The fastest small-business sales wins from Claude aren't "AI writes my outreach." They're killing bad deals faster, walking into every call with a 90-second brief, and scoring inbound against the buyers you've actually closed — not the buyers in a slide deck. This page is the blueprint, with twelve prompts, a project template, the inbound triage workflow walked through end-to-end, and a real intent-signal outbound pipeline.
Block 60 minutes every Friday afternoon. Open your Sales Playbook project, paste your current pipeline, and walk it with Claude. The goal isn't to advance every deal — it's to identify which deals you should walk, which one stalled deal deserves the unsticking action, and what your one ask for next week is. The bias is toward killing, not carrying.
The temptation in sales is to keep every deal alive forever, just in case. That's how pipelines bloat and reps spend their week on deals that will never close while the deals that would close get under-attended. The Pipeline Hour exists to reverse that bias once a week, in a structured way, with a sparring partner who has no ego invested in any specific deal.
In the first 30 minutes you walk every open deal. In the next 20 you draft the one action for the one deal that matters most this week. In the last 10 you do the lost-deal post-mortem on whatever closed-lost this week. Done.
"Walk my pipeline. Which one deal should I close-lost this week — not maybe, but actually call it? Use my won/lost patterns in the project. Be willing to disagree with me."
"Which open deal has been idle longest with no real movement? Draft the unsticking action — the specific message, what it says, and what to do if they don't respond in 5 days."
"Which deal am I dragging out longer than I should — referencing the list in the project? Tell me the conversation I'm avoiding and how to start it on Monday."
"Which open proposal is priced lower than our won-deal pattern says it should be? What's the conversation to raise it before it closes?"
"For the deals that closed-lost this week, give me the 5-line post-mortem on each: stage it died, real reason, signal we missed, what we'd do differently, whether this is a pattern."
"Which deal needs the owner / a senior person to step in this week? Draft the brief for them — 60 seconds of context, the ask, what good looks like."
Why this works. Pipelines don't decay because reps are lazy. They decay because nobody wants to be the person who killed the deal. The Pipeline Hour structures that decision and externalizes it — you and Claude, looking at the same data, in the same room, once a week. Walking a deal is a discipline. The Pipeline Hour is where you practice it.
The single biggest source of generic AI sales work is generic context. Most reps describe their ICP in marketing language ("operations leaders at growing SMBs") and then ask Claude to score leads against it. The result is scoring noise. The fix is the won-deal log — twelve real closes, with the real reason each one bought.
Download the template. Fill it in — 60 to 90 minutes the first time, because the won-deal and lost-deal logs take work. They're worth it. Upload your last 12 won proposals and 12 lost proposals as knowledge files; Claude can pattern-match the structure even from your raw docs.
Open Claude → Projects → New Project → "Sales Playbook." Paste the file. Upload the proposals. Now every scoring prompt scores against your actual close pattern, every outreach prompt drafts in your voice, every objection-handler prompt references what has actually worked.
The won/lost deal log is the section that does the work. Without it, every prompt below scores against an idealized ICP and produces a confident-looking lead score that's worth nothing. With it, "score this lead" returns "this lead looks like Beta Inc and Charlie Corp — both bought in 60–80 days after their champion moved from a competitor; check if your champion here has that pattern" — which is a real answer.
Used inside your Sales Playbook project so Claude already has the ICP, the won/lost patterns, the objections, and your voice in context. Fill in the [bracketed] specifics and copy-paste.
I have a meeting with [NAME] from [COMPANY] in [TIME]. The meeting is about [TOPIC]. Using public information (LinkedIn: [URL], company site: [URL], any recent news), give me: 1. Three things I should know about them before walking in 2. Two things about their company that aren't obvious from the homepage — what's actually going on (hiring, funding, tech stack changes, recent press) 3. Where they sit relative to our ICP and won-deal pattern — closer to a buyer who closed or a buyer who walked? 4. The most likely thing they want from this conversation 5. Three good discovery questions to ask, drawing from our discovery framework in the project 6. One thing I should NOT say given my position Length: under 400 words. I want to skim this in 90 seconds before walking in.
New lead came in: [NAME, COMPANY, ROLE, ANY FORM-FILL CONTEXT, ANY URLs]
Score them against our actual won-deal pattern (not the aspirational ICP). Output:
1. Closest match: which of our last 12 wins does this lead most resemble, and why? Name the specific deal.
2. Closest miss: which lost deal does this look like, and what was the failure mode there?
3. ICP fit: 1–10. Be specific about why — title, company shape, trigger event match, budget authority.
4. The single signal that would move the score up by 2 points.
5. Recommendation: discovery call, async qualifying questions first, or walk. Bias toward walk if the lost-deal match is strong.
