Learn to do your own job with AI — task by task

Not a course about AI. You practice the tasks your job is actually made of, in the AI tool your employer already allows, and every result is checked against what it has to contain. About fifteen minutes a day.

Take the 3-minute quiz

How it works

  1. Diagnose

    See which tasks of your job AI already does or helps with, from the O*NET task list, and roughly how many hours a week that adds up to — as a range, with the method. AI and every job

  2. Quiz

    Twelve short screens, about three minutes: what you need, which AI tool you may use, where your week goes, and three quick checks of how you work with AI.

  3. Path

    Your tasks first, at 10, 15 or 25 minutes a day. Each module: why it matters, a worked example, the task by hand and then with AI, a check, and a template to keep.

  4. Proof

    Pass a task and catch the mistake planted in an AI answer, and you get a verified skill record: a public page showing what was checked, when and how, that you can add to LinkedIn.

  5. Pay — only if it is worth it

    The quiz, the plan and the first two modules are free, with no card. CVBLADE Pro opens the rest, and you decide after you have done the free ones.

Why it is different

Your tasks

Every path is built from the O*NET task list of one occupation — 819 of them. A bookkeeping clerk practices reconciling a vendor statement, not writing a poem about spreadsheets.

Your tool

Copilot, ChatGPT, Gemini or Claude — whichever your employer allows. We hand you the material and the prompt; you run it where you will use it at work, so the skill comes with you.

Checked by practice

You paste your result and it is checked against the facts it must contain and the mistakes it must not. Then you find the error planted in an AI answer — because catching what AI gets wrong is half the skill.

Honest numbers

Time saved is your own measurement: a slice of the task by hand, then the whole task with AI. Hours estimates are ranges with their method. We never promise a job or a raise.

Start with an office job

812 occupations have a path today. These were written and tested first.

Tasks: O*NET Database 31.0. Hours: a range from published experiments applied to each job’s task mix — see the method on each plan.

What it costs

Free

The diagnosis, the quiz, your readiness score, your plan, the first two modules of your path and one verified record. No card.

CVBLADE Pro

Every module of every path, verified records for each one you pass, and the market data and job radar. No trial that turns into a charge, no weekly plans.

Laid off?

Everything in Pro is free for 90 days. No card, and we take your word for it — when the 90 days end there is nothing to charge.

What happens to your data

Your answers stay on your device

The tasks you pick, your plan and what you paste into a module stay in your browser. We receive an anonymous summary with no name or email attached: occupation, state, score, what you need, how often you use AI and which tool. Your résumé, if you checked one, never left your browser in the first place.

Your employer sees training, never you

If your company pays for your seat, it sees modules passed, verified skills and time saved. Never your quiz answers, your résumé, or anything about a job search.

Training a team?

A free ten-minute AI readiness audit for your team: which tasks of their roles AI touches, and how many hours a week that could add up to. Seats from three people, priced on the page.

CVBLADE for teams

Sources

Occupations and their tasks: O*NET Database 31.0. Pay: BLS Occupational Employment and Wage Statistics, May 2025. The hours ranges apply the published results of controlled experiments: support agents with an AI assistant resolved 14% more issues an hour, and newer agents 34% more (Brynjolfsson, Li and Raymond, 2023); professionals finished writing tasks 40% faster with 18% higher quality (Noy and Zhang, Science, 2023); consultants were 25% faster on tasks AI handles well — and 19 percentage points more likely to get wrong the ones it does not (Dell’Acqua et al., 2023). That last finding is why every module ends with catching the AI’s mistake.