aicroft
Playbooks · 4 workflows, all runnable

Recipes, not articles. Tools named, costs included.

Each playbook is a workflow we've run ourselves: the problem, the exact steps, the real monthly cost, and the outcome you can verify. Run it yourself — or have us build it for you.

01Everyone
About an hour₹0/monthBeginner

Private AI on your own laptop

A ChatGPT-style assistant that runs on your machine, costs ₹0/month, and never sends your data anywhere.

The problem

Subscriptions add up, and some work — client documents, finances, unpublished writing — shouldn't leave your computer at all.

The stack — by name
  • Ollama
  • Llama 3
  • Open WebUI
Watch the demoHave us build it for you
The steps
  1. 01

    Install Ollama

    One download from ollama.com (Mac, Windows, Linux). It manages models the way an app store manages apps.

  2. 02

    Pull an open model

    `ollama run llama3` downloads Meta's Llama 3 — capable enough for drafting, summarising, and everyday questions.

  3. 03

    Add a chat interface

    Open WebUI gives you a familiar chat window on top, with history and file uploads — all still local.

  4. 04

    Move one real task over

    Pick something private — summarising client notes, drafting invoices, rewriting emails — and make local AI its home.

You end with

A working private assistant on your own hardware — and the confidence that you can run AI without renting it.

02Creators
A weekend to learn₹0 with free tiersBeginner

A finished reel from raw footage

Turn a folder of clips and a rough idea into a publish-ready short — without learning an editor's timeline.

The problem

Footage piles up because editing is the bottleneck: structure, cuts, music sync, and captions take hours of skill you may not have.

The stack — by name
  • ChatGPT or Claude
  • CapCut
  • ffmpeg
Have us build it for you
The steps
  1. 01

    Script with an assistant

    Give ChatGPT or Claude your rough idea and clip list; ask for a beat-by-beat structure — hook, build, payoff — under 45 seconds.

  2. 02

    Cut to the beat

    CapCut's auto-captions and beat-marking do the mechanical work; your job is choosing which clip answers each beat.

  3. 03

    Automate the assembly

    Going further, a script pipeline (ffmpeg + a beat-detection step) can cut clips to music automatically — we publish ours.

  4. 04

    Repurpose once it works

    The same structure becomes a YouTube short, a LinkedIn clip, and a carousel script — one edit, three posts.

You end with

A repeatable production line: raw folder in, publishable short out — in an evening instead of a weekend.

03Small business
One afternoon≈ ₹1,600/monthIntermediate

First-draft replies for your inbox

Every routine customer email arrives with a draft reply already written — you approve, adjust, send.

The problem

Half of business email is the same ten questions in different words. Typing the answers again is hours of skilled time spent on unskilled work.

The stack — by name
  • Gmail
  • Claude or ChatGPT
  • n8n or Zapier
Have us build it for you
The steps
  1. 01

    Collect your real answers

    Pull your 20 most common questions and the replies you actually sent. That's the voice the AI must match.

  2. 02

    Write the standing instruction

    A single prompt in Claude or ChatGPT that carries your tone, policies, and the rule: 'if unsure, say a human will follow up.'

  3. 03

    Wire it to your inbox

    n8n (self-hosted, free) or Zapier watches incoming Gmail, drafts a reply with the AI, and saves it as a Gmail draft — never auto-sends.

  4. 04

    Keep the human approval

    You read, edit, send. The AI does the typing; you keep the judgment and the relationship.

You end with

Around four hours a week back, faster replies for customers, and a tone that's still unmistakably yours.

04Professionals
30 minutes to set up₹0 on free tiersBeginner

The daily AI operating routine

Five small AI habits that compound — planning, drafting, summarising, learning — without adding a single new app.

The problem

Most people use AI like a search engine: one-off questions, no memory, no compounding. The value is in routines, not queries.

The stack — by name
  • ChatGPT or Claude
  • Notion
  • Your calendar
Have us build it for you
The steps
  1. 01

    Morning: plan in one message

    Paste your calendar and task list into ChatGPT or Claude; ask for the day in three priorities and the first 30-minute block.

  2. 02

    Draft everything once

    Emails, updates, proposals — write the ugly version fast, let the assistant shape it, then make it yours. Never send raw output.

  3. 03

    Summarise instead of skipping

    Long reports and threads go to the assistant first: 'five bullets and the one decision I need to make.'

  4. 04

    Learn with a tutor, not a feed

    Pick one topic a week and interrogate it — 'explain like I know X but not Y' beats an hour of videos.

  5. 05

    Keep a prompt shelf in Notion

    When a prompt works, save it. Ten good saved prompts are worth more than a hundred tips.

You end with

A personal operating routine where AI quietly handles first drafts and summaries — and you notice the week getting lighter.

Missing the workflow you need?

Tell us the job — “I want to turn ___ into ___” — and the best requests become the next published playbook.

Request a playbook
@aicroft

Rather have it built for you?

Every playbook here is something we install for businesses — scoped, costed, and handed over working. The first call is free.

Book a free AI callSee how we work

replies within 48 hours · aicroft.in@gmail.com