Don't soften the score to be encouraging. A 4 is a 4. I want to reach out to [PERSON] at [COMPANY]. The signal I have: [PASTE — could be a job posting, a press release, a podcast they were on, a tech stack change, a LinkedIn post, ANYTHING specific]. Draft a cold email that: 1. Opens with the specific signal — not "I came across your company" but the actual thing I noticed 2. Connects the signal to a real problem we solve (in the project) 3. Names one piece of evidence that we've helped someone similar (use the won-deal log; cite a real one) 4. Ends with a low-friction ask — not "want to hop on a call?" but something more specific 5. Under 90 words total. Subject under 7 words. Rules: No "hope this finds you well." No "I wanted to reach out." No generic flattery. If the signal isn't strong enough to ground a real email, tell me — don't fake it.
Attached: my current pipeline (deal name, stage, days in stage, deal size, last activity, next action). Walk it with me. For each deal, give me: 1. Status read: real or stalled? (Days in stage and last meaningful activity tell the truth.) 2. The one thing I'm probably wrong about on this deal 3. The single next action Then aggregate: - Which deal should I kill this week — call it close-lost? Name it. - Which stalled deal deserves the unsticking action this week? Name it. - Which deal am I probably under-pricing relative to the won-deal pattern? - Which one deal needs an escalation to a more senior person? - Which one deal am I dragging out longer than I should (cross-reference the project's "dragging" list)? Bias toward walking deals, not carrying them. If you're not confident, say "real" — but don't soften the call when you're confident the deal is dead.
Deal: [NAME]. Stage: [STAGE]. Days idle: [N]. Last meaningful contact: [DATE + SUMMARY]. What I know about the buying process: [ANY CONTEXT — committee, champion strength, competing priorities]. Draft the unsticking action: 1. Diagnose first — what's the most likely reason this is stalled? (Champion lost momentum, internal priority shift, considering a competitor, can't get budget approval, decided not to decide.) 2. Based on the diagnosis, the right move — a specific message, a specific call, or "walk and put on follow-up in 90 days." 3. If a message: draft it. Under 100 words. Direct, no false casualness, no "just checking in." 4. The 5-day rule: if no response by [date], what's the next move — and what's the move after that? 5. The point at which I should call it dead. If your diagnosis is "this deal is dead," say so. The unsticking action might be the close-lost email.
Deal: [NAME]. Closed-lost on [DATE]. Their stated reason: [WHAT THEY SAID]. My actual notes from the cycle: [PASTE]. Give me the 5-line post-mortem: 1. The stage it actually died at (not where it sat in CRM — where the buyer mentally checked out) 2. The real reason — which often differs from the stated reason. Use the lost-deal pattern in the project. 3. The earliest signal we missed 4. What we'd do differently — one specific change, not "improve discovery" 5. Whether this is a pattern (cross-reference the lost-deal log) or a one-off End with: should this go into the lost-deal log as a new pattern entry? If yes, draft the entry.
Attached: notes from the discovery call(s) with [PROSPECT]. Deal size estimate: [$X]. Draft the proposal using our standard structure (in the project). Rules: 1. Lead with their words — the way they described the problem in discovery, not our version of it 2. Frame outcomes, not features. Use the exact metrics they said they care about. 3. Scope what we're doing AND what we're explicitly not — boundaries protect closure 4. Price per our pricing in the project. If you're recommending a tier different from what they asked about, flag it. 5. Include the one piece of evidence (case study, customer quote, data) most relevant to their specific situation. Don't dump our generic proof points. 6. End with the decision they're being asked to make and by when — not "let me know what you think." After the draft, tell me the one section a competitor would attack and how to harden it.
Just got off a call with [PROSPECT]. Notes / transcript: [PASTE]. Draft the same-day follow-up. Structure: 1. One sentence that names what we agreed on — not "great chatting" 2. The 2–3 specific things they said they care about (use their language) 3. What I'm sending now (attach / link) 4. What I'm sending later (with a specific date) 5. The next step we agreed to, with a specific time proposed 6. One question that surfaces something I didn't fully understand on the call — earned curiosity, not stalling Length: under 150 words. No "thanks again for your time." Direct, specific, structured. Also: update the CRM note with the 3 things I should remember about this prospect for the next call — separately from the email.
