Why your ChatGPT prospect list Is still sitting in a Google Doc

I asked ChatGPT to build me a prospect list in January. Fifty SaaS founders in the B2B space, based in North America, between ten and fifty employees, who'd recently posted about struggling with customer acquisition.

It gave me a beautiful list. Names, company names, LinkedIn URLs, a one-sentence summary of why each person was a fit. Formatted perfectly. Ready to go.

That list is still in a Google Doc. I opened it once, three days after I created it, got through two of the profiles, wrote half a message to the first one, and then my kid needed something and I closed the laptop.

That was eleven weeks ago.

I'm not proud of this. I'm also not unusual. Every founder I've mentioned this to has laughed and said some version of "yeah, I have one of those too." The Google Doc full of prospects that were going to be the start of a real outreach campaign. The ChatGPT-generated list that felt like progress when it was created and has been slowly rotting ever since.

The gap between a prospect list and actually doing something with it

ChatGPT is remarkably good at the first 10% of outreach. It can research industries, identify personas, suggest targeting criteria, draft templates, and produce a list that looks like it came from a junior analyst who spent a week on it.

The problem is that the first 10% is the part that wasn't actually hard.

The hard part was always sitting down, opening LinkedIn, finding the specific post that person wrote last week, crafting a reply that references what they actually said, and clicking send. The hard part is doing that five times. Tomorrow. And the day after. For a month.

ChatGPT solved the research problem. Nobody was stuck at research. They were stuck at execution.

I know this because I spent six years building workflows and automation systems before starting MyPip. I've seen hundreds of prospect lists. CRM exports. Enrichment spreadsheets. Beautifully structured databases of potential customers with phone numbers, email addresses, company revenue, tech stack, intent signals, and employee count.

Almost none of them were ever fully actioned. Not because the data was bad. Because the person who made the list got busy, and the list got old, and by the time they came back to it the context had changed and starting felt harder than it did the first time.

Why AI-generated prospect lists don't convert into customers

There's a specific failure mode with using ChatGPT for prospecting that I want to name because I think it's making the execution problem worse, not better.

When you ask an AI to generate a prospect list, you get something that feels complete. It has structure. It has detail. It looks like a finished product. You save it, feel productive, and move on to the next thing on your to-do list.

But a prospect list isn't a finished product. It's a starting line. The value is zero until someone sends the first message.

The completeness of the AI output creates an illusion that the work is done. I've watched this happen with my own behaviour. The afternoon I generated that fifty-person list, I felt like I'd accomplished something meaningful. I hadn't. I'd created a document. The same way reorganizing your desk feels productive but doesn't actually move anything forward.

The second problem is staleness. ChatGPT doesn't have access to what your prospects posted yesterday. It doesn't know that the founder on row twelve just tweeted about a failed product launch and is probably more receptive to a conversation about growth right now than they were last month. It doesn't know that the company on row thirty-seven just laid off their marketing team and is suddenly looking for ways to do outreach without dedicated staff.

A prospect list generated on Tuesday is already less relevant by Friday. By the following Tuesday it's significantly less relevant. By the time you actually get around to acting on it, three weeks later, most of the context that would have made your outreach feel personal and timely has evaporated.

You end up sending generic messages to a stale list. Which is exactly the outcome you were trying to avoid by using AI in the first place.

The execution system nobody builds

The outreach workflow has three parts. Finding the right people. Creating a personalized approach for each one. And then actually doing the outreach, daily, in a way that builds relationships over time.

ChatGPT handles the first part reasonably well and the second part okay. Nobody handles the third part. Not ChatGPT. Not the outreach tools that cost $200 a month. Not the courses and templates and playbooks that fill your bookmarks.

The third part is an execution system. Not a tool. Not information. A system that puts specific actions in front of you every morning and makes doing them easier than not doing them.

Think about why going to the gym with a personal trainer works better than having a workout plan on your phone. The plan is the same. The exercises are the same. The difference is that the trainer removes every decision except one: do the rep or don't.

The plan on your phone requires you to decide when to go, what to do first, how many sets, how heavy, whether to swap in a different exercise. Each decision is a small opportunity to quit. By the third decision, most people have negotiated themselves out of the workout entirely.

