You have a discovery call in 45 minutes. You open Google, spend half an hour reading the company’s homepage, skimming their LinkedIn, and trying to find anything recent in the news — then close 12 tabs with a handful of half-useful notes. There is a better way, and it takes a tenth of the time.

Learning how to research a prospect using AI is not about cutting corners. It is about shifting from the exhausting manual task of data-gathering to the high-value cognitive work of strategic thinking. Instead of asking “what does this company do?”, you arrive at the call already asking “how does our specific offering map to the three strategic risks they disclosed in their last annual report?” That is the difference between a rep who sounds prepared and one who sounds like a peer.

The reality is that most professionals who try AI for prospect research get mediocre results — not because the AI is bad, but because they approach it like a search engine. They type a company name and expect magic. What you actually need is a structured three-step process: feed the AI real source material, extract sales-relevant signals rather than generic summaries, and then force it to map those signals directly against your specific value proposition. This article teaches you exactly that system, with five ready-to-use prompts for each professional scenario.

One thing to address immediately: AI can and does hallucinate during research tasks. We will cover exactly how to prevent this, which tools are most reliable for which tasks, and what data you should never put into a public AI chatbot. This is not a “cool trick” article — it is an operating procedure you can rely on before every client meeting.

The Problem: Why Manual Research Fails at Scale

A sales professional with five discovery calls per week spends roughly 3–4 hours just on pre-call research. That is an entire working morning, every week, spent reading websites, skimming LinkedIn profiles, and scanning Google News — before a single conversation has even happened. According to research on B2B seller time allocation from Gartner, the majority of a sales rep’s working time is spent on non-revenue-generating activities. Pre-call research is one of the largest contributors.

The problem is not effort — it is structure. Most manual research produces the wrong type of output. You end up knowing what the company does (something you could have learned in 30 seconds from their homepage), but not knowing what they are struggling with right now, which of your product’s features actually maps to their current priorities, or what specific recent event gives you a natural, timely reason to reference in your opening question. That is the gap AI closes — not by replacing your judgment, but by synthesising raw source material into sales-relevant intelligence in under two minutes per source.

Why Generic “Tell Me About This Company” Prompts Fail

A common mistake is opening ChatGPT and typing “Tell me about Acme Corporation.” This produces a Wikipedia-style paragraph that recaps founding year, headquarters location, and a generic description of their product category. It is the equivalent of reading the first paragraph of their Wikipedia page. It is not wrong — it is useless for sales.

The issue is that vague prompts produce vague outputs. The AI has no idea you are about to get on a call with their VP of Operations. It does not know which of your three product features you are leading with. It does not know you need to reference a specific pain point in the first 60 seconds to establish credibility. When you give the AI context — your role, your product, the type of insights you need, and the format you need them in — the output quality changes completely. The prompts in this article are built around that principle.

The Danger of AI Hallucinations in B2B Sales

This is the section most “AI for sales” articles skip, and it is the most important practical point in this guide. AI hallucination in a B2B sales context means the AI invents a funding round that never happened, attributes a product launch to the wrong quarter, or names a decision-maker who left the company eighteen months ago. You quote this information on a call, and the prospect knows instantly that you did not do real research — you outsourced it to a chatbot and trusted the output blindly.

The fix is called zero-shot grounding: you instruct the AI to base its output only on the source material you provide, and to explicitly flag anything it cannot find rather than filling the gap with inference. Every prompt in this article includes this guardrail. The rule is simple: never ask AI to retrieve facts from memory. Always bring the data to the AI, and tell it to work only within what you have provided.

The 10-Minute AI Pre-Call Brief (The Core Workflow)

Here is the system. It has three steps, and the order matters. Most professionals jump straight to step three (asking the AI to give them a pitch angle) and then complain that the output is generic. The reason the output is generic is that they skipped steps one and two. The AI cannot synthesise what it has not been given.

Step 1: Feed the Context (Website, PDF, or LinkedIn)

Before writing a single prompt, you need to decide what source material you are giving the AI. There are three types, each suited to different scenarios:

• Company website URL: Best for SMBs and private companies. Use an AI with live web browsing (Google Gemini Advanced or Microsoft Copilot) and paste the URL directly. Works well for understanding their core offer, target market, and messaging tone.
• 10-K or Annual Report PDF: Best for publicly traded companies. Download the PDF from the SEC EDGAR database and upload it directly to Claude or ChatGPT Plus. These tools can now process full-length financial documents and extract strategic risks, investment priorities, and financial headwinds in table format.
• LinkedIn profile text: Best for individual stakeholder research before a meeting with a specific person. LinkedIn blocks AI web scrapers, so you must manually copy the “About” and “Experience” sections and paste them into the chat window.

