Ismail
Islam

Problem

Leads fall through the cracks. Follow-ups get forgotten, no-shows go unaddressed, and manually tracking every contact across a pipeline doesn't scale.

Solution

A fully automated 4-stage sales pipeline: lead ingestion, AI-written follow-up sequences, meeting booking detection, and automatic no-show re-engagement. All running on its own with Google Sheets as the CRM.

How it works

Leads are pulled from a source sheet into the CRM automatically. A follow-up agent uses Claude to write personalised emails and sends up to 3 of them, 2 days apart. When a prospect books via Calendly, follow-ups stop. If they don't show up, a no-show agent fires a re-engagement email and updates the record.

Problem

I'm a content creator who doesn't use social media. No TikTok, Instagram, Snapchat, none of it. I know how distracting and unproductive it is, so I cut it out entirely. But I still needed to stay on top of everything happening in AI without spending hours hunting for it manually.

Solution

Built a fully personalised AI newsletter that completely replaces what social media would have given me, but with zero noise. Tailored to exactly what I care about, delivered on schedule, nothing irrelevant.

How it works

Runs on a cron schedule, pulls from configured AI and tech sources, summarises with AI, formats into a clean newsletter edition, and delivers directly to my inbox. Always informed, always productive.

Problem

Email overload kills focus. Triaging, categorising, and drafting replies to dozens of emails daily is a cognitive drain.

Solution

An AI agent that monitors the inbox, classifies each email by urgency and category, drafts responses for approval, and handles routine replies autonomously.

How it works

New emails trigger the workflow. AI classifies intent and priority, generates a context-aware draft, and either sends automatically for low-stakes emails or surfaces the draft for review.

Problem

Finding specific information across hundreds of documents requires manual searching, slow, error-prone, and scales badly.

Solution

A retrieval-augmented generation agent that ingests Google Drive documents into Supabase vector store and answers natural language queries with source citations.

How it works

Documents are chunked and embedded into Supabase pgvector. When queried, the agent retrieves semantically relevant chunks and synthesises a grounded answer, no hallucination, always sourced.

Problem

Producing consistent short-form video content is time-intensive, scripting, editing, and publishing daily is unsustainable manually.

Solution

An end-to-end pipeline that generates scripts, produces voiceovers via ElevenLabs, assembles video assets, and schedules YouTube Shorts for upload.

How it works

Trigger fires on schedule or input topic. AI generates a short-form script, ElevenLabs synthesises audio, Creatomate assembles the video, and the workflow uploads directly to YouTube via API.

Problem

Producing UGC-style ad creatives at scale requires actors, studios, and significant budget, out of reach for most brands.

Solution

An AI video ad pipeline using Google Veo and OpenAI Sora to generate photorealistic UGC-style footage from prompts and assemble complete ad creatives.

How it works

Script and product brief enter the workflow. AI generates scene prompts, Veo and Sora render video segments, assembly pipeline stitches clips with captions and music, outputting a finished ad.

Problem

Manual invoice processing is repetitive, error-prone, and creates accounting bottlenecks, especially with high invoice volume.

Solution

An AI document processing pipeline that extracts invoice data, validates fields, and pushes structured records into the appropriate accounting system automatically.

How it works

Invoice arrives via email or upload trigger. AI vision model extracts vendor, amount, date, and line items. Workflow validates data integrity and posts to the accounting integration with zero manual touch.

Problem

n8n workflow failures generate cryptic errors that require manual debugging, halting automation pipelines and wasting diagnostic time.

Solution

A meta-workflow that catches errors across all pipelines, feeds the error context to Claude AI, and returns a plain-English diagnosis and suggested fix.

How it works

n8n's built-in error trigger fires on any workflow failure. The error payload, node context, and execution trace are sent to Claude, which analyses the root cause and delivers a structured fix recommendation, removing guesswork from debugging.