I haven't heard back from [PROSPECT] in [N] days. Last meaningful contact: [SUMMARY]. They were at stage [STAGE] when they went quiet. Draft a re-engagement message that: 1. Doesn't say "just checking in" — earn the response with something specific 2. Acknowledges the silence directly without being passive-aggressive 3. Gives them an easy "out" — yes I'm still interested, no I'm not, the timing's not right 4. Includes one new piece of value or context that wasn't in the last conversation (a relevant insight, a customer signal, a new piece of evidence) 5. Under 80 words If the silence has been long enough that the right move is to send the breakup email instead, draft that and tell me to send it.
Deal: [NAME]. They're asking for a [X]% discount or [OTHER CONCESSION]. The deal is otherwise at [STAGE]. Their stated reason: [WHAT THEY SAID]. Help me think about this: 1. Is the ask real, or is "price" the polite version of a different concern? Check against our objection library. 2. If I give the discount, what's the most likely outcome — does the deal close, or does it move to a new objection? 3. What's a counter-offer that holds price but moves something else? (Annual prepay, case study, multi-year, logo rights, smaller scope.) 4. What's our discount latitude per the project, and is this deal worth using it on? 5. The line I won't cross — what's a "no" to this ask? End with: what would you do, if this were your deal?
Prospect said: [PASTE OBJECTION VERBATIM]. Context: [STAGE, RELATIONSHIP, WHAT THEY'VE SEEN FROM US]. Help me handle it: 1. Decode it: what are they likely actually saying? Map to our objection library if there's a match — but flag if this is a new variant. 2. The wrong response — what most reps would say back, and why it'd fail here 3. The right response — drawing from what's worked in our objection library, calibrated to this specific stage and prospect 4. The follow-up question that gets us to the real concern underneath this objection 5. The "they didn't budge" path — if my response lands flat, what's next End with: should this be added to the objection library as a new variant?
Pull the last 20 closed deals from the project — 12 won, lost ones, and any open deals over [N] days.
Refine our ICP:
1. What pattern across the 12 wins is not in our current ICP description? (Often the surprise — geography, trigger event, decision-maker shape.)
2. What's in our current ICP description but contradicted by the wins? (The aspirational lines.)
3. Lost deals — were they bad-fit, or did we lose winnable deals to execution? Be specific.
4. The one disqualifier we should add — the kind of prospect we should walk from on first contact.
5. The one qualifier we should add — the signal that should fast-track a discovery call.
End with the proposed updated ICP paragraph for the project. I'll edit and approve. Prompt 01 is the gateway drug. Every other prompt is good. Prompt 01 — the pre-meeting brief — is the one that turns AI-curious reps into AI-leveraged reps in a single week. Run it before every external call for two weeks. By week three you won't take a call without it.
The integration with the highest ROI for sales: when a new lead hits your form, Claude pulls the HubSpot record, enriches it from public sources, scores it against your won-deal pattern, and writes a 60-second pre-meeting brief — before you've even seen the notification. Setup: ~60 minutes. The first week you use it, you'll stop taking discovery calls cold.
In your Sales Playbook project → Connect apps → HubSpot (read access on Contacts, Deals, Forms, Companies; write access on Contact notes if you want Claude to attach the brief directly). If you're on a different CRM — Pipedrive, Close, Attio — the same pattern works wherever you have an integration available.
Read access does most of the work. Write access on notes is the convenience layer — the brief auto-attaches to the contact instead of you copy-pasting it. Worth it.
Save this as "New inbound — triage" inside the project. You'll run it for every new lead.
A new lead just came in. Their HubSpot contact ID / details: [PASTE OR LET CLAUDE PULL FROM HUBSPOT].
Produce the triage in this exact structure:
**1. ICP fit (1–10) with reasoning**
Score against the won-deal pattern in this project — not the ICP description. Name the specific past deal they most resemble. If they look more like a lost-deal pattern, say so.
**2. The signal in the form fill / source**
What did they say or do that's the actual buying signal? (The form-fill question they answered, the page they came from, the referrer.) If the signal is weak, name that.
**3. Public enrichment — what's actually going on at the company**
Pull from LinkedIn and their company site: recent hires, recent press, tech stack hints, growth signals. Three bullets max. No filler.
**4. Three discovery questions tailored to this lead**
From our discovery framework, the three most relevant for what we know about them. Not generic; specific to their situation.
**5. The pre-meeting brief (under 200 words)**
What I'd want to read 90 seconds before the call. Hook, pain hypothesis, what to listen for, one thing not to say.
**6. The routing call**
- "Book the discovery call" — they're a real lead, the brief is ready
- "Async qualify first" — score is borderline, send these 2 questions before booking
- "Walk" — they look like a lost-deal pattern, here's the polite decline
Attach the full output to the HubSpot contact as a note (if write access is enabled). Don't wait for a fresh lead. Run the prompt against the last 5 inbound leads you've handled — including the ones you walked from and the ones that closed. Two checks:
ICP fit: 8/10. Closest match: Beta Inc — ops director at a 80-person services firm, came in with a specific pain (multi-system reporting takes 4 days/month), closed in 67 days. This lead matches on company shape, role, and the specific pain in the form-fill.