One rule that applies to all three: before you paste anything, add a grounding instruction. The first line of every prompt should tell the AI to work only from the material you have provided and to flag any gaps rather than filling them with inference.

Step 2: Extract the Triggers and Pain Points

This is where most professionals stop when doing research manually — they find out what the company does, and then they improvise on the call. The AI’s real value in this step is identifying what the company is actively struggling with right now, not just what they sell. There are two high-quality sources for current pain points that most salespeople overlook:

Job postings are one of the most reliable signals available. A company posting for a “Director of Data Governance” is telling you they have a data management problem they cannot solve internally. A company with 15 open engineering roles in the same team is scaling and probably breaking its internal processes. Paste a company’s open job listings into the AI and ask it to extract the implied pain points — you will learn more in two minutes than you would in an hour of website reading.

Trigger events are recent occurrences that give you a natural, timely reason to open a specific conversation: a funding round, a product launch, a leadership hire, an acquisition, or even a company-wide restructuring announcement. Any AI with live web access (Gemini Advanced, Copilot, or ChatGPT with browsing enabled) can surface the last 90 days of relevant signals in under a minute.

Step 3: Map Your Value Proposition

This is the step that separates a good pre-call brief from a great one, and it is the step that AI makes dramatically faster than manual research. Instead of writing your pitch from your own product’s perspective, you ask the AI to cross-reference the prospect’s pain points with your specific product capabilities and generate a hypothesis: “Given what I know about this company’s current situation, here is why our product is specifically relevant to them right now.”

The key instruction is to feed your product capabilities into the prompt first, before asking for the mapping. If you do not give the AI your value proposition, it will invent a generic one. The prompt templates below show exactly how to structure this cross-reference.

How to Research a Prospect Using AI: 5 Copy-Paste Prompts

Each prompt below is ready to use in ChatGPT Plus, Google Gemini Advanced, Microsoft Copilot, or Anthropic Claude. They all include zero-shot grounding instructions to prevent hallucination, and they all produce output in a scannable format you can reference during a live call. Fill in the bracketed sections with your actual information before running the prompt.

A practical note on workflow: keep two browser tabs open — one for the AI, one for the prospect’s LinkedIn or company website. Copy what you need into the AI window, run the prompt, and paste the output into a simple notes document you can reference on-screen during the call. The whole process, including prep, takes 8–12 minutes once you have done it twice.

Prompt 1: The Executive Bio and Icebreaker Generator

Use this before any meeting where you have been given a specific person to speak with — a VP, a Director, or a C-suite executive. It eliminates generic small talk by generating an icebreaker rooted in their actual career history, and it tells you how they are most likely measured in their current role (which is the pain point frame you need to open a productive conversation).

Act as a B2B sales strategist preparing me for a discovery call. Base your entire response on the text I provide below — do not invent or infer any information not present in this text.

PROSPECT INFORMATION (copied from their LinkedIn profile):
Name: [Prospect Name]
Title: [Current Job Title]
Company: [Company Name]

LinkedIn About Section:
[Paste their About section here]

Last 2–3 Job Descriptions:
[Paste their experience entries here]

MY PRODUCT / SERVICE IN ONE SENTENCE:
[Example: "I help operations leaders at mid-market manufacturers reduce manual reporting time by 40%."]

Please generate:
1. THREE metrics this person is most likely evaluated on in their current role, based only on their job description and title
2. Their likely communication style (technical / visionary / results-focused / relationship-driven) — with a one-sentence justification
3. ONE specific icebreaker question based on their career history that is not generic — it must reference something specific in their background
4. ONE tailored opening angle connecting their likely priorities to what I sell

If any of the above cannot be answered from the provided text, write "INSUFFICIENT DATA" for that item — do not guess.

Prompt 2: The SMB Website Instant Discovery Brief

Use this for smaller or private companies where a 10-K is not available. This prompt works best with Google Gemini Advanced or Microsoft Copilot, both of which can browse live URLs. It is designed to replicate what a good discovery call prep would look like — understanding their business model, their competitive angle, and two smart questions you can lead with.

I have a discovery call in 15 minutes. Please browse this URL and provide a structured pre-call brief based ONLY on what is publicly available on their website. Do not invent details not present on the site.