Outreach works the same way. A prospect list in a Google Doc requires you to decide who to contact first, what to say, which channel to use, whether to write the email now or research them more first, whether their post from two weeks ago is still relevant enough to reference.

Each of those micro-decisions is an exit ramp. The more decisions between you and the action, the less likely the action happens.

An execution system collapses all of those decisions into a single moment: here are your five actions for today. Do them or don't.

How to actually use your prospect list (instead of letting it rot)

If you have a ChatGPT prospect list sitting in a doc right now, you're not going to action the whole thing. Accept that. The list as a monolith is part of why it hasn't moved. Fifty names feels like a project. One name feels like a task.

Take the top five names. Not the top five based on any sophisticated criteria. Just the first five where you can find a recent LinkedIn post or tweet from the past week. That recency is what matters. Not the quality of the initial match. The freshness of the context.

For each of those five people, write one action you could take in under sixty seconds. Comment on their post. Reply to their tweet. Send a short email that references something specific they said recently. Not a template. Not a pitch. Something that shows you actually read what they wrote and had a thought about it.

Do those five things tomorrow morning. Before your first meeting. Before you open your inbox. Before Slack has a chance to hijack your attention.

Then do five more the next day. Different people, same approach. Fresh context, specific actions, sixty seconds each.

After a week, you've done 25 outreach actions. Some of those people will have noticed your name appearing twice. A few might reply. One might turn into a conversation.

That's more output than most founders get from an entire quarter of intending to do outreach.

Where AI actually helps with outreach (and where it doesn't)

I'm not down on AI for outreach. I'm down on using AI for the wrong part of the problem.

AI is excellent at monitoring. Scanning hundreds of LinkedIn profiles, Twitter feeds, company blogs, and job boards to identify which of your target prospects did something relevant in the last 48 hours. That's a research task that takes a human hours and a machine minutes.

AI is good at personalization context. Pulling together the three things about a person that would make your outreach feel specific. Their recent post about a hiring challenge. Their company's product launch last week. A comment they left on a competitor's thread. That context turns a cold email into a warm one.

AI is mediocre at writing the actual message. It can draft something that's grammatically correct and reasonably personalized, but it still sounds like AI wrote it. The recipient can feel it. The best outreach is still written by the founder, even if it's only two sentences, because those two sentences have a specificity and authenticity that no model replicates consistently.

AI is terrible at being the execution system. It gives you information and then walks away. It doesn't show up in your Slack at 7:30 in the morning with five specific actions. It doesn't track that you engaged with someone three times this week and they still haven't responded so maybe it's time to try a different angle. It doesn't notice that you skipped yesterday and gently remind you that consistency matters more than perfection.

The valuable AI for outreach isn't the AI that writes your emails. It's the AI that tells you who to talk to today and puts the action one tap away.

The difference between a list and a system

I have a theory about why the Google Doc prospect list persists despite being obviously ineffective. It's because making the list feels like control. You're organizing the chaos of "who should I be talking to" into something structured. That sense of control is satisfying. It scratches the itch.

But control without action is just anxiety in a spreadsheet.

A system is different from a list because a system has a clock. It runs on a schedule. It puts something in front of you at a specific time and asks for a specific response. A list waits passively. A system shows up actively.

Email has a clock. Slack has a clock. Your calendar has a clock. A Google Doc does not.

Every founder I've spoken to who actually maintains a consistent outreach practice has attached it to something with a clock. A daily reminder, a morning routine, a message that arrives in an app they already check compulsively. The practice survives because it lives inside a cadence they already follow, not in a document they have to remember to open.

If your prospect list is in a Google Doc, move it into whatever app you check first in the morning. Take five names, attach five actions, set them to appear at 7:30am. That alone, just changing the delivery mechanism, will get more of that list actioned than any amount of refining the list itself.

The list was never the bottleneck. The bottleneck was always the moment between knowing what to do and actually doing it.

I still have that original fifty-person list from January, by the way. I went through it last week. Fourteen of the companies had changed something significant since I made the list. Three had been acquired. Two had pivoted to different markets. One founder had left the company entirely.

Eleven weeks of decay. A perfectly researched list that was already wrong before I ever used it.

Next time I'll skip the list and just start the conversations.

Or more likely, I'll have something do the research for me every morning and just show up with today's five. That seems to be the only version of this that actually works.