Is My Job Safe From AI?

It's a split picture. The parts of the job that are really 'build a spreadsheet and summarize it' are already being done by AI tools inside Excel, Bloomberg, and standalone copilots. The parts that involve defending a number in a room are not.

Typical rung and task breakdown →

The execution half of the job — copy, creative variants, campaign reporting — is heavily automated already. The strategy half — knowing which campaign to run and why it will or won't work for this specific audience — is not, and is becoming more valuable precisely because execution got cheap.

Typical rung and task breakdown →

This is one of the more exposed frontline roles. Chatbots and AI voice agents now resolve a large share of tier-1 tickets — password resets, order status, simple refunds — without a human touching them. What's left is disproportionately the hard, emotional, or ambiguous cases.

Typical rung and task breakdown →

The document-heavy half of the job is genuinely at risk — contract review, discovery, and legal research are some of the strongest current use cases for AI. The client- and court-facing half, and the professional judgment about what actually matters in a filing, is not.

Typical rung and task breakdown →

Production design — resizing, templated social assets, stock-style illustration — is now genuinely faster and cheaper with AI. Original art direction, brand systems, and understanding why a design should look a certain way for a specific audience are not replaceable the same way.

Typical rung and task breakdown →

Scheduling and inbox triage are largely solved by AI tools now, but the trust-based, judgment-heavy parts of the role — knowing what an executive actually needs before they ask — are harder to replace than most people assume.

Typical rung and task breakdown →

Bookkeeping-level accounting — reconciliation, categorization, standard journal entries — is one of the most automated categories of white-collar work today. Advisory work — tax strategy, audit judgment, client-specific guidance — is not.

Typical rung and task breakdown →

Scripted, repetitive test execution is heavily automated by AI-assisted testing tools now. Exploratory testing — the ability to think like a user and find the bug nobody scripted for — is still a distinctly human skill.

Typical rung and task breakdown →

This one splits down the middle. The mechanical coordination — status reports, schedule updates, meeting notes — is being absorbed by AI features inside the tools PMs already use. The harder half of the job, aligning stakeholders and making tradeoff calls under pressure, is not, and that half is what companies keep paying for.

Typical rung and task breakdown →

The production side — writing queries, cleaning data, building dashboards — is automating faster than almost any analytical skill set. What isn't automating is knowing which question to ask, whether the data can actually answer it, and when a suspicious number means a broken pipeline rather than a real trend.

Typical rung and task breakdown →

Volume writing is exposed; voice isn't. Commodity copy — product descriptions, SEO posts, routine email sequences — is already AI-generated at most companies. Original brand voice, campaign concepts, and writing built on real customer reporting still need a human, and they're where the remaining budgets are concentrating.

Typical rung and task breakdown →

Less than the headlines suggest. Screening, policy questions, and onboarding paperwork are automating quickly, but the situations that define the role — investigations, terminations, conflicts between a manager and their report — are ones companies explicitly want a human handling, for legal reasons as much as human ones.

Typical rung and task breakdown →

The mechanical half of this job is largely gone already. Transaction categorization, bank reconciliation, and invoice data entry are already handled natively by tools like QuickBooks and Xero. What's left — judgment calls on messy transactions, catching fraud, and being the person a client can call — is harder to replace.

Typical rung and task breakdown →

Few office roles are more exposed than this one. OCR plus LLM extraction now reads a clean, printed invoice or form and puts the right values in the right fields about as reliably as a person typing them in. What's left is the messy input and the accountability for getting it right.

Typical rung and task breakdown →

More exposed than Executive Assistant, less than tier-1 customer service. AI tools now handle scheduling, inbox drafting, and data entry well, but a single-location office still needs a person physically present to greet visitors, handle mail, and make judgment calls when something breaks.

Typical rung and task breakdown →

The production half of the job — writing code from a clear spec — is now genuinely fast and cheap with AI coding assistants. The systems half — architecture decisions, debugging novel failures, and translating a vague business need into a working product — is not, and that's the half clients and employers still pay for.

Typical rung and task breakdown →

The answer changes a lot depending on which kind of translation you do. High-volume document work between common language pairs — marketing copy, manuals, UI strings — is already handled by neural MT and LLMs at near-human quality. Certified, legal, medical, and live interpreting work is not, because liability and real-time judgment still require an accountable human.

Typical rung and task breakdown →

Unevenly, but yes in places. Self-checkout kiosks, in-store apps, and chatbots now handle a lot of the 'do you have this in stock' and basic pre-purchase questions that used to eat up floor time. Actually selling, handling money, and being a visible presence on the floor are much harder to replace.

Typical rung and task breakdown →

Will AI replace recruiters? Not entirely, but yes on the sourcing and screening side. AI now reads resumes, ranks candidates, and drafts outreach messages faster than any human — including the tools recruiters themselves rely on. What it can't do is read whether a candidate will actually survive this team, or talk someone off the ledge during a counteroffer.

Typical rung and task breakdown →

Will AI replace insurance claims adjusters? For straightforward claims, largely yes — AI now processes routine auto and property claims end-to-end using photo damage assessment and policy-matching, often without a human touching the file. Complex, disputed, or high-value claims still need a person who can investigate and make a defensible call.

Typical rung and task breakdown →

Will AI replace loan officers? In part. Automated underwriting engines already approve or deny straightforward loans in minutes using credit data alone, cutting out much of the manual file review loan officers used to do. What's left is guiding borrowers through complex or borderline cases and actually winning the business.

Typical rung and task breakdown →