Gains:
- Ability to evaluate potential customers with ideal customer profile (ICP) and qualification framework (e.g. BANT)
- Ability to produce a structured account/person briefing from publicly available information
- Ability to mark qualification output with assumptions that need to be verified
The most expensive mistake in sales is spending too much time with the wrong person. If a sales rep spends his day meeting with a lead who never needs your product, real opportunities fade away. That's why good selling starts with screening before you sell: who's really a good candidate and who's not?
This job is called qualification. Artificial intelligence (AI) plays two powerful roles here: it extracts a quick account brief from publicly available information (website, industry, news) and evaluates the lead according to a defined framework. But beware: the information AI produces is an initial assumption, not proven fact. In this unit, you will learn how to maintain accuracy and confidentiality while accelerating lead research.
Note: Account/person information generated by AI should not be processed into the CRM as real without being verified, and especially should not be used as the basis for a decision (e.g. "this company has no budget"). Collecting non-public personal data is both an ethical and legal limit.
Attribution Frameworks: ICP and BANT
Two basic concepts:
- ICP (Ideal Customer Profile): Describes the characteristics of the customer who benefits most from your product, is the easiest to close and stays the longest (sector, size, need). "Who should we sell to?" is the answer to the question.
- BANT: Four questions that measure a lead's maturity: Budget (is there a budget), Authority (is he a decision maker), Need (is there a real need), Timing (is the timing appropriate). Evaluating each lead in these four makes it clear who to prioritize.
AI works great at fitting a lead into these frameworks and extracting missing information as “questions to ask.”
Step by Step: Qualifying a Lead
- Identify your ICP. Write down 4-5 concrete characteristics of your ideal customer.
- Collect publicly available information. Website, "about" page, industry, job openings.
- Remove briefing with AI. Produce a structured account statement.
- Score ICP compliance. How well does the lead fit the ICP?
- Mark the assumptions. Keep AI-predicted (unproven) information separate.
- Prepare opening questions. Questions to ask in the first interview to fill in the BANT gaps.
Copiable Prompts
The basic prompt that extracts an account briefing from publicly available information:
Role: You are a sales research assistant. Below is PUBLIC text about a company. Produce a structured account briefing from this:- What the company does (1 sentence)- Industry and estimated size- Possible needs/pain points (related to our product)- ICP fit: [High | Medium | Low] + single sentence justification - list separately ASSUMPTIONS (inferences without evidence) that need to be verified FITTING information not in the text; If you don't know, write "unclear". Our product: {{ product_summary }}Company text: {{ publicly_public_text }}
Prompt that evaluates the lead according to BANT:
Evaluate the following lead information within the BANT framework: - Budget: we know / unknown - Authority (authority): is the person a decision maker? - Need: is there evidence? - Timing (time): is there a trigger event? Mark "known / unknown" for each dimension. Suggest a question to ask at the first meeting that will cover the unknowns. Lead information: {{ information }}
Prompt that produces a personalized first contact opening:
Based on this account briefing, write a personalized one-sentence ice-breaker for the first contact. Let it be based on a concrete observation specific to their work, not general praise. Do not exaggerate, do not use unsubstantiated claims. Briefing: {{ briefing }}
Prompt that also suggests "elimination" in qualification:
If this lead seems low compatible with our ICP, say so clearly and give 1-2 items of justification why it may not be worth pursuing. Don't force rapport; be honest to allocate our resources to the right place.ICP: {{ icp }} | Lead: {{ lead }}
Weak Prompt / Strong Prompt
poor approach
Strong approach
"Tell me about this company"
Returns public text + requests structured briefing
Produces fictitious turnover/number of employees
Says "uncertain", marks assumptions separately
Shows every lead as "qualified"
Fairly eliminates low compatibility
General opening ("great company")
Concrete opening specific to their business
The difference is that the robust approach uses AI as a “hypothesis generator” rather than a “source of fact”: each output is a blueprint to be verified.