Company URL: [Paste URL here]

My service/product: [One sentence description]
My target customer profile: [e.g., "Operations managers at companies with 50–500 employees"]

Generate a Pre-Call Brief with these exact sections:

SECTION 1 — WHAT THEY DO (1 sentence: their core offer + target audience)
SECTION 2 — HOW THEY MAKE MONEY (their apparent business model: subscription, project-based, product sales, etc.)
SECTION 3 — COMPETITIVE ADVANTAGE (their stated or implied differentiator based on their messaging)
SECTION 4 — LIKELY PAIN POINTS (2 pain points I might be able to solve, inferred from their messaging — label as "HYPOTHESIS: not confirmed")
SECTION 5 — TWO SMART OPENING QUESTIONS (industry-specific, not generic — tied to Section 4 hypotheses)

If the website blocks access or lacks detail on any section, write "DATA NOT AVAILABLE."

Prompt 3: The 10-K / Annual Report Synthesizer

Use this for publicly traded companies before enterprise-level meetings. Download the company’s 10-K directly from the SEC EDGAR search tool, upload the PDF to Claude or ChatGPT Plus, and run this prompt. Claude handles extremely large documents particularly well — you can upload a 150-page annual report and get a clean three-point synthesis that maps directly to your product.

This prompt is the one that earns the most credibility on C-level calls. Referencing a company’s specific disclosed strategic risks — by name, from their own filing — signals a level of preparation that almost no competitor rep will match.

Act as a seasoned Enterprise Account Executive. I have uploaded the latest annual report (10-K) for [Company Name]. Base your entire response ONLY on the document I have uploaded — do not reference external information or assume facts not present in this report.

MY PRODUCT / SOLUTION:
[Describe your product in 2-3 sentences. Be specific — include the problem it solves and who it serves.]

TASK: Read the "Management's Discussion and Analysis (MD&A)" and "Risk Factors" sections of the uploaded document.

Please produce:

1. TOP 3 STRATEGIC INITIATIVES OR RISKS (from the document)
   For each, provide:
   — The specific initiative or risk in one sentence (quoted or closely paraphrased from the document)
   — A 2-sentence hypothesis on how my product specifically addresses this issue
   — Confidence level: HIGH (directly stated) / MEDIUM (implied) / LOW (inferred)

2. ONE KEY FINANCIAL SIGNAL (revenue trend, margin pressure, or cost concern mentioned in MD&A)
   — Quote the relevant passage (under 30 words)
   — State how this creates urgency for my product category

3. ONE ICEBREAKER QUESTION I can open the call with, referencing the above

Format as a clean bullet-point table I can reference during the call.
Flag any item where data was not found in the document with: [NOT FOUND IN DOCUMENT]

Prompt 4: The Recent News and Trigger Event Sweep

Use this with any AI that has live web access — Google Gemini Advanced, Microsoft Copilot, or ChatGPT with browsing enabled. A trigger event is a recent occurrence that gives you a natural, timely reason to reach out or reference something specific in your opening. The most valuable triggers are leadership changes, funding rounds, acquisitions, product launches, and major contract wins or losses. This prompt isolates those signals from the noise of general PR content.

Search the live web for recent news, press releases, and announcements about [Company Name] from the last 90 days. Focus specifically on sales-relevant trigger events. Ignore generic marketing blog posts, product update newsletters, and sponsored content.

TRIGGER EVENTS I am looking for:
- Leadership changes (new C-suite, VP-level hires or departures)
- Funding rounds or M&A activity
- Major new product launches or discontinuations
- Geographic expansion or market entry
- Major customer wins, contracts, or partnerships announced
- Regulatory changes or compliance issues that affect their industry

MY PRODUCT / SERVICE: [One sentence on what you sell and who you serve]

Please provide:
1. UP TO 3 TRIGGER EVENTS found in the last 90 days
   For each: Source name + date + one-sentence summary + why this is sales-relevant for me

2. ONE CONVERSATION OPENER that references the most significant trigger naturally — not salesy, just contextual

3. KNOWLEDGE GAP FLAG: If no trigger events are found in the past 90 days, say so clearly. Do not invent or approximate recent news.

Cite the source URL for each event found.

Prompt 5: The On-the-Fly Competitor Battlecard

Use this when a prospect mentions mid-call that they are also evaluating a competitor. You do not have 20 minutes to pull a battlecard — you have 30 seconds while they are talking. This prompt is designed to run in under 60 seconds and give you the two or three specific points you need to confidently continue the conversation without trash-talking the competition.

I am in a sales conversation and the prospect has just mentioned they are also evaluating [Competitor Name]. I need a quick battlecard based on publicly known market positioning — not invented weaknesses.

