Real things.
Actually built.

A customer service manager built a workflow that reads every inbound email, classifies it (enquiry, complaint, invoice, etc.), drafts an initial response, and routes it to the right team member — all before a human sees it.

Every recorded meeting gets automatically transcribed, sent to Claude, which extracts a summary, lists action items by owner, and posts both to the team Slack channel and individual to-do lists.

An HR team uses AI to write job descriptions from a short brief, post them, then screen inbound CVs against the role requirements — producing a ranked shortlist with reasons before any human reviews.

A consulting firm built a tool that reads a client brief, pulls relevant case studies from their library, and generates a first-draft proposal with executive summary, methodology, and indicative pricing — in the firm's exact tone and format.

An individual built a personal tool where they paste their payslip, and the AI decodes every deduction, explains what it means, calculates their true take-home in different scenarios, and builds a suggested monthly budget from it.

An investor built a workflow that automatically downloads quarterly earnings reports when published, runs them through Claude, and delivers a structured summary — key metrics, management commentary highlights, and risks — before markets open.

An accounts team built a pipeline that reads all inbound invoices, checks them against purchase orders, flags mismatches, and drafts the supplier communication for any discrepancy — eliminating hours of manual matching work monthly.

A content team writes one long-form article and runs it through an AI pipeline that generates 10 social posts, 3 email newsletter versions, a LinkedIn article, and a thread format — all adapted to each platform's voice and format automatically.

A marketing team built a library of master prompts that capture their exact brand voice, tone guidelines, and formatting rules. Every piece of content — from product descriptions to email subject lines — gets run through the same template, producing on-brand first drafts every time.

A training company replaced expensive video production with an AI pipeline: scripts written by ChatGPT, narration by ElevenLabs, visuals from Midjourney, assembled into finished videos — at a fraction of the traditional cost.

A PhD student built a process where Perplexity surfaces 20 relevant papers on a topic, Claude reads and synthesises them into a structured literature review with themes, gaps, and contradictions identified — in an afternoon instead of weeks.

A strategy team built a daily digest that monitors competitor websites, news mentions, and job postings — using AI to identify signals of strategic moves, product launches, and hiring patterns — delivered as a morning briefing every day.

An individual built a personal AI workflow: describe symptoms, get a list of likely questions the doctor will ask, relevant medical history to mention, and questions worth raising. They go into appointments better informed and more productive.

Job seekers paste a job description and their CV, and Claude generates likely interview questions (technical, behavioural, role-specific), model answers, and then runs a mock interview — asking questions, evaluating answers, and giving feedback in real time.

A parent set up a personalised AI tutor for their child — with a system prompt defining the child's age, level, learning style, and areas of struggle. The AI explains concepts at the right level, creates practice problems, and never just gives the answer.

Developers are building complete production applications — backend, frontend, database, and deployment — using Claude Code or Cursor as the primary builder. The developer acts as architect and reviewer, the AI writes most of the code.

A business analyst connected Claude to their company's PostgreSQL database via MCP. They type questions like "which customers haven't placed an order in 90 days?" and get the answer — no SQL knowledge required. The AI writes and runs the query.

A SaaS company deployed an AI agent trained on their documentation, support history, and product knowledge. It handles 70% of inbound tickets autonomously — answering questions, troubleshooting issues, and escalating only genuinely complex cases to humans.

A solo consultant runs what used to require a 4-person team. AI writes the proposals, handles client onboarding emails, produces reports, creates presentation decks, and manages social content — all from a set of master prompts. The human does the thinking; AI does the execution.

A product manager with no coding experience described what they wanted in plain English: "a dashboard that shows my team's weekly output, with a table and chart." v0 generated the React component. Claude Code built the backend. It was live in three days — no developer hired, no agency engaged.

Before a major pitch, a strategy consultant runs a Deep Research prompt: "Analyse our competitor's pricing strategy, recent product launches, and customer complaints across review sites." Twenty minutes later they have a sourced, structured brief — intelligence that used to take a junior analyst two days to compile.

Students and professionals upload dense PDFs — academic papers, industry reports, training manuals — to NotebookLM and generate a two-host AI audio discussion of the content. They listen while commuting instead of reading. The "Audio Overview" feature creates something that sounds like a real podcast episode.

A niche newsletter operator launched a companion podcast without a microphone or recording setup. Each week: Claude writes the episode script from the newsletter, ElevenLabs narrates it with a cloned host voice, Sora generates a short video clip for social. The whole production takes 45 minutes.

A B2B sales team built an AI agent that monitors inbound form submissions 24/7. The agent looks up the company using Perplexity, scores the lead against their ICP criteria, drafts a personalised first email, and — if the lead scores high enough — books an intro call automatically. Humans only see pre-qualified, pre-scheduled meetings.