Three Mini Cases
Case 1 — The fake turnover trap. A representative asked: "What is the turnover of this company?" he asked; AI confidently said "250 million TL annually". However, this was completely fabricated; The company never disclosed this information. When the team realized this, they updated the prompt with the rule "If you don't know, write vaguely, make up numbers". In subsequent briefings, the AI wrote "uncertain" in the turnover field, preventing an erroneous proposal based on an incorrect budget assumption.
Case 2 — Honest elimination saved time. A list of 30 leads was passed through the ICP prompt. AI flagged 9 leads as “low fit” and gave justification (e.g. “too small team, below minimum number of users of our product”). Instead of spending time on these 9, the representative focused on 21 qualified leads; In that quarter the quality of the pipeline (sales pipeline) increased significantly.
Case 3 — Concrete opening door opened. AI inferred from a company's recent job postings that "they are building a fast-growing support team" and based its opening sentence on this: "I was curious how you manage the increasing volume of requests as you grow your support team." Compared to a general opening, the response rate was noticeably higher; because the message touched on their real situation.
Tip: Turn each “assumption” the AI infers into a question in the first meeting. Instead of saying, “I think your support volume is increasing,” ask, “How has your support volume changed lately?” ask. This way, you confirm the assumption and get the customer talking.
Privacy and Ethical Boundary
Attribution is made only with publicly available and business information. It is both unethical and unlawful to use a person's private social media posts, personal life or data obtained without permission as a source of research (KVKK, i.e. personal data protection legislation). The rule is simple: do not collect and enter into AI any personal information that is not written on the company's official website or on open commercial resources.
Caution: AI-generated "need/pain point" predictions should not be presented to the customer as if they were known for certain. Saying "You have this problem" instantly breaks trust if it's wrong. Instead, “this challenge is often seen in your industry, how are you doing?” Proceed open-ended.
Catching Triggering Events
In sales, timing is often more important than the message itself. A “trigger event” is a publicly visible development that indicates a company may need your solution right now: the hiring of a new executive, the opening of a new office, the acquisition of new investment, rapid team growth, or a new product launch. These events are a “doorway”; Because companies experiencing change are in the period when they are most open to new solutions. You give the AI an account's public news and postings and ask "are there any triggering events that can be linked to our solution?" you may ask.
With such a trigger in place, your first contact message becomes much stronger: An opening like “I saw you opened your new operations center; managing support volume at this scale is often challenging” feels both relevant and timely. Still, the same discipline applies: verify that the triggering event is real and current, and don't be laughed at by mistaking old news for "new." Timing is a good friend, but only with verified information.
Common mistakes
- Accepting the turnover/employee number made up by AI as real without verifying it.
- To qualify each lead; skipping honest screening and wasting resources.
- Conducting research with the assumption that "everyone is our customer" without defining an ICP.
- Searching for non-public personal data and entering it into AI (KVKK violation).
- Presenting AI's pain point prediction to the customer as absolute truth.
- Skipping validation by not turning assumptions into interview questions.
In summary
- Good selling starts with elimination; ICP and BANT make it clear who to prioritize.
- AI extracts rapid briefing from publicly available information; but its output is an assumption to be verified.
- Prevent making up numbers and information with the "if you don't know, write vaguely" rule.
- Eliminate low-conformance leads honestly; Allocate your resources to the right place.
- Use only publicly available, business information; Protect personal data limit (KVKK).
Application task
Take a real target company's publicly available "about" page and feed it to the briefing prompt. Check three things in the output: (1) did the AI make up any numbers/information (if so, strengthen the "uncertain" rule), (2) does the "assumptions" list actually isolate unevidenced inferences, (3) does the ICP compliance rationale make sense. Then turn each highlighted assumption into an open-ended question that you will ask in the first interview.
checklist
- [ ] I described my ICP with 4-5 concrete features.
- [ ] I have used only publicly available, business information.
- [ ] I prevented the AI from making up numbers/information with the "fuzzy" rule.
- [ ] I marked assumptions separately from proven information.
- [ ] I eliminated low-compatibility leads honestly.
- [ ] I turned each assumption into a question I would ask in the interview.