How to learn AI fast enough to put it on your CV.

The four tools worth learning in 2026

Learn one from each category deeply. Skip the rest.

The 14-day learning plan

1. Days 1–2
   Fundamentals (3 hours max)

   Watch Andrej Karpathy's "Intro to LLMs" on YouTube (1 hour). Read OpenAI's prompt engineering guide (30 mins). Do one free ChatGPT prompt-writing course. That's it. Stop.

2. Days 3–4
   Pick a boring task

   Find a repetitive, well-defined task inside your target role — brief drafting, meeting summarising, request triage, data cleanup. Boring is the point: it's the task the team actually pays humans to do.

3. Days 5–10
   Build the tool

   Use Lovable, n8n or a ChatGPT Custom GPT. Ship a v1 you can hand to a friend without explaining. Iterate on real inputs, not toy examples.

4. Days 11–14
   Document the proof

   Three-minute demo video. One-page case study with a quantified outcome. CV bullet. LinkedIn post. Interview script. This is the bundle recruiters click.

Why courses alone don't work

A Coursera certificate says you watched videos. A working link says you can ship. Recruiters at competitive schemes see hundreds of CVs claiming "AI proficiency"; the applicants who cut through send a URL that opens a working tool. Free courses are fine as background reading — but if your evidence is a certificate, you're competing against the median. If it's a demo, you're competing against no one.

Frequently asked

Do I need to learn to code to use AI at work?

No. The AI tools employers care about — ChatGPT, Claude, Gemini, Copilot, Notion AI, Lovable, n8n — all have no-code or low-code interfaces. The bottleneck for graduates isn't Python; it's knowing which boring task to point AI at and being able to prove the outcome. If you can write a clear brief in English, you can build a useful AI tool in an afternoon.

Which AI tools should a graduate actually learn in 2026?

Four categories cover 90% of graduate-scheme use cases. Chat & reasoning: ChatGPT (GPT-5) and Claude. Document work: NotebookLM for research synthesis, Claude Projects for long briefs. Building: Lovable for internal tools and web apps, n8n or Zapier for workflow automation. Meetings: Otter or Granola for notes. Learn one from each category deeply rather than dabbling in twenty.

How long does it take to learn AI well enough to put on a CV?

Two weeks of focused build time — roughly 20–30 hours — is enough to ship a working tool tied to a real job task. That's the whole thesis of the CareerGlowUp Build. Passive courses take longer and produce less; you learn AI by using AI on a real problem, not by watching someone else use it.

Are AI courses on Coursera or LinkedIn Learning worth it?

For fundamentals (what is a model, what is a prompt, what is a token) they're fine — most are free and 2–4 hours. For proving skills to employers, they're weak signal: a certificate says you watched videos, not that you built anything. Employers rank a linked demo of a working tool above any certificate. Use free courses for background; use the Build for evidence.

What's the fastest way to prove I know AI on a job application?

Ship one small, boring, useful tool tied to the target role, record a three-minute demo video, and put the link in your CV, LinkedIn, and cover letter. That single asset outperforms every combination of course certificates and buzzwords because it's verifiable — the recruiter clicks the link and sees the thing work.