Direct Answer
Charles Broomfield's video is persuasive because it is not built around a benchmark. It is built around three changes he experienced at work: organizational redesign, compressed software creation, and compressed creative production. Together, they show why waiting for an AI to reproduce an entire job title is the wrong test. Businesses can remove roles by redesigning workflows, and a capable person can now produce more without buying every specialist hour that production previously required.
The evidence does not support certainty that most knowledge jobs disappear in five years. It does support a narrower warning: exposure is rising, entry-level pathways appear especially vulnerable, and organizations are explicitly planning around smaller teams. The useful response is preparation, not panic.
The key distinction: AI rarely needs to replace every task in a job before it changes headcount. It only needs to remove enough work, shorten enough cycles, or make a different organizational design economical.
Watch Charles Broomfield's Video
Video and experience credit: Charles Broomfield, published through Charles Level Up on 15 July 2026. Follow him on LinkedIn. The career history and personal results below are his account, not statements from Google, AWS, or his other employers. This article independently checks the broader claims and adds a practical response framework.
Source Note: Testimony, Company Data, and Research Are Different
Broomfield describes work in national security, AWS security risk management, Capital One, and Google abuse detection. Those details establish the perspective he brings to the video, but they do not make every prediction an employer position or a measured result. This article separates three evidence layers:
- Personal testimony: what happened to his team, what he built, how long it took, and what creative demo he saw.
- Company-confirmed facts: Amazon's published workforce reductions, CEO statements, and financial results.
- Independent labor evidence: current research on exposure, productivity, and employment patterns.
That separation matters. A first-hand story can reveal a mechanism before a large dataset can measure it, but it cannot establish how common the mechanism is.
The Three Warning Signs
1. A role can disappear before AI can perform it end to end
Broomfield says his AWS team managed systemic security risks that crossed organizational boundaries. He initially described his layoff as being replaced by AI, then corrected the framing: the work was too collaborative to automate directly, but the team could still be treated as a manual bottleneck while the surrounding process was redesigned.
This is the strongest idea in the video. The unit of automation is not always the job. It can be a handoff, approval layer, reporting cycle, queue, or coordination mechanism. Remove or standardize enough of those pieces and a company may need a different shape of team even when no agent can reproduce the old role.
2. An eight-hour prototype changes the cost of trying
After the layoff, Broomfield says he built a mobile-friendly expense tracker in about eight hours despite having never built an app or programmed in JavaScript. The prototype included authentication, Google sign-in, categorized expenses, notes, editable history, import and export, charts, filters, and cloud backup.
That does not prove the app was production-ready or that a non-developer can safely ship financial software. It does prove something economically important: the cost of reaching a convincing prototype has fallen. A founder can test an idea before hiring a team. An internal operator can build a workflow before entering a procurement cycle. A small company may buy fewer external hours for exploratory work.
The new bottleneck moves toward requirements, judgment, verification, security, distribution, and maintenance. Fast generation raises the value of knowing what should exist and how to prove it works.
3. Creative automation reaches beyond text
The third experience was a demonstration that turned part of a script into polished motion graphics aligned with a creator's visual style. Broomfield compares that output with visual work that would previously have taken him hours or cost hundreds of dollars from an experienced editor.
The near-term change is not that every editor disappears. It is that rough assembly, asset generation, basic animation, format adaptation, and variation become cheaper. Skilled creative work still includes story structure, pacing, rights clearance, visual continuity, factual review, taste, and final quality control. But when one person can cover more of the pipeline, staffing and pricing pressure follows.
