The AI experiments that failed and what I learned.

I was going to build an AI-powered lead qualification system that would score incoming enquiries, categorise them by fit, and automatically draft personalised responses. I would wake up to a tidy inbox full of pre-written replies ready to send.

A Google Form that fed into a Sheet, with a ChatGPT prompt that was supposed to analyse each response and populate scoring columns.

ChatGPT could not directly access my Google Sheet in any reliable way. I tried Zapier, Make.com, and Apps Script that would pass data to the API. Each integration had its own quirks, rate limits and mysterious failures.

When I finally got something working, the scoring was wildly inconsistent. One day it would give a consulting enquiry a 3/10. The next day, the same type of enquiry would get an 8/10. The AI had no memory, no consistency, and no actual understanding of what made a good lead for my business.

The personalised responses were generic garbage. I could have written that template in 30 seconds without AI.

1. AI cannot maintain context across disconnected systems. Each time the API runs, it starts fresh unless you feed it the right context every time.
2. Simple automation beats complex AI. Basic Google Sheets formulas worked better because they were deterministic.
3. Personalisation at scale is usually fake. Real personalisation requires knowing the person. AI personalisation is often just mad-libs with nicer wording.

I use a simple form with smart conditional logic that routes people to different follow-up paths based on their answers. No AI. No scoring algorithms. Just clear questions and honest triage.

Take my podcast episodes and automatically turn them into blog posts, LinkedIn posts, email newsletters and social media snippets. One episode becomes 20 pieces of content without lifting a finger.

A workflow where I fed ChatGPT the podcast transcript and asked it to create all these different formats in one go.

The blog posts read like they were written by a robot. Because they were. The tone was off, the flow was choppy, and key points got lost or over-simplified.

The LinkedIn posts were worse. They all had the same structure: provocative question, three bullet points, call to action. After posting three of them, people started asking if I had been hacked or had outsourced my content.

The email newsletters did not sound like me at all. Subscribers started unsubscribing.

1. Your voice is harder to replicate than you think. AI can mimic surface-level style, but not the rhythm, examples and small asides that make content sound like you.
2. One-shot generation is lazy and it shows. Asking AI to do everything at once gives you AI’s default outputs.
3. Repurposing is not automation. It is editorial judgement. You still need to decide which ideas are worth expanding and which formats need more context.

• I use AI to transcribe and create a rough outline of key themes.
• I personally pick which themes deserve to become content.
• I write the first draft myself or heavily edit what AI produces.
• I use AI for specific tasks like shortening text or suggesting headline variations.
• Every piece gets a final “does this sound like me?” check before publishing.

The result takes longer, but people actually read it.

I would record client meetings, have AI transcribe them, and automatically generate summaries with action items that I would send to the client. Professional, efficient, impressive.

I used Otter.ai for transcription and ChatGPT to create summaries from the transcript.

The summaries were technically accurate but completely useless. They captured what was said, but not what mattered.

AI would note that “Steve mentioned possibly updating the requirements document by next Friday” when what actually happened was that the client pushed back hard on that timeline and we agreed to revisit it in two weeks.

It would list every topic in chronological order, making a 45-minute meeting look like we covered 15 things, when really we spent 40 minutes on one core issue. Clients did not read them.

1. AI captures words, not meaning. It does not understand emphasis, pushback, body language or unsaid context.
2. Summaries need editorial judgement. What matters is what changed, what was decided and what happens next.
3. Clients want your interpretation, not a transcript summary. They were in the meeting too. They need clarity.

I take notes during the meeting and spend 10 minutes afterward writing a short human email:

• Here is what we agreed.
• Here is what changed from our last conversation.
• Here is what you are doing, and here is what I am doing.
• Here is when we are talking next.

AI sometimes helps me structure rough notes, but the thinking and framing are mine.