1.Data Prep Checklist
Get your data ready in just a few minutes. Four inputs, three required, one optional.
Pick the Right AI Model
Use the advanced thinking/reasoning model — the kind that pauses to think before answering. Our testing showed that GPT 5.5 Thinking had the best results.
- •Microsoft Copilot: Change the model selector from "Auto" to GPT 5.5 → "Think Deeper"
- •ChatGPT: Select "Thinking," not the fast option
- •Google Gemini Advanced: Select the Pro model
- •Claude Pro: Select Claude Opus 4.6 or 4.7
Warning: Do not use a fast model. Fast models fabricate evidence — they'll invent quotes and conversations that never happened. Your whole analysis depends on real evidence, so this step matters.
Your Four Inputs
CRM Export (Required)
You'll want to limit this to 25 accounts at a time, so think about filtering by Owner, Region, Vertical, or other ways you may segment accounts. Export your accounts as a CSV file (not a regular Excel .xlsx file — AI tools work better with CSV). Include: Account name, revenue/contract value, services they buy, locations you service, total locations they have, last activity date, primary contact name. Keep it to 25 accounts or fewer. The volume of data — especially with transcripts — can overwhelm the AI. Start with your most important accounts. You can always run a second batch. One row per account. Plain English column headers. Remove any accounts you don't want scored.
CRM Notes (Required)
Export or copy-paste rep notes, emails, meeting summaries, and service visit notes. Format: CSV with columns for Account Name, Date, Rep Name, and Note Content. Include dates on every note — notes without dates are much less useful. Go back 12 months minimum. Even a few notes per account helps. The analysis flags lower confidence for thin records.
External Research (Required)
Open a separate AI chat, paste in your account names, and use the External Research Prompt in the next section below. Copy the output. That's it.
Call Transcripts (Optional)
If you use Gong, Chorus, Fireflies, or similar: export transcripts from the past 6–12 months. Focus on QBRs, renewals, and discovery calls — skip routine scheduling calls. Upload limit: Most AI tools cap file uploads at 10 at a time. Upload in batches — the prompt will ask you to confirm when you're done before starting. The analysis works fine without transcripts. You'll get Medium confidence instead of High, and the output will tell you where transcripts would have helped.
| Input | Required? | Format |
|---|---|---|
| CRM export | Yes | CSV (not .xlsx) |
| CRM notes | Yes | CSV |
| External research | Yes | Text (copy-paste) |
| Call transcripts | No | Text files, 10 per upload |
Max 25 accounts per run. CSV format only. Use a thinking model.
2.External Research Prompt
Open a separate AI chat (GPT is best for this one). Copy/paste this prompt below, then paste your list of account names (or drag the CSV file into the chat). Remember, maximum 25. If it doesn't process all of them, just follow up with any that it missed. You can leave the results there for now.
You are a corporate research analyst. I am giving you a list of companies. Your job is to search the web and find recent, verified information for each one.
Rules
1. Run a separate web search for each company. Do not rely on your training data. If you have a deep research or multi-step search mode, use it.
2. Every factual claim must include the URL where you found it. Place the URL in parentheses immediately after the claim. If you cannot provide a real URL for a claim, do not include the claim.
3. Thin results are expected and correct. Some of these are regional or mid-market companies with limited news coverage. When you cannot find recent, sourced information for a section, write exactly: "No verified recent information found." I expect to see this for at least a few companies. That is the right answer.
4. Do not fall back to your training data. If your web search returns nothing, leave the section empty. Do not fill it with what you "know." Your training data may be outdated or wrong.
5. Do not invent names, dollar figures, dates, job titles, or events. If you are uncertain whether a fact came from a search result or from your training, leave it out.
What to Research
For each company, cover these two areas. Skip any section where you cannot find sourced information.
Company Strategy & News (past 12 months)
Press releases, leadership changes, expansions, mergers, acquisitions, or strategic initiatives.
Industry Trends & Risk Landscape
Threats, regulations, competitive dynamics, and market conditions relevant to this company's industry and geography.