Claim Audit: What Holds Up and What Needs Qualification
| Claim in the video | Verdict | What the record supports |
|---|---|---|
| Amazon cut about 30,000 roles across the last two rounds. | Confirmed | Amazon announced approximately 14,000 corporate reductions in October 2025 and 16,000 more in January 2026. |
| AI is part of Amazon's workforce logic. | Confirmed | CEO Andy Jassy wrote that AI efficiency is expected to reduce Amazon's total corporate workforce over the next few years. The October 2025 reduction note also connected a leaner structure with the speed enabled by AI. |
| AWS revenue reached about $128.7B, up 20% in 2025. | Confirmed | Amazon's full-year results report AWS segment sales of $128.7B, up from $107.6B in 2024. |
| AWS cut 10% of its workforce. | Not independently verifiable | Amazon publishes segment revenue but not a matching AWS employee series that proves this percentage. |
| Broomfield's team was eliminated only because of AI. | Personal interpretation | The organizational mechanism is plausible and consistent with Amazon's stated direction. No public company source confirms the sole cause of this specific team decision. |
| AI is better than most entry-level software engineers. | Too broad | The app is a meaningful first-hand demonstration, not a controlled comparison. Capability varies by task, codebase, model, security requirements, and review process. |
| Many knowledge jobs will not exist in five years. | Forecast, not fact | Current evidence shows rising exposure and early-career pressure, but not enough to make a reliable five-year occupational extinction forecast. |
What the Best Current Evidence Says
Exposure is broad, but exposure is not job loss. The ILO and NASK estimate that roughly one in four jobs worldwide has some generative-AI exposure, rising to about one in three in high-income countries. Their central conclusion is that transformation is more likely than full replacement because most occupations contain tasks that still require human input.
Measured displacement remains limited, but the distribution is uneven. The ILO's June 2026 review finds real yet uneven productivity gains, while large-scale job displacement remains limited so far. It flags growing inequality, weaker opportunities for younger workers, and changes to autonomy and job quality as the more immediate risks.
Entry-level workers are the clearest warning signal. Stanford's Digital Economy Lab, using ADP payroll data, reports that employment since ChatGPT's release grew in every exposure group, but more slowly in the two most AI-exposed groups. For workers aged 22 to 25, employment declines in exposed occupations have persisted and deepened. The lab also cautions that some earlier movement reflected factors beyond AI, so the data should not be read as a single-cause proof.
Coding productivity is not one number. METR's early-2025 randomized study found experienced open-source developers took 19% longer with then-current AI tools on familiar repositories. Its 2026 update found small productivity benefits with later agents, while retaining substantial uncertainty and selection caveats. Both findings can coexist with Broomfield's rapid prototype: greenfield, bounded app generation is a different task from safely modifying a mature codebase.
Evidence-led conclusion: AI is already changing the economics of tasks and early-career pathways. The size, speed, and permanence of aggregate job loss remain uncertain. Plan for workflow redesign without pretending the labor-market outcome is already settled.
Audit Your Role at the Task Level
Do not ask only, "Can AI replace my job?" Break a normal week into tasks and score each one. This produces a more useful map than a job-title forecast.
| Task type | Typical examples | Near-term response |
|---|---|---|
| Repeatable and digital | Summaries, first drafts, data cleanup, routine analysis, asset resizing | Automate a bounded version and measure time, quality, and exceptions. |
| AI-accelerated, human-owned | Code changes, research synthesis, campaign concepts, financial models | Use AI for options and execution; retain tests, source checks, and accountable approval. |
| Coordination-heavy | Cross-team risk decisions, stakeholder alignment, negotiation, incident response | Document why the coordination exists. Remove needless handoffs while preserving controls that manage real risk. |
| Trust and consequence-heavy | Hiring, legal judgment, security acceptance, medical or financial decisions | Keep named human ownership, evidence trails, and escalation paths. |
| Physical or situational | Field work, care, installation, live facilitation, complex on-site judgment | Use AI around the work for preparation and documentation rather than assuming full substitution. |
For each task, record frequency, time, error cost, data sensitivity, verification method, and the person who remains accountable. The biggest opportunity is usually high-frequency, low-consequence work with a clear acceptance test. The biggest danger is delegating consequential work because the output looks polished.
Three Ways to Prepare Without Pretending You Can Predict 2031
1. Build a realistic financial buffer
Broomfield says his unusually high savings rate and long runway made his AWS layoff less stressful. That is a personal strategy, not a universal target. The transferable principle is to create a cash reserve appropriate to your circumstances. The US Consumer Financial Protection Bureau notes that even a small emergency fund can reduce dependence on credit after an income shock. This is general education, not personal financial advice.
