You get this email because you subscribed to it at compoundleverage.com or are one of our customers or prospects. If you're no longer interested in receiving our emails, you can unsubscribe using the link in the footer.

I built a proposal automation system for a business owner last month as a pilot project.

Google Docs integration, Google Drive, data pulling, auto-generation into a template that ran from a simple interface - everything she'd asked for to streamline her proposal process.

When she tested it, the results were disappointing. Not because of what I built, but the input provided was missing key elements.

"I see what you were doing, but the output was skinnier than I expected. I went back to copying and pasting in ChatGPT.”

Here's what was missing:

Her concerns were real: She didn't put in upfront work, which is why she got “skinny” results.

But she never asked the critical question: "What does this technology need from me to give me the result I want?"

The whole point of doing a template was to save time on formatting, integrating, and organizing to get a cohesive response that sounds good, so she could increase her capacity.

But without the upfront investment in examples, context, and instructions, the system couldn't deliver what she desired. Her problem lay in her prompt, depth of knowledge, data, and not the pilot solution.

The lesson? If your goal is to reduce your workload with AI, you have to consider what you must do to reach the outcome you desire.

I use this simple framework: Eliminate, Simplify, Automate, Outsource.

You have to think about what the technology needs from you to succeed. The planning before doing that's what transforms time-consuming work into systems that scale.

What task are you trying to shortcut without asking: "What does AI need from me to get the results I want?"

Try this prompt in ChatGPT to plan your next AI project:

"I want to use AI to help with [describe your task]. My goal is to [specific outcome you want]. Help me think through: 1) What context and examples do I need to provide? 2) What constraints should I include? 3) What format do I want the output in? 4) What could go wrong if I don't prepare properly? Ask me follow-up questions to help me understand what you need from me to get the results I want."

Marvin

PS: Whenever you're ready, here are three ways to get your time back: