Ali Fakhar
فارسیGet in touch

How to research a prospect with AI before a sales call

  • Level: Advanced
  • A few days to set up

Ali Fakhar · Updated: Sept 2026

The short answer

Start from the enquiry and the company’s own pages, collect facts with their sources, then ask a model for a one-page brief: what the company does, what they asked, what is still unknown and which questions to ask. A person checks every claim before the brief reaches sales.

On this page

At a glance

The marketing problem
A new enquiry often reaches sales with little useful context. Manual research takes time, and a confident AI summary can hide assumptions about the company. A sourced brief gives the first call a better start without inventing what the buyer wants.
Who this is for
B2B marketing and sales teams that research incoming enquiries before a discovery call.
Tools involved
  • n8n
  • HubSpot
  • Claude API
  • Gemini API

What you need

Before you start

  • An agreed brief template: company overview, the enquiry, unknowns, questions for the call
  • Access to the CRM record and the enquiry form
  • A list of sources you trust for company facts
  • A model that can read long documents (ChatGPT, Claude or Gemini)

What goes in

  • The enquiry, in the person’s own words
  • The company website: home, about, product and news pages
  • Verified company data from your enrichment tool, if you have one
  • Longer public documents, such as annual reports or case studies

How it works

  1. 01 · TriggerCapture the enquiry
  2. 02 · ToolCollect company facts with sources
  3. 03 · AI stepDraft the one-page brief
  4. 04 · AI stepRead the long documents
  5. 05 · You decideCheck and correct
  6. 06 · OutputAttach the brief to the CRM record
  • Trigger
  • Tool
  • AI step
  • You decide
  • Output

Step by step

  1. 01 Trigger

    Capture the enquiry

    Start from the submitted form and the company domain. Keep what the person actually asked, word for word, at the top of the brief.

  2. 02 Tool

    Collect company facts with sources

    Gather facts from the company’s site and your data provider. Record the URL and date for each fact, and keep observations apart from assumptions.

  3. 03 AI step

    Draft the one-page brief

    Give the model the enquiry and the sourced facts and ask for a structured brief. Nothing without a source goes in as a fact.

    Prompt to copy

    You are preparing a salesperson for a first call with [company name]. Using only the sources below, write a one-page brief with these sections: 1. What the company does (two or three sentences, each with its source URL) 2. What they asked us, quoted from the enquiry 3. What we do not know yet (budget, timeline, decision makers, current tools) 4. Five questions to ask on the call, each linked to something in the sources Rules: if a fact is not in the sources, write “unknown”. Do not guess company size, budget or priorities. Enquiry: [paste the enquiry] Sources: [paste facts with URLs]

  4. 04 AI step

    Read the long documents

    For annual reports, case studies or long PDFs, ask a model with a large context window to pull out only what matters for the enquiry, with page references.

    Prompt to copy

    Read the attached [document type] from [company name]. List up to five points that matter for a conversation about [our service area]. For each point give the page number and a short exact quote. If nothing in the document is relevant, say so.

  5. 05 You decide

    Check and correct

    Open each source and confirm the claim. Remove anything unsupported. If the prompt caused the error, send the brief back to the drafting step with the correction.

  6. 06 Output

    Attach the brief to the CRM record

    Save the reviewed brief on the contact or deal record, where the person running the call will look first.

Where a person decides

A person checks every claim against its source and approves the brief before sales uses it.

What to watch for

  • Public information can be out of date or promotional.
  • A website does not tell you budget, urgency or who decides.
  • Models fill gaps confidently, so unknowns must stay marked as unknown.

How to tell it is working

  • Every material claim links to a source you can open.
  • The salesperson says the brief changed how they prepared.
  • Preparation time is compared with doing the research by hand.
  • Corrections are logged, so the prompt improves.

An illustrative example

Input

Illustrative: a fictional analytics company asks about automating research before discovery calls. Its team size and budget are unknown.

Output

A short sourced company overview, the enquiry in its own words, “budget: unknown”, and questions about call volume and the current research process.

Interactive example · fictional dataHubSpot / n8n / AI

One lead. Six steps. The judgement between them.

01Incoming lead
Input
Aster Analytics — fictional B2B software company. Request: “We need faster preparation for discovery calls.”
Output
A research task linked to the enquiry. No buying intent or budget is inferred.
Evidence
Sample contact form, supplied for this demonstration.
My decision

Keep the original words. A request for information is not a qualified opportunity.

02Company research
Input
Sample company page: reporting software for operations teams. Team page lists 18 people. No turnover or budget information.
Output
Known: reporting software; operations audience. Unknown: total headcount, turnover, budget and buying timeline.
Evidence
Fictional company and team-page excerpts. These are fixtures, not live web research.
My decision

Store the evidence with each fact. A team page is not a reliable company-size estimate.

03Evidence-backed brief
Input
The form and the two source excerpts.
Output
Draft: “Aster has 18 employees and is ready to invest in AI.” This contains two unsupported conclusions.
Evidence
Deliberately flawed model-style output, written for this demonstration.
My decision

Flag both claims. The model has converted partial evidence into false certainty.

04Suggested fit
Input
Known problem: slow call preparation. Relevant offer: research workflow implementation.
Output
Potential fit. Reason: a repeatable research task with a clear handoff. Confidence: limited until a conversation.
Evidence
A transparent rule applied to the sample enquiry; not a predictive score.
My decision

Do not reject or prioritise solely on an invented company size. Ask about volume and current process.

05Human review
Input
The draft brief, source excerpts and suggested fit.
Output
Corrected brief: “Aster builds reporting software for operations teams. The contact wants faster call preparation. Team size and budget are unconfirmed.”
Evidence
Every statement can be traced to the sample form or company excerpt.
My decision

I remove unsupported claims and add two questions: how many calls each week, and what research is required?

06CRM handoff
Input
Reviewed brief, evidence references and open questions.
Output
A contact note and a task for the account owner. No automated prospect message is sent.
Evidence
Illustrative CRM record. No real person or account is contacted.
My decision

The account owner decides the next action. The workflow prepares a better conversation.

A pre-authored demonstration. It does not research live companies or send messages.

Want this running in your team?

AI workflow implementation. Turn a repetitive marketing task into a workflow your team can use with confidence.

View service
Research and insight

How to do competitor analysis with AI

Compare competitors’ positioning, offers and proof side by side with one repeatable prompt, and check every claim against the source page.

All AI in marketing workflows ↗
Analytics preferences

Current preference: off