2. Build relationships before you need a referral
Broomfield reviewed his application history and found that referrals made him about five times more likely to receive an interview. That figure is his own sample, not a labor-market benchmark. The durable lesson is still sound: maintain relationships with classmates, colleagues, clients, and collaborators by being useful and reliable long before a job search starts.
A network is not a list of strangers collected during a crisis. It is accumulated trust. Share useful work, make thoughtful introductions, help people solve real problems, and keep lightweight contact with people whose work you respect.
3. Practice the conversion skill: interviewing
A referral can open a door; it cannot prove judgment, communication, or role fit. Broomfield attributes much of his own career mobility to repeated practice speaking, explaining, and interviewing. Build a small evidence bank with six stories: a hard problem, a failure, a conflict, a decision under uncertainty, a measurable result, and a time you changed your mind.
AI can help rehearse, but do not let it manufacture experience. Ask it to challenge vague claims, identify missing metrics, and run follow-up questions. Your examples must remain true and defensible.
A 30-Day Career Resilience Plan
- Days 1-7: map the work. List recurring tasks from the last four weeks. Mark what is repeatable, sensitive, consequential, relationship-heavy, or easy to verify.
- Days 8-14: automate one safe task. Choose a low-risk workflow, define an acceptance test, run it five times, and log time saved plus corrections required.
- Days 15-21: create proof. Turn the result into a one-page case study: problem, old process, new process, safeguards, result, limitations. Remove employer-confidential information.
- Days 22-25: reconnect. Contact five people with a useful update or specific offer of help. Do not begin with an ask.
- Days 26-28: rehearse. Run two mock interviews and improve the six evidence stories. Record yourself so you can hear vague or inflated answers.
- Days 29-30: reduce one point of fragility. Improve your cash buffer, update your portfolio, document a critical skill, or identify one adjacent role that values your domain judgment.
One additional rule matters for anyone experimenting at work: use approved tools and data boundaries. Never paste customer records, credentials, incident details, internal code, or proprietary documents into an unapproved model. Career resilience should not create a security incident.
Video Chapters
- 00:00 - Three experiences that changed Charles's view
- 00:26 - The AWS security role he thought was irreplaceable
- 00:58 - Process redesign instead of direct replacement
- 02:18 - Building an expense app without JavaScript experience
- 03:12 - The eight-hour prototype and its features
- 06:04 - AI-generated motion graphics and creative work
- 08:09 - Why knowledge-work automation feels different
- 09:59 - Three ways to prepare
- 10:40 - Financial resilience
- 11:08 - Building a professional network
- 16:21 - Interview skill as career leverage
- 18:58 - Final preparation checklist
Bottom Line
Broomfield's most useful warning is not that a chatbot will take every job. It is that workflow economics can change faster than titles, training systems, and career assumptions. A team can disappear because the organization wants a different process. A prototype can be built before a specialist is hired. A creator can internalize work that used to be purchased.
The counterweight is equally important. Current evidence does not show a completed job apocalypse, and a polished demo is not a production system. Treat the transition as a live risk-management problem: learn the tools, verify consequential outputs, protect sensitive data, build proof that you can improve a workflow, and strengthen the financial and professional relationships that give you options.
Sources
- Creator source: Charles Broomfield, "I saw the future of AI... it scared me", plus the attached transcript.
- Creator credit: Charles Level Up on YouTube and Charles Broomfield on LinkedIn.
- Amazon CEO Andy Jassy: generative AI and the expected reduction in corporate workforce
- Amazon: approximately 14,000 corporate role reductions, October 2025
- Amazon: approximately 16,000 additional role reductions, January 2026
- Amazon 2025 full-year results: AWS sales of $128.7B, up 20%
- ILO and NASK: one in four jobs potentially exposed to generative AI
- ILO 2026 review of empirical evidence on jobs, productivity, and work organization
- Stanford Digital Economy Lab: current employment effects by AI exposure and experience
- METR: early-2025 AI and experienced open-source developer productivity
- METR Frontier Risk Report, February-March 2026: updated developer productivity evidence
- Consumer Financial Protection Bureau: emergency-fund guide