and owns release-blocker calls, is durable. The production-tier test author is not.
Third, routine sysadmin and infrastructure scripting. Provisioning Terraform modules, writing Ansible playbooks, building deployment pipelines, drafting runbooks, generating monitoring config. AI infrastructure-as-code tools now produce these from natural-language descriptions. The systems administrator and the junior DevOps engineer, both heavy in this work, are the seats getting squeezed. Senior platform engineering, where the call is which tooling stack to standardise on and how to design for blast radius, is appreciating in the opposite direction.
Fourth, documentation and developer-relations content. API references, getting-started guides, release notes, code comments, internal wikis. AI documentation tools now generate the first draft from the codebase itself. Technical writers and the routine documentation work inside engineering teams are compressing fast. The senior developer-relations seat that owns external advocacy and community work is durable for the same reason senior brand work in marketing is durable: it sits at the layer AI tooling cannot reach.
What is appreciating
The flip side of the split is that architect-tier and AI-native engineering roles are getting more valuable, not less. Three groups are clearly appreciating in the structural data.
AI engineers and machine learning engineers are the cleanest case. The supply of people who can ship a working LLM application end-to-end, debug an inference pipeline at scale, fine-tune against a domain dataset, or design retrieval architectures, is well below demand. The wage premium on these roles has widened sharply since 2024 and the projection for the next two years shows the gap widening further. This is the only category in tech where the role itself was created by the AI shift, not just reshaped by it.
Cybersecurity specialists, security architects, and CISO-track roles are the second group. AI tooling has expanded the attack surface faster than it has expanded the defensive layer. Threat modelling, incident response, security architecture, and the irreducibly human stakes calculation around breach impact, all sit firmly above the line where AI tooling can substitute. The compression that hits routine security operations does not reach the architect-tier seats that own the threat-model and the audit response.
Solutions architects and senior platform engineers are the third group. The seats that decide what to build, how to design for scale, where to draw the boundaries between services, and how to choose between competing tooling stacks, are appreciating because the productivity gains AI tooling unlocks at the producer tier flow upward to the architect who decided what to produce. Every additional unit of generated code makes the architect's design call more consequential, not less.
Why the curve is faster in tech than elsewhere
Three structural conditions make technology compress faster than any other category we track. First, the tooling itself is built by the people inside the field. Adoption inside engineering teams is near-instant because the buyers are the users and the friction between trying a new tool and shipping with it is hours, not quarters. Second, there is no licensure perimeter. No state bar, no medical board, no regulatory body forces a human signature on production code. The market alone decides whether the AI output is good enough, and by mid-2026 the market has decided that for an expanding share of the work it is. Third, the cost calculus inside engineering teams is tight. Senior engineers cost five to ten times what AI tooling costs at the throughput AI tooling now achieves on routine surfaces. The headcount math gets done quickly when the seat being replaced is producing the kind of output AI is now genuinely good at.
What this means if you sit inside technology
If you are an AI engineer, security architect, solutions architect, senior platform engineer, or principal-tier IC, the strategic answer is to keep extending your scope. Take more of the architecture calls. Own more of the build-versus-buy decisions. Build proof points that the system designs you owned moved the business outcome. The compression is happening below you. Your job is to make sure your seat sits firmly above the line where AI tooling stops, and that the organisation reads you as the person who decides what gets built rather than the person who builds it.
If you are a frontend developer at the production tier, a routine QA engineer, a junior DevOps engineer, a systems administrator, or a database administrator working primarily on routine optimisation, the honest reading is that your seat is exposed and the compression is already underway in plenty of teams. The strategic answer is not to leave technology. It is to deliberately migrate inside the category toward the architect surface. That looks like taking ownership of one piece of the system design work, getting fluent in AI tooling so the productivity gain accrues to you rather than to your replacement, and getting in the room when architecture and platform calls happen. Six to twelve months of that, run with intent, will move your structural reading meaningfully.
If you are an engineering leader running a team, the structural play is to redesign the headcount mix. The teams that win in the next two years will look different from the teams that won in the last two: fewer producers, more architects, more AI-native specialists, and a tooling line item that does the production work the producers used to do. The same logic that applies to individuals running a six-month plan applies to an engineering org running a hiring plan.
The honest read
Technology is not dying. Software engineering is not being replaced. But the production layer of the field, which has been the entry point into engineering careers for two decades, is getting compressed faster than the field has had time to restructure. That is the structural read. If you work in technology today, the question is not whether the field changes. It changes either way. The question is whether you migrate to the appreciating part of the field on purpose, or wait for the market to do the migration for you.
The full per-role detail, including the moderate-exposure roles in the middle of the curve, the specific migration paths between production tiers and architect tiers, and the six-month plan applied per role, is mapped at the technology hub. Read the full AI Risk In Technology Jobs page for the per-role reading and the structural moves that hold value over the next two years.
Run your structural assessment
The AI Career Index is free to take. Get your structural snapshot in 10 minutes, including your durability score, archetype, and projection.
Take the Free Assessment