Output Format
Use this exact format for every company:
## [Company Name]
**Company Strategy & News**
- [Specific claim] (URL)
- [Specific claim] (URL)
- *No verified recent information found.*
**Industry Trends & Risk Landscape**
- [Specific claim] (URL)
- [Specific claim] (URL)
- *No verified recent information found.*
Important
I am feeding this research into another AI for a business analysis. Fabricated information will corrupt the entire downstream analysis. Accurate but thin results are far more valuable to me than detailed but unreliable results. If you only find one real fact for a company, give me that one fact. Do not pad the output.
Companies to Research
Companies to research are attached and/or below.
Copy the output from this step — you'll paste it into the analysis chat in the next step.
3.Master Analysis Prompt
Open a new AI chat. Upload your CRM export (CSV), CRM notes, and the external research output from the previous step. Then paste this prompt.
You are helping a sales leader at a B2B services company identify which existing accounts have the strongest expansion opportunities — meaning opportunities to sell new services, cover new locations, or capture new projects. This is not about retention or defending existing revenue.
Your Data
I will provide you with the following data. Read all of it carefully before scoring anything.
Required:
1. CRM export — a spreadsheet or CSV with structured account data (revenue, services, locations, contracts, activity dates, etc.)
2. CRM notes / activity log — rep notes, emails, meeting summaries, and service visit records
3. External research ��� company news, industry trends, leadership changes, and other publicly available information about the accounts
Optional (if provided):
4. Call transcripts — recordings or transcripts of conversations between reps and customers
If I am uploading transcripts in multiple batches, wait until I tell you I have finished uploading before starting the analysis.
Critical Rules
Before you begin, internalize these rules. They override any default behavior.
Evidence integrity:
- Every piece of evidence you cite must come directly from the data I provided. Do not infer, fabricate, or embellish.
- When you quote internal evidence (notes, transcripts), use the exact words from the source. Do not paraphrase into something that sounds better.
- If a quote appears in the data, cite it with the date it was recorded. If you cannot find a direct quote, describe the evidence factually and cite the source and date.
- Do not invent services, conversations, people, or events that do not appear in the provided data. If the data says the company buys "fire alarm and sprinkler" services, do not describe them as buying "HVAC" or "janitorial" or anything else.
Scoring integrity:
- Use the full 0–100 range. If all accounts cluster between 50–70, re-examine your scoring — you are not differentiating enough.
- Most accounts should score 2��3 on Evidence Convergence (Dimension 1). A score of 5 requires three genuinely independent signals from different data sources pointing to the same expansion opportunity.
- This analysis is about expansion only. A competitor threatening existing business is a retention issue. Score accounts on their expansion evidence, not their churn risk.
- Score every account in the CRM export. Do not skip any.
Output integrity:
- Every "Why It's Interesting" must cite specific evidence with a date — not generic statements like "good relationship" or "growth potential."
- Every "Suggested Next Action" must name a specific person to contact, a specific topic to raise, a specific reason grounded in the evidence, and a timeframe.
- If you cannot estimate revenue based on actual data patterns in the CRM, say "Insufficient data to estimate" — do not guess.
- When transcripts are not available, rely only on CRM notes and CRM export data for internal evidence. Do not create, reconstruct, or paraphrase conversations that do not appear in the provided notes. If no direct quote exists for an account, describe the factual evidence from the notes (e.g., "Rep noted a service gap on 03/15/2026") rather than fabricating a quote.
- Do not end your response with offers to drill down, follow-up questions, or conversational filler. End with the executive summary.
Scoring Rubric
Score each account on five dimensions, each rated 1–5.
