TATT: The “AI assistant” for prospect research

Most agency owners doing prospect research are still Googling company names, scrolling LinkedIn, and skimming "About Us" pages hoping something useful jumps out. It's slow, it rarely surfaces anything genuinely interesting, and by the time you get on a discovery call you've got a vague sense of what the company does but nothing that makes them feel like you've actually done your homework.

Today I want to show you a specific way to use AI as a pre-call research assistant that takes about 25 minutes and will make you sound sharper in the first five minutes of any sales conversation than most agencies sound across the entire call.

This isn't "just paste their website into ChatGPT." That produces shallow summaries that don't help you. This is a structured research protocol, specific inputs, specific prompts, a clear output that builds a layered picture of a prospect before you ever say hello.

Why standard prospect research fails you

The goal of pre-call research isn't to know facts about a company. It's to enter the conversation with a hypothesis about their actual problem. There's a big difference. Knowing that a company was founded in 2018, has 45 employees, and sells B2B SaaS tells you almost nothing useful. Having a hypothesis that says "their paid acquisition is likely leaking at the bottom of funnel because their review presence is weak relative to their ad spend", that's something you can lead with. That hypothesis earns you the right to ask better questions, which is what moves deals forward.

The research framework I'm about to walk you through is designed to produce hypotheses, not fact sheets.

The Four-Layer Research Stack

Before you touch AI, you need to collect raw inputs across four layers. Think of each layer as a different signal type.

Layer 1 Public digital footprint. This is the stuff anyone can see: their website, landing pages, active ad creatives (pull from Meta Ad Library and Google's ad transparency tools), their organic search presence (a quick SEMrush or Ahrefs free trial snapshot), and their content output over the last 90 days. You're not analyzing yet, you're collecting. Copy URLs, paste ad copy, note what pages exist and which ones are thin.

Layer 2 Review and reputation data. Pull their Google Business profile reviews, G2 or Capterra reviews if they're a software company, Trustpilot if relevant. Look at both the star rating and the actual language customers use. This is gold. Customers tell you in plain language what's working and what isn't. Specifically look for patterns in negative reviews and in the language used in positive ones, that language tells you what they're actually selling even if it doesn't match their marketing copy.

Layer 3 Team and leadership signals. Spend five minutes on LinkedIn. Look at the founder or CMO's recent posts. Look at what they're engaging with. Look at who they've hired in the last six months, job titles are a signal of where they're investing. If they've hired two performance marketers in the last quarter, they're scaling paid. If they just brought on a Head of Content, something shifted in their strategy. You're reading intent.

Layer 4 Competitor context. Pick two or three of their direct competitors. Do a quick surface-level comparison. Are competitors running more ads? Do competitors have better review volume? Is there a gap in organic presence? You're not doing deep competitive analysis here, you're just building enough context to understand where your prospect sits in their market.

The AI Synthesis Protocol

Once you've collected your raw inputs and this should take 15 to 20 minutes max if you're focused, you're going to run a structured AI session. I use Claude for this, but ChatGPT works the same way.

Open a fresh conversation and start with a context-setting prompt. Something like this:

"I'm preparing for a discovery call with [Company Name]. They are a [brief description]. I'm going to paste several pieces of raw research and I want you to help me synthesize them into a pre-call brief. Don't summarize what I give you, I want you to identify patterns, surface likely problems, and generate hypotheses about where their marketing or growth has friction. I'll feed you the research in chunks."

Then feed it in layers. Paste their website copy and landing pages first. Ask: "Based on this, what is the core value proposition they're leading with, and does the messaging feel differentiated or generic?" Then paste ad creative copy. Ask: "What's the narrative they're pushing in paid? Does it align with or contradict the website messaging?" Then paste the review data. Ask: "What themes emerge in customer language, and what does this suggest about the gap between what they promise and what customers actually experience?"

After each layer, you're not looking for long essays. You're looking for pointed observations. Tight. Specific.

Then you run a synthesis prompt: "Based on everything I've shared, generate three hypotheses about where this company likely has friction in their marketing or growth right now. Frame each hypothesis as a statement I could test in conversation, not as a certainty."

This is where it gets useful. You'll get output like: "Their ad creative appears to be focused heavily on features rather than outcomes, while customer reviews consistently reference time savings, there may be a messaging misalignment that's affecting conversion." That's a hypothesis you can walk into a call with.

What a finished brief looks like

Your output from this entire process should be a single-page document. It doesn't need to be pretty. Mine has five sections:

1. What they're saying (their core positioning in their own words)

2. What their customers are actually saying (the real language from reviews)

3. What their paid activity suggests about current priorities or budget allocation

4. Three competitor comparison points worth noting

5. Two to three hypotheses about where they likely have friction

That's it. Print it or have it open on a second screen going into the call.

A real example of this in action

A few months ago I was prepping for a call with an e-commerce brand in the home goods space. Mid-market, decent ad spend, had been running for about four years. Standard research would have told me they sell furniture online, they're profitable, they're growing. Not helpful.

Running this protocol surfaced something specific: their Google reviews were full of customers raving about the quality of the product but mentioning slow delivery and lack of updates during shipping. Their ads were running on a "quality, delivered fast" message. Their competitors were leaning hard into "white glove service" language. There was a direct conflict between what they were promising in paid and what customers were experiencing, and their competitors had already identified that gap and were exploiting it.

I walked into the call and within the first few minutes said "I noticed your ad creative is leaning on speed as a differentiator, I'm curious whether that's been a friction point at all in customer experience." The CMO paused and said "that's actually our number one ops headache right now."

That's the conversation you want. That's what a hypothesis does that a fact sheet can't.

The whole protocol collecting inputs, running the AI synthesis, building the brief takes about 25 minutes once you've done it a few times. The first time will take longer as you build the habit, but the output quality is dramatically better than anything you'll get from passive browsing.

One more thing: save your synthesis prompts as a reusable template. Build a prompt doc you can copy and paste for every new prospect. Once you have the structure dialed in, the speed improves significantly.

I have linked a set of prompts to work through each layer and out put the brief.

When I did my 50 prospect out reach this is the framework I used to craft my emails and phone scripts sent 50, got 14 clients out of it worth roughly 13k total sales just under 10k MMR in 2 weeks.

How are you currently doing pre-call research, and where in this framework do you think you'd get the most leverage? Drop it below.

https://docs.google.com/document/d/1Z6juNTgYOFPwJJOVdmUvQGNCqvQpbF1mYDKpW-AXPUA/copy?usp=sharing

4 comments

TATT: The “AI assistant” for prospect research