SpeedRun Ep 05 Projects with SOW from Notes and Emails -Tales from the Sky Lounge
Episode Summary
A consultant demonstrates using Claude Desktop with MCP integrations to automatically generate client proposals by synthesizing meeting transcripts, emails, and web research. The AI system searches Google Drive for templates and meeting notes, analyzes Gmail exchanges, researches the client online, and produces a complete statement of work—all from a single prompt.
Pull Quotes
“What's that worth? To write that good of a response and to synthesize all that data all in one place and then just put it on the paperwork.”
On working with early-stage AI integrations
“Everything's beta. So this is like the bleeding edge of technology. We're going to get cut a little bit. I'm kind of over it, but I do have a solution.”
On AI explainability compared to traditional machine learning
“Here's show your thinking. It's a critical leap forward. It makes everybody a lot more comfortable.”
“I just like Plaude because it's a hardware device and it's just less intrusive on all the Zoom calls to go, 'Oh my god, there's all these note takers.'”
On creating reusable AI workflows
“I have it saved as a project now. So all I have to do is go in here and scope of work, draft agreement, blah blah blah. There we go.”
Transcript
Okay, so the next thing I want to explore is, can we set up templatized projects? So these chats where they're called projects in Claude Desktop. Can we set up a repeatable process? And I thought, okay, cool. I have one client that I pitched and had a couple meetings. I use my Plaude recorder as a hardware device. So I've got this thing set up to where after every meeting, it automatically transcribes and then when the transcription is ready, it takes everything and plops it into a meeting notes folder on my Google Drive for my company. So I said, "Hey, I wonder if we can look at emails and I wonder if we can look at these meeting transcriptions and do anything kind of cool with that." So this is my experiment in that. So now I've got a couple MCP connections, Google Drive. Got one automation that's external to AI. It's just Zapier and it just clocks everything over. But I got this cool little recording device. Could be Firefly Otter, notes, read, whatever you got. Just like Plaude because it's a hardware device and it's just less intrusive on all the Zoom calls to go, "Oh my god, there's all these note takers." So here is that adventure. So I went in, I said, "Hey, prompt me. This is the general thing. Prompt me for a company I want to provide a proposal for. Look in requirements, needs, outcomes, to-do items in all these places. Look in my meeting notes folder on Google Drive. Just find it. Let's see if it works. Any documents record meetings with that company. Look at the last five emails exchange with a client. Surfs the web. Go get some background and write a proposal based on the template that's in my Google drive." I kind of gave up on Notion for a second because the Notion local MCP was getting a little flaky on me. Everything's beta. So this is like the bleeding edge of technology. We're going to get cut a little bit. I'm kind of over it, but I do have a solution. I think remote MCP is going to solve all of this stuff for us. So I said, "Hey, here's a client." And it said, "Okay, cool." Searched the Google Drive, found the template, found some other stuff. Here's meeting notes. Looked in my Gmail, found the people, looked on the web for who they are, what their motivations are, what they write about. Found 10 different articles. I guess I found 10 logos. I don't know. And then it wrote me a proposal and so here it is. Let's check it out. Oh god, this is awesome. And it put my name in there. Said nothing about it. Put the date. That's very courteous. Statement of work. Executive summary. And this is all in the format of that template, right? It did really good stuff and a strategic opportunity. And then here's the things that they said they wanted to do in the meeting. And here's some drill down on that. And boy, spot on with exactly what they wanted. Everybody wants a Genetic AI, some traditional data stuff implementation. Yeah, let's just go ahead and get an onboarding timeline. How cool is this? Engagement terms, blah blah blah. And then off we go. Oh, why partner with my company? Tech CXO conclusion. Yeah, I just signed here. How cool is that? And then it came and it said, "By the way, from your meeting notes, I figured out these things from the proposal stuff. I pulled out these and then from emails and it picked up on this guy and this other guy and figured out what my hourly rate was. Hey, just because then from the web research. How cool is that? And then here's explain your thinking. That's been a real sticking point with ML in the past, before we got all hot on LLMs. Why did you make that decision? I don't know. It just the computer just mathed it and it was all good. But here's show your thinking. It's a critical leap forward. It makes everybody a lot more comfortable. Would you like to make any adjustments? Nah. Looks good. Awesome. Please insert it on the correct template with art and save it on the Google Drive.
What's that worth? To write that good of a response and to synthesize all that data all in one place and then just put it on the paperwork and then so, I have it saved as a project now. So all I have to do is go in here and scope of work, draft agreement, blah blah blah. There we go. And then I've inserted now a couple of different documents that are my service offerings, my CV, couple case studies. What else you got? And then so everything I could throw at this thing is only 5% of the memory. And then I got a couple, here's the prompt which you just saw. Oh, and then I added craft an email. So how cool is that? Anyway, this life is good. This is going to be awesome.
Want to go deeper?
Work With MeRelated Episodes

SpeedRun Ep12 Conference Attendee Email Lead Parsing Tales from the Sky Lounge
A developer experiments with using Claude AI to automate conference lead generation by parsing a PDF attendee list and creating a complete HubSpot marketing campaign. Claude autonomously researches the conference, segments attendees, writes email sequences, and designs lead scoring workflows—but hits output limits when generating the final CSV, revealing practical constraints in AI-driven business development automation.

SpeedRun Ep13 Email Marketing Write an Agent Tales from the Sky Lounge
A CTO demonstrates how Claude AI can autonomously design and build a complete email marketing automation system from a single meeting recording. Starting with meeting notes and a one-line goal, Claude generated technical architecture, wrote Python code, created file structures, and produced sample email sequences—all through conversational prompts without traditional software development.

SpeedRun Ep 09 Resource Use for ICP + GTM Plan -Tales from the Sky Lounge
A fractional CTO demonstrates using Claude AI with deep research mode to generate a comprehensive business development plan. By uploading marketing materials, case studies, and service offerings, the AI produces detailed ideal customer profiles, multi-channel outreach strategies, budget allocations, and personalized contact sequences—illustrating how experienced professionals can leverage AI to automate strategic planning while maintaining the judgment to guide and refine outputs.