Dimension 1: Evidence Convergence & Signal Strength (Weight: 30%)
How many independent data sources point to a real expansion opportunity, and do they reinforce each other?
| Score | Criteria |
| 5 | Three or more independent sources (CRM data, notes, transcripts, external research) all point to the same expansion opportunity. Signals are specific and mutually reinforcing. |
| 4 | Two independent sources show clear, aligned signals. A third source is absent or neutral. |
| 3 | One strong signal supported by circumstantial evidence from a second source. Or two moderate signals that are loosely related. |
| 2 | One clear signal from a single source with no corroboration. |
| 1 | No clear expansion signals in any available data. Or signals that contradict each other. |
What counts: A customer asking about additional services. A rep noting a competitor failure that opens a door. A CRM record showing a gap between services owned and services available. An external news item about expansion that would directly increase demand for the type of services this company provides. A transcript where the customer raises a new need unprompted.
What does NOT count: Generic positive sentiment. Stale activity with no follow-up. A large account with no engagement evidence. Routine service events (inspections, equipment fixes). A customer saying "maybe later" or "not right now." External news that is only loosely connected to demand for these services.
Be strict. A score of 4 requires two clearly distinct signals from different data sources. A score of 5 requires three. When in doubt, score lower.
Dimension 2: Coverage Gap & Whitespace (Weight: 20%)
The gap between what the account currently buys and what they could buy — services, locations, or both.
| Score | Criteria |
| 5 | Large gap in both services and locations. Account uses a small fraction of what is available. |
| 4 | Significant gap in either services or locations. |
| 3 | Moderate gap. Several services but untapped locations, or most locations but missing key services. |
| 2 | Small gap. Fairly well-penetrated with only minor expansion possible. |
| 1 | Fully penetrated. All services across all locations. |
Assess this by comparing services owned vs. services your company offers, and locations serviced vs. total locations. A large gap with no engagement signals is just arithmetic — it does not make an opportunity real.
Dimension 3: Relationship Momentum & Engagement (Weight: 20%)
Recency and depth of engagement. An opportunity in an active relationship is more capturable than one in a cold account.
| Score | Criteria |
| 5 | Substantive engagement within 30 days. Multiple touchpoints. Customer is initiating conversations, requesting proposals, or discussing future needs. |
| 4 | Contact within 60 days with meaningful dialogue. Customer is responsive but not driving next steps. |
| 3 | Contact within 90 days. Interactions are routine (service visits, check-ins) — stable but not advancing. |
| 2 | Stale contact (90–180 days). Interactions were transactional. |
| 1 | No meaningful contact in 180+ days. Relationship would need rebuilding. |
Look at: who initiated the last contact, whether it was operational or strategic, whether the customer raised new needs, and the trend over 6–12 months.
Dimension 4: Timing & Urgency (Weight: 15%)
Whether there is a time-sensitive reason to act on the expansion opportunity.
| Score | Criteria |
| 5 | Immediate trigger: customer has requested a proposal for new services, contract renewal within 90 days where expansion can be bundled, competitor actively bidding, or customer-stated deadline. |
| 4 | Near-term trigger: budget cycle opening next quarter, planned project with a rough timeline, or vendor evaluation underway. |
| 3 | Moderate: renewal within 6–12 months, customer mentioned future plans without a firm date. |
| 2 | Weak: general awareness of future needs but no timeline or trigger. |
| 1 | None: no contract events, no stated plans, no external pressure. |
Equipment failures and service issues are operational, not expansion timing triggers. A vague mention of future needs with no timeline is a 2 at most.
Dimension 5: External Signals & Market Context (Weight: 15%)
Whether publicly available information amplifies or dampens the expansion opportunity.
| Score | Criteria |
| 5 | Strong signal with a direct, concrete link to demand for the type of services this company provides: new facilities being built, physical expansion, or regulatory changes that specifically require these services. |
| 4 | Positive context with a plausible link: company is growing in ways that could reasonably increase service needs. |
| 3 | Neutral: no external signals, or signals only loosely related to these services. |
| 2 | Mildly negative: cost-cutting, industry contraction, or tightening spend. |
| 1 | Negative: financial distress, leadership instability, or industry downturn making expansion unlikely. |
Relevance test: The signal must relate to demand for the type of services this company actually provides. If the connection requires more than one logical leap, score 3.
