AI Is Collapsing the Bottom Rungs of IT and Software Work

This is one of the few industry assessments in the full set that is not abstract. It points directly at the part of the labor market most exposed to AI right now: software and IT work.

The source file makes that explicit. This is the industry closest to Kane’s own operating lane, so the interpretation has to be precise.

Its topline conclusion is blunt: the sector sits around 50% overall AI replacement exposure, but the change is not distributed evenly. AI is not removing the entire software function. It is hollowing out the bottom, compressing the middle, and pushing value upward into architecture, systems judgment, security, integration, and AI-native work.

That is why the market feels contradictory. Software labor is still strategically central. But fewer people are needed to ship many of the same outcomes.

The Core Signals Are Already Here

The source highlights a small set of numbers that capture the structural break:

  • roughly 46% of new code on GitHub is now AI-assisted
  • graduate hiring at the top 15 U.S. tech companies is down 55% from 2019
  • one senior engineer with AI can produce the output of what used to require a three-person team
  • Forrester expects around 70% of DevOps tasks to be automated by 2026
  • 65% of developers expect their role to be redefined by 2026
  • hiring for roles such as GenAI Engineer and MLOps Specialist is growing 2-3x year over year

That combination matters more than any single statistic. It means AI is not just speeding up coding. It is changing team design, hiring logic, and what companies think they are paying engineers for.

The Highest-Risk Jobs Are the Ones Built Around Repeatable Implementation

The source’s exposure map is straightforward. The most threatened roles are the ones where work is already:

  • structured
  • testable
  • well-documented
  • and easy to scaffold

The Most Exposed Roles

Role Estimated AI replacement rate Why it is exposed
Manual Test Engineer 90% Test generation, execution, and regression are increasingly automated
Junior Frontend Developer 80% Site scaffolding and component generation are now commodity workflows
L1 Technical Support 80% Password resets, FAQs, and ticket triage are classic chatbot tasks
HR-like entry equivalent in software: Junior Backend Developer 75% CRUD, API scaffolding, and routine service logic are heavily AI-assisted
SOC Analyst (L1) 75% Alert triage and first-pass investigation are becoming system-driven
DevOps Engineer 70% Standard CI/CD, monitoring, and infra workflows are becoming automated
System Administrator 70% Configuration management and routine operations are moving into tooling
BI Analyst 70% Dashboard generation and anomaly surfacing are becoming self-serve

This is not a theoretical future. The source treats these as current or near-term shifts.

The strongest pattern is that entry-level and operationally repetitive jobs are losing labor value first. That is why the most important phrase in the file is not “AI writes code.” It is “the bottom rungs are disappearing.”

Junior Development Is the First Casualty

The sharpest finding in the source is the vulnerability of junior engineering work.

The junior frontend developer sits at 80% exposure, the junior backend developer at 75%. The reason is not that software is easy. It is that a meaningful portion of entry-level software work used to consist of:

  • boilerplate generation
  • CRUD endpoints
  • UI scaffolding
  • routine styling
  • documentation lookup
  • and fixing small isolated bugs

That was once apprenticeship work. Now it is increasingly machine-native work.

The consequence is bigger than job loss. It is a pipeline problem. If junior engineers do less implementation, how do they become senior engineers with production intuition? The source does not fully expand that point, but it is implicit in the hiring collapse it cites.

Architecture and Leadership Stay Safer Because They Carry Irreversibility

The least exposed roles in the file are not low-tech. They are high-accountability.

The Least Exposed Roles

Role Estimated AI replacement rate What keeps it human
CTO / Technical Director 15% Strategy, leadership, business alignment, and org design
System Architect 20% Global tradeoffs across large systems and services
Engineering Manager 20% Hiring, coaching, conflict management, and delivery alignment
AI Researcher 20% Novel algorithmic work and theory-building
Security Architect 20% Risk, compliance, and system-level defensive design

This is the clearest dividing line in the entire assessment.

AI can generate implementation options. It cannot easily own the irreversible consequences of system-level decisions:

  • service boundaries
  • trust models
  • platform bets
  • staffing design
  • vendor tradeoffs
  • compliance posture
  • or long-range technical debt

That is why architects and technical leaders remain relatively insulated. Their value is not mainly in code production. It is in choosing which code should exist, how the system should behave under pressure, and what the business should accept as risk.

DevOps and Ops Work Are Not Disappearing. They Are Being Consolidated

The source assigns 70% exposure to DevOps engineers and system administrators, and 65% to DBAs. That will sound too high to anyone who only thinks in terms of whether the job title disappears.

But the file’s actual argument is subtler. AI does not need to eliminate the role name to break the labor market. It only needs to reduce how many people are required for the same output.

