Built to interrogate the evidence, not automate executive judgment.
The rule
Use AI to compress evidence, not manufacture certainty. Every conclusion should trace back to customer language, observed behavior, or an explicitly labeled inference.
Do not paste confidential customer information into tools that have not been approved by your company. Remove personal information, commercial terms, and sensitive account details before conducting an analysis.
Prompt finder
Four questions. Then the right prompt.
Nothing you enter leaves this page. The finder narrows the library, it does not generate prompts.
Question 01 of 04
What are you trying to pressure-test?
Answer the first question and recommendations appear here.
Research guardrails
AI should compress evidence, not manufacture certainty.
01
Evidence before inference
Separate what customers actually said or did from the interpretation. Every conclusion should trace back to language, behavior, or an explicitly labeled inference.
02
Unique customers, not mentions
One vocal customer should not become “the market.” Ask for the denominator: how many unique accounts, out of how many reviewed, over what date range.
03
Contradictions matter
Show the evidence that challenges the strongest conclusion. A synthesis that resolves every tension has usually removed the most useful part of the data.
04
Frequency is not strategic importance
The most common theme is not necessarily the most consequential one. Prevalence, momentum and consequence are three separate measurements.
A simple confidence rubric
Require a rating on every claim. An unrated finding is an opinion in a research costume.
High
Multiple independent accounts; consistent across relevant segments or periods; clear direct evidence; limited contradiction.
Medium
A recurring pattern, but concentrated in one segment, stage, channel, or small sample; some contradiction or missing context.
Low
Isolated examples, seller-led language, unclear denominators, strong selection bias, or inference without direct support.
PrivacyDo not paste confidential customer information into tools that have not been approved by your company. Remove personal information, commercial terms, and sensitive account details before conducting an analysis.
Browse all
Twelve prompts. One research standard.
Search or filter the full library. Open any prompt to customize the variables and copy it with the research setup block attached.
12 prompts
01Positioning
Use whenYou need positioning grounded in how customers describe the problem
Answers the question underneath every positioning debate: which words are actually the customer's, and which ones did we put in their mouth?
A language map with excerpts, unique-customer counts, segment differences, and five recommended message inputs.
Relevant motions: B2B sales-led, B2C or ecommerce, Product-led, Self-serve, Partner-led, Community-led, Marketplace, Enterprise, Hybrid, Not sure
01Positioning
Use whenYou have a positioning statement, homepage, pitch, or campaign claim
Tests relevance, distinctiveness, credibility and commercial consequence against what buyers actually said, then rewrites only what the evidence can carry.
A go/no-go recommendation, credibility risks, and a proof plan.
Relevant motions: B2B sales-led, B2C or ecommerce, Product-led, Self-serve, Partner-led, Community-led, Marketplace, Enterprise, Hybrid, Not sure
04GTM strategy
Use whenLeadership is deciding where to focus, invest, or change course
Breaks a bet into its assumptions, labels each supports / contradicts / mixed / absent, and sizes the smallest test that could falsify the riskiest three.
A segment–motion matrix with fit, risks, confidence, and missing evidence.
Relevant motions: B2B sales-led, B2C or ecommerce, Product-led, Self-serve, Partner-led, Community-led, Marketplace, Enterprise, Hybrid, Not sure
04GTM strategy
Use whenYou need more than a list of reasons from CRM or analytics fields
Compares outcome groups, won/lost, retained/churned, repeat/one-time, to find the patterns that appear before the outcome becomes obvious.
Evidence
Sales conversations · Customer success conversations · Surveys or interviews · Support tickets · …
Expected output
Leading indicators, strategic implications, and three changes worth testing first.
Relevant motions: B2B sales-led, B2C or ecommerce, Product-led, Self-serve, Partner-led, Community-led, Marketplace, Enterprise, Hybrid, Not sure
The method
A strong prompt is a research protocol.
Paste the setup block before any prompt in this library. It forces the model to separate evidence from interpretation and prevents one loud call from becoming “the market.”