Composite Score
Calculate the opportunity score for each account:
Weighted Raw Score = (D1 × 0.30) + (D2 × 0.20) + (D3 × 0.20) + (D4 × 0.15) + (D5 × 0.15)
Opportunity Score = ((Weighted Raw Score - 1) / 4) × 100
Round to the nearest whole number.
Priority Tiers
| Tier | Score Range | Meaning |
| Act Now | 75–100 | Strong convergent evidence, active engagement, and clear timing. Act within 1–2 weeks. |
| Develop | 50–74 | Solid signals but missing a key element. Invest effort over the next 30 days. |
| Monitor | 25–49 | Weak or scattered signals. Check in quarterly. |
| Maintain | 0–24 | Little to no expansion evidence. Focus on retention and service quality. |
Confidence Level
| Level | Criteria |
| High | Three or more data sources contributed meaningful, specific, recent signals. |
| Medium | Two data sources contributed meaningful signals, or data is available but thin. |
| Low | Only one or two sources available, or data is stale or lacks specificity. |
Revenue Estimation
Estimate the dollar value of the expansion opportunity using these methods:
1. Service gap: For each missing service, estimate annual value based on what similar accounts in the CRM spend on that service.
2. Location gap: Current spend / locations serviced × unserviced locations.
3. Stated projects: If notes or transcripts reference a specific project, estimate based on scope described.
Present as a range. Cite which method you used. If the data does not support an estimate, say "Insufficient data to estimate."
Missing Transcripts
Call transcripts are the only optional input. If none were provided:
- Maximum Evidence Convergence score is 4, since one data source is absent. An account can still score 4 if CRM data, CRM notes, and external research all align strongly.
- Set confidence to Medium at best. Note that transcripts would improve the assessment.
- Missing transcripts means less evidence — it does not mean negative evidence. Do not penalize accounts for data you do not have.
- The analysis is still useful. CRM data, rep notes, and external research can reveal coverage gaps, timing triggers, relationship patterns, and market context.
Output Format
Produce a table with these columns, sorted by Opportunity Score descending:
| Column | What to include |
| Rank | Position in priority order |
| Account Name | From the CRM data |
| Opportunity Score | 0–100 composite score |
| Priority Tier | Act Now / Develop / Monitor / Maintain |
| Dimension Scores | Show as D1/D2/D3/D4/D5 |
| Estimated Opportunity Value | Dollar range or "Insufficient data to estimate" |
| Confidence Level | High / Medium / Low |
| Why It's Interesting | 1–2 sentences citing specific evidence with dates from the provided data |
| Key Internal Evidence | A direct quote or specific data point from the CRM notes or transcripts, with the date. Must appear in the data you were given. |
| External Signal | From external research if provided. Otherwise "No external research provided." |
| Suggested Next Action | Name the person, the topic, the reason (citing evidence), and a timeframe. |
After the table, include:
Confidence explanation (2–3 sentences, plain English): Explain what the confidence levels in the table mean and why they are set the way they are. Be specific about which data sources were available and which were not. For example: "Confidence is rated Medium across all accounts because this analysis is based on three of four possible data sources — CRM data, rep notes, and external research. Call transcripts were not provided. Adding transcripts would strengthen the evidence base and could raise confidence to High for accounts where conversations confirm the signals seen in the notes."
Executive summary (3–5 sentences): How many accounts are in each tier, the key themes you see across the portfolio, and total estimated pipeline value if calculable.
Output as Spreadsheet
Format the results as a table I can copy and paste into Excel or Google Sheets. Use a clean table format with one row per account and the column headers listed above.
Important Limits
- Maximum 25 accounts per run. More than 25 accounts — especially with transcripts — produces too much data for a single analysis. If you have more, run your top 25 first, then do a second batch.
- CRM data must be a CSV file, not a regular Excel (.xlsx) file. CSV is the most reliable format for AI tools to read accurately.