That is already plausible across:

  • CI/CD setup
  • infra-as-code generation
  • alert handling
  • config management
  • index and query tuning suggestions
  • capacity forecasts
  • routine incident triage

The result is not “no DevOps.” It is “much less DevOps headcount for the same environment.”

This is one of the most important patterns in white-collar automation: the job survives, but the staffing model does not.

Data Work Is Also Being Split by Task Structure

The source puts data engineering, BI, analysis, and warehousing mostly in the 60-70% range, while data science stays closer to 50%, and AI/ML engineering sits lower still.

That makes sense.

The more a role is about:

  • building repeatable pipelines
  • generating reports
  • standard dashboard logic
  • query translation
  • or templated transformations

the more exposed it becomes.

But once the job moves into:

  • causal reasoning
  • experiment design
  • difficult business framing
  • model deployment tradeoffs
  • or domain-specific AI systems

human leverage rises again.

This is the same pattern the source identifies across the whole IT stack: execution compresses first, judgment compresses later if at all.

Security Is Not Safe, but It Is Still Judgment-Heavy

Security roles in the file are mixed.

  • security engineer: 40%
  • penetration tester: 45%
  • SOC analyst (L1): 75%
  • security architect: 20%

That is a useful breakdown because it mirrors what AI does well and badly.

AI is already strong at:

  • alert triage
  • anomaly detection
  • vulnerability scanning
  • log patterning
  • and basic investigative support

It is much weaker at:

  • adversarial creativity
  • systemic threat modeling
  • control design
  • and crisis judgment under ambiguity

That is why the first-layer SOC role gets hit hard while the architecture layer holds up.

Product, Design, and Support Are Being Repriced

The file is also strong on the non-engineering side of software work.

Roles such as:

  • product manager
  • UI designer
  • UX researcher
  • interaction designer
  • L2 support
  • network engineer
  • project manager
  • scrum master

are mostly placed in the 30-65% exposure range, not the extreme top or bottom.

That is the right shape. These jobs are not immune, but they still retain value where they touch:

  • stakeholder management
  • prioritization under tradeoffs
  • user insight
  • system coordination
  • or complex troubleshooting

What shrinks is the execution support around them:

  • report generation
  • ticket triage
  • component drafting
  • prototype generation
  • status updates
  • and workflow reminders

So the role survives, but the human contribution gets pulled upward toward interpretation and coordination.

AI-Native Roles Are the Main Counter-Trend

The source’s most optimistic section is the AI/ML cluster:

  • AI/ML engineer: 35%
  • NLP engineer: 40%
  • computer vision engineer: 40%
  • AI researcher: 20%
  • MLOps engineer: 35%

These roles are exposed to AI as a tool, but not displaced by it in the same way because they live on the side of the stack that builds, deploys, tunes, and governs AI itself.

That is why the file explicitly treats GenAI Engineer and MLOps Specialist as high-growth roles. This is the same pattern seen in other digital sectors: the people closest to the orchestration layer gain leverage while the people closest to the repetitive execution layer lose leverage.

The Real Break Is Not “Coding vs Not Coding.” It Is Judgment vs Throughput

The strongest way to read this source is not by job title. It is by task type.

High-exposure tasks

  • scaffolding
  • repetitive implementation
  • first-pass troubleshooting
  • standard testing
  • routine ops
  • document and dashboard generation
  • basic monitoring and reporting

Lower-exposure tasks

  • architecture
  • technical strategy
  • multi-system integration
  • security design
  • stakeholder negotiation
  • product prioritization
  • org leadership
  • AI lifecycle management

That is the core lesson. AI does not care about prestige. It cares about structure.

What This Means for the IT Labor Market

Software and IT are not being “destroyed by AI.” They are being re-tiered.

The bottom tier loses first:

  • junior coders
  • manual testers
  • first-line support
  • routine BI
  • basic admin-style ops work

The middle tier gets compressed:

  • full-stack development
  • standard DevOps
  • data engineering
  • support engineering
  • applied security

The top tier stays more resilient:

  • architects
  • technical leaders
  • security architects
  • AI-native engineers
  • product and platform decision-makers

That is why the sector feels so different from one end to the other. A new graduate and a principal architect are technically in the same industry, but they are no longer standing in the same labor market.

What This Means for Kane

The source is unusually direct on this point. Kane’s strongest zone is not low-end coding labor. It is the part of the software stack with lower exposure:

  • system architecture
  • requirements definition
  • cross-functional product judgment
  • technical decision-making
  • AI integration into business workflows

That is the right conclusion.

The most commoditized offer in this market is “I can build the straightforward app or dashboard.” The stronger offer is:

  • “I can decide what should be built.”
  • “I can design the system.”
  • “I can turn AI into business operations.”
  • “I can connect product, engineering, and workflow automation.”

In this sector, AI does not erase the need for expertise. It raises the bar for what kind of expertise still prices well.

Sources