Dave Killeen: [00:00:00] Hello, Dave Colleen here at Field CPR at Pendo and EMEA, and a very warm welcome back to the 5:00 PM episode three. For those of you just joining, welcome. This is the podcast where I'll be doing deep dives on various different AI tools, tricks, and shenanigans, so that you don't have to, I'll be bringing you super short, practical demos under 10 minutes every week or so to solve some of your biggest challenges by dancing more effectively with ai.

Alright, let's have a little group therapy session first about a few of the core pains of being a product geek. One, we can see ideas perfectly in our head, but we struggle to bring them to life, to tell a story around them that gets stakeholders to enthusiastically buy in. And two, our discovery and validation cycles are just too slow and way too costly.

And by the time we've scheduled interviews, we built clickable prototypes. The market has just already shifted. It just takes too long. And three, most of us struggle to give the LLM the right context [00:01:00] to get anything truly useful back. We push what we think is the right information into the ai, and I call this context pushing.

Today we're going to flip context, pushing on its head. We're gonna be talking about a new method I call, context pulling. This is where you engineer the AI to. Take responsibility to actively pull the required information out of you. Have it ask you for the context it thinks it needs, so that it can get the job done that you're asking you to do.

And so to do this today, we're gonna build a virtual product manager with a prompt that will take the vaguest ideas and turn it into a fully specked working prototype in a matter of minutes. This is here in the arms. Let's dive in. Okay, so my starting point was a very vague idea. Help me create a cost delay calculator for our Pendo sales team to bring into conversations with prospects so that they could figure out that for every month that goes by and they haven't got Pendo.

They're losing out on their ability to increase their revenue, reduce their cost, and reduce their [00:02:00] risk. So to bring this to life, I created a virtual PM prompt. And the prompt's job isn't just to wait for me to give it the perfect spec. Its job is to interrogate the hell outta me until it can figure out what my crappy idea is until that.

Becomes an amazing spec and this is context pulling in action. Okay, so right now we're in Claude Projects. You can do this with other lms, but lemme just show you what I'm doing here. So I go into the instruction section of the Claude Project, and here's the prompt I created using Claude Console. I'll be doing an episode in this quite soon where we.

Give it a very vague prompt and generates a brilliant one back for you. And this is what it generated for me. As you can see, it's structured into kind of four main parts. First, it tells the AI to act as an experienced pm and its initial job is just to ask me clarifying questions. It doesn't do anything else until it's interrogated the hell outta me.

And then once I'm ready, I use that magic command spec doc. And then when that's done, it then triggers the AI to take all of my answers and generate a proper product spec. And then the [00:03:00] next phase is even more bonkers. I then type in design flows, and when it does that, it activates a virtual designer who comes in and maps out all the flows based on our chat.

And then finally, I'll then say, create PRD. And that command bundles everything into one comprehensive document that's ready to go. And that's the structure. It's a simple multi-stage workflow, all controlled by me in one prompt. I'll show you more now. So what I'll first do is I'll open up Claude on the mobile.

And the reason mobile is because you can just drop a whole bunch of voice notes in very easily. It's so effective. Claude doesn't have the ability to drop voice notes in through web and on mobile. It just works really well. Have a look at episode one. There's more detail there about this particular use case.

It's fantastic to watch it. You get a whole ton of leverage from it. And so as you can see, this is just a transcribed voice note. It's not even like. Picking the spelling up correctly. I'm just literally giving it what I want, right? And then it asks me the clarifying questions. And so then what I'll do, and again, why Mobile works great for this is I'll go back in on voice on mobile and go, right for the first question, use your own judgment, blah, blah, blah, blah, for question [00:04:00] two, and you get the idea, right?

So you just go through everything and tell it what you want. And it gets to a point here where actually it's still not happy with me. I'm being told off and saying, I want more. Questions answered. And so I go back in, I answer all of that, and then we go ahead and it specs the doc for me. So that's the first stab at the spec done.

And then down below, then I'll say, okay, design flows. And because the prompt is engineered in the way I just showed you, it will then start designing all the flows for me. And then once that's all there, scroll through all of this as madness. I'll say summarize it then gives me a summary. And then I'll go in and I'll say, please, now can you gimme the entire PRD?

And it just goes through everything. Now, just fast forward a second and now what you see here is the entire PRD. It even got the logic on how to calculate the cost of delay and it's just fantastic. And so I then dump that into lovable. I'll show you that quickly now. And then job's done. Okay, so you just, after a few minutes of a dance with the l and m and that brilliant prompt, now we have the cost of delay calculator and we can show this to teams to get some sense that this would actually be of any use in the business whatsoever.

So [00:05:00] we'll go in here, I'll show them how to use it. You'll go in, you'll put in a whole bunch of metrics in the company that we're working with, and then that will then calculate. To what extent Pendo will have an impact on their business when it comes to revenue, cost, and risk based on us changing metric fundamentals for them through their use of Pendo, and that will then articulate the value of Pendo.

Bear in mind, this is the V one or the prototype, just to get some early feedback, but you get a sense of where you can go with this magical. So the big strategic shift here is moving from that context, pushing mindset to context pulling, letting the AI take the responsibility for asking you the right questions for the context it things it needs from you based on the goal you've given it.

And this solves for those PM pain points, right, that we started with. You can now create compelling stories, working demos. You can validate those ideas with customer-centric evidence. Before committing those precious resources, you can therefore make smarter roadmap decisions without all of that waste.

And so here's my challenge to you. Try this approach [00:06:00] with one of your vague ideas. Embrace the iterative dialogue. Use voice on mobile to answer those questions quickly, and use the prototype to show, not just tell, and to make it easy for you to get started, I've put this exact prompt that we've shown here today.

Into the YouTube description for this episode so that you can copy it and try it for yourself. So that's it for me. Thank you for joining another episode of the Vibe pm Please spread the vibes and I'll see you next time. Thank you.