Getting Started
Start by confirming you understand your role and asking me for the first piece of data: The CRM Export. Remind me that the CRM export must be a CSV file (not .xlsx) and that I should include no more than 25 accounts. After I provide it to you, ask me for CRM notes. After that ask me for external research and remind me I can get the prompt for doing external research here: https://www.bluecollarailabs.com/thewalkthrough/opportunity-prioritization/starter-kit. After that, ask me for transcripts, but be sure to tell me that if I don't have transcripts it's ok we can proceed without them. Make sure to explain I can only upload 10 files at a time so transcripts may have to be in batches, in which case you'll confirm I'm done before starting the analysis.
The prompt will walk you through uploading your data step by step — just paste it and follow along.
4.Go Deeper
The analysis you just ran is a starting point — not the finish line. You have a ranked list, scores, and suggested next steps for every account. That's already useful. But the real edge is in what you ask next.
This is where you use your curiosity as an edge. The AI still has all of your data loaded. It knows the scores, the evidence, the transcripts — everything. All you have to do is keep the conversation going.
Here are a few things to try: Copy any of these into the same chat where you ran your analysis (or ask anything else you want):
Give me a full action plan for [account name from your results].
The analysis gave you a one-line next step for each account. This takes it further — a full playbook with specific outreach angles, timelines, and talk tracks you can hand directly to a rep.
Give me details on the dimension scores for [account name].
If you're curious how the AI arrived at a score — what evidence it used, how it weighted each dimension — just ask. You'll see exactly what went into the number and where the data was strong or thin.
What else should I be looking into for these results?
When you're not sure where to go next, ask. The AI can suggest angles you haven't thought of — specific accounts to compare, patterns across your list, or gaps in the data worth filling. This is the habit that makes everything else work: stay curious, keep asking.
Watch For
If any of these show up in your results, that's a good reason to dig in. Pick an account and ask a follow-up question to learn more.
| If you see... | It probably means... | What to try |
|---|---|---|
| All scores between 50–70 | Not enough differentiation — the model didn't have enough data to separate accounts clearly | Re-run with a thinking model, or add more data (transcripts, CRM notes) for those accounts |
| Quotes or evidence you can't find in your original data | Wrong model type — fast models fabricate evidence | Switch to a thinking/reasoning model and re-run the analysis (see the Data Prep Checklist) |
| Generic reasons like "growth potential" | The model didn't engage with your specific data — it fell back on filler | Check that your files uploaded correctly and re-run |
| Vague actions like "follow up" or "schedule a call" | The data was too thin for the model to get specific | Add CRM notes or transcripts for those accounts and re-run |
| Missing accounts from your list | The model skipped some | Ask: "Can you score the accounts you missed?" |
Every time you work with AI, ask one more question than you think you need to. That's the habit that compounds.
5.Next Steps
Share It With Your Managers and Reps
Send the ranked table and the full analysis to your managers and reps so they can take action. There's a lot they can do with this:
- •Use the suggested action steps directly. Every account has a specific next step — they can run with those right away.
- •Give the full analysis to AI and create action plans. They can take the output into their own AI chat and build out detailed playbooks, outreach timelines, and talk tracks for any account on the list.
- •Dig into the recommendations and evidence. If they want to understand how an account was ranked — or disagree with one — they can ask the AI to break down the scores and show the evidence behind them.
This isn't just an update to their account strategy. It's a chance for your managers and reps to start leveraging AI themselves.
Act on It
Act Now accounts: this week. These have real signals with real timing. Assign owners and get moving.
Develop tier: Figure out what's missing and work with the rep to close the gap over the next 30 days.
Monitor and Maintain: Don't invest selling time here unless something changes.
Run It Again
You're capped at 25 accounts per run, but you can run as many batches as you want. Try slicing by:
- •Rep territory — give each rep their own ranked list
- •Vertical or segment — compare healthcare accounts vs. commercial vs. education
- •Account tier — run your A accounts first, then your B accounts
How often: Re-run weekly, monthly, or quarterly — whatever fits your pipeline rhythm. Update your CRM notes and pull fresh external research each time. The more current your data, the sharper the results.