MY PRODUCT: [Company name + one-sentence description]
COMPETITOR: [Competitor name]
PROSPECT'S ROLE: [e.g., VP of Finance at a 200-person manufacturing company]

Generate a 60-second battlecard with:

1. CORE POSITIONING DIFFERENCE (1 sentence: what fundamental problem do we each solve, and for whom?)
2. TWO AREAS WHERE [Competitor] IS KNOWN TO BE WEAKER (based on public G2/Capterra reviews or known market positioning — label as "market perception, not confirmed")
3. ONE CONSULTATIVE QUESTION I can ask the prospect that naturally highlights our strength without directly criticising [Competitor] — phrase it as genuine curiosity, not a trap
4. ONE TRAP TO AVOID (something I should not say in this comparison)

If you do not have reliable public information on [Competitor], say so — do not fabricate weaknesses.

See the Difference: Vague Prompt vs. Structured Research Prompt

ChatGPT vs. Copilot vs. Claude vs. Gemini: Which AI is Best for Sales Research?

The answer depends entirely on what type of research task you are running. Each tool has a genuine structural advantage for a specific scenario, and using the wrong tool for the task is one of the most common reasons professionals get mediocre research output. Here is a practical breakdown for 2026.

The short version: use Claude when you have a large document to analyse (the only tool that handles 100+ page PDFs reliably). Use Google Gemini Advanced when you need live web research on recent company news. Use Microsoft Copilot when you are handling sensitive account data that must stay inside your organisation’s M365 environment. Use ChatGPT Plus when you need icebreaker generation, stakeholder profiling, and tone-matching for specific individuals.

One important caveat: LinkedIn blocks AI web scrapers across all platforms. No AI tool can read a live LinkedIn profile URL, regardless of what the tool claims. You must manually copy the profile text and paste it into the AI chat window. For more on each tool’s capability differences, our comparison of Microsoft Copilot vs ChatGPT for professional work covers the full picture across more use cases.

Data Privacy: What NOT to Tell the AI

This is not a legal disclaimer — it is a practical workflow issue. Most professionals either overshare (pasting internal deal notes into public ChatGPT) or undershare (refusing to use AI at all because they heard it was risky). The right answer is in the middle, and it depends on which tool you are using and what type of data you are handling.

What many people overlook is that the risk is not the AI reading the data — it is the data being used to train future model versions. On free or unmanaged AI tiers, your inputs can contribute to training data. On enterprise-managed platforms like Microsoft Copilot in an M365 environment, your data stays inside your organisation’s security boundary and is not used for any external model training. Understanding this distinction determines which tool you reach for depending on the sensitivity of what you are researching.

The Traffic Light Framework: What to Share and What to Keep Private

For anything in the amber or red category that you genuinely need AI to help analyse, the correct tool is Microsoft Copilot integrated into your organisation’s Microsoft 365 environment. Saving AI research briefs in Google Docs works well for green-category material and keeps your prep organised across multiple accounts.

The One Rule for Preventing Hallucinations

Every prompt in this article includes this instruction, and it is worth stating separately: always tell the AI exactly what to do when it cannot find the answer in your source material. The default AI behaviour is to fill information gaps with plausible-sounding inference. In a general knowledge context, that is acceptable. In a B2B sales context, it can make you look uninformed on a client call.

The instruction is: “If the answer is not present in the material I have provided, write [DATA NOT AVAILABLE] — do not guess, infer, or approximate.” This single line changes the behaviour of every major AI tool and turns hallucination from an unpredictable risk into a clearly flagged gap you can verify before the call.

Frequently Asked Questions

Your Next Steps

• 1

  Pick your next scheduled call and run Prompt 2 or 3 right now

  Do not wait until the morning of the call. Find the company’s website or download their 10-K today, and run the relevant prompt. Compare the output to what you would have assembled manually — the difference in depth and format will be immediately clear.

• 2

  Save the prompts as reusable templates in your AI tool

  In ChatGPT, use the Custom Instructions feature to store your product description so you never have to retype it. In Copilot, save the prompt as a Copilot Lab template. This removes the last remaining manual step and means you only need to paste the source URL or profile text each time you run a research session.

• 3

  Set up your document storage for briefs

  Create a folder in Google Docs or OneNote labelled “Pre-Call Briefs.” After each AI session, paste the brief output there with the prospect’s name and call date. This builds a searchable reference library and means you can review past research before follow-up calls without starting from scratch. Saving and organising AI briefs directly in Google Docs works particularly well if you use Gemini for live web research.

• 4

  Extend the system to your full sales workflow

  The same structured prompting approach works across the entire sales cycle. The research workflow in this article feeds naturally into writing personalised LinkedIn outreach before the call, and into building follow-up email sequences and handling objections after it. Each step uses the same principle: give the AI specific context, constrain the output format, and verify before using.