Reusable setup block
Use before every analysis
You are acting as a senior GTM researcher. Analyze only the first-party customer evidence I provide, which may include sales or success conversations, emails, surveys, interviews, support tickets, product feedback, reviews, community discussions, or behavioral notes. Do not use general market knowledge unless I explicitly ask for it. Separate direct evidence from inference. Preserve the customer's exact language when it is distinctive, but redact personal or confidential information. For every conclusion, show: (1) supporting excerpts or observations, (2) number of unique customers, accounts, or users, (3) relevant segment and journey stage, (4) whether the evidence is recurring, emerging, isolated, or contradictory, and (5) confidence: high, medium, or low. Do not equate frequency with strategic importance. Flag selection bias, company-led language, missing context, and insufficient sample size. If the evidence cannot support an answer, say so.
Ask for this output: A concise synthesis, an evidence table, contradictions, confidence, and the next research question.
This setup block tells AI how to handle evidence, uncertainty, contradictions, sample bias, and company-led language before it begins the analysis.
Output standard
Ask AI to show its work.
End any prompt with this format. It makes the analysis easier to audit, compare, and bring into a leadership discussion.
Return the answer in six sections: 1) Executive takeaway: three sentences maximum. 2) Evidence table: finding, unique-account count, denominator, segment/stage, representative excerpts, counterevidence, confidence. 3) What changed or differs: only where a comparison is valid. 4) Implications: positioning, campaigns, sales, product, and GTM strategy; include only functions actually affected. 5) Unknowns and bias: what the dataset cannot tell us. 6) Next move: the smallest decision or test supported by the evidence. Use plain language. Do not hide uncertainty behind polished prose.
Ask for this output: A traceable readout that a leadership team can challenge and act on.
Final check
Can every important claim be traced to evidence? Did the analysis expose contradictions? Is uncertainty visible? Does the finding change a decision? If not, keep researching.
Follow-up prompts
Interrogate the first answer.
The first synthesis is usually too smooth. Paste any of these after an analysis to force sharper thinking.
01Challenge the pattern+
Paste after any analysis
What evidence contradicts your strongest conclusion? Show the accounts and explain whether they are exceptions, a distinct segment, or a sign the conclusion is too broad.
02Audit the sample+
Paste after any analysis
How might the way these calls and emails were selected distort the result? Which buyer types, stages, outcomes, or channels are underrepresented?
03Remove company contamination+
Paste after any analysis
Re-run the analysis using only language or behavior initiated by the customer, not terminology introduced by sales, marketing, research, support, or success teams. Which themes weaken, disappear, or become stronger?
04Test strategic consequence+
Paste after any analysis
If this pattern is true, what decision should change? If no decision changes, explain why the finding is interesting but not strategically useful.
05Quantify carefully+
Paste after any analysis
Count unique accounts, not mentions. Show the denominator, segment mix, date range, and whether a few highly vocal accounts are driving the pattern.
06Find the earliest signal+
Paste after any analysis
At what point in the buyer journey does this pattern first appear? What observable language or behavior could teams use as a leading indicator?
07Design the falsification test+
Paste after any analysis
What new evidence would prove this conclusion wrong? Propose the fastest credible way to collect it.
08Translate the finding into action+
Paste after any analysis
Turn the finding into one positioning change, one campaign test, one sales-enablement change, and one strategic question. Rank by likely value and reversibility.
About
Created by Leah Russo
VP Marketing and GTM leader
Leah Russo works at the intersection of positioning, customer intelligence, campaigns, and revenue strategy. She created this library to help GTM leaders use AI as a research accelerator without mistaking polished synthesis for customer truth.
Companion resource for the talk “How to Pressure-Test GTM Strategy Using AI.”
Your customers have already given you the research.
The evidence is sitting inside conversations, emails, feedback, objections, behavior, and decisions. AI can help you surface it. Your job is to interrogate what it means.