Signal in form fill: "Spending 3 days a month reconciling between QBO and our PM tool" — specific, named, time-quantified. Strong signal.
What's going on at the company:
Three discovery questions: (1) Walk me through what the 3-day reconciliation actually involves. (2) What did the new Director of Finance say should change first? (3) What have you tried before us, and what didn't work?
Pre-meeting brief: They're a strong-fit warm inbound, ops-led with a new finance hire involved. The pain is specific and quantified. Listen for: who else is in the buying group (likely the new Director of Finance plus the COO). One thing not to say: don't lead with our PM-tool integration — they just adopted it and are looking for a fix, not a replacement.
Routing call: Book the discovery call. Brief is attached to HubSpot.
When a lead hits the form, you run the prompt. By the time you've finished your coffee, you have the score, the brief, and the routing call. You stop taking discovery calls cold. You stop spending 20 minutes manually qualifying a lead that scored 3/10. Multiply by every new lead this quarter — that's the integration paying for itself.
Keep yourself in the loop on the "walk" decision. The triage proposes it; you confirm. Walking a lead is the kind of irreversible move where the human stays in the loop forever.
What this isn't. It isn't an autonomous SDR. It's a 5-minute prep step that runs in parallel with you reading the lead notification. The judgment Claude can't bring is the read on the human in the meeting — but it can do 90% of the prep before you do any of it. Automate the typing, keep the deciding.
Once L1–L3 are real, L4 is where Claude operates on a schedule, manages the inbox, and runs full outbound pipelines without you holding the prompt. Three sketches, ordered from least to most aggressive.
Every morning at 7am, Claude scans your pipeline against your won-deal pattern and the rules in your project. It DMs you one Slack message: which deals shifted overnight, which stalled deals crossed your "should have moved by now" threshold, which proposals haven't been opened in 5 days, and the single highest-ROI action for today.
Movement overnight: Acme Corp opened the proposal (3rd time). Beta Inc rescheduled the demo to Thursday (this is the 2nd reschedule — soft signal of declining urgency).
Stalled threshold crossed: Charlie Corp is now at 16 days since last meaningful contact. Per your stall threshold (14 days), this needs an unsticking action today. Draft is ready in the thread.
Proposal aging: Delta LLC's proposal sent 6 days ago, opened twice, no response. Per your follow-up rhythm, the day-7 nudge fires tomorrow — preview it now if you want to adjust.
Today's highest-ROI action: The Charlie Corp unsticking message. Their champion went quiet 3 weeks after their CFO transition — historical pattern in your lost-deal log suggests this is when they decide internally. Catch it before that.
Inbound replies — to your outbound, to your proposals, to your follow-ups — auto-classified. Positive replies route to you immediately with a suggested next step. Soft passes get auto-replied with a polite "noted, can we revisit in Q3?" message that goes into your CRM as a future-task. Out-of-office and "wrong person" replies get auto-routed without ever touching your inbox.
The Reply Manager is the highest-leverage L4 for solo founders doing sales — it's the difference between "I spend 90 minutes a day in my inbox" and "I spend 15 minutes a day on the 8 replies that actually matter."
The most aggressive — full automated outbound based on real-time buying signals, not bought lists. A real working pipeline a serious sales operator can build today:
Signal scraped: A 200-person logistics company posted a job for "Director of Revenue Operations" yesterday. JD specifically calls out "build out reporting infrastructure across HubSpot, NetSuite, and our 3PL tools."
Subject: The RevOps director hire — three things they'll find in their first 30 days
Body: Maya — saw the RevOps director req went up yesterday. The "build out reporting across HubSpot, NetSuite, and 3PL tools" line in the JD is going to be the new hire's first 90-day fire. Three things they're going to find in week 2 (we've done this with [similar company] — happy to walk you through what we learned): [1] the 3PL data won't reconcile cleanly with NetSuite without a transform layer, [2] HubSpot deal stages drift away from NetSuite stages in 4–6 weeks if not stitched, [3] the "single source of truth" conversation will happen and someone has to lose. Worth a 20-minute call before they start, or right after?
Why it's not template-able: Claude read the actual job description, identified the specific tool stack, and built a credible "future fire" argument from it. The angle came from the prospect's own posting — not a hook library.
The outcomes are real. New intent signals get pulled daily, enriched, scored, outreached, and managed automatically. What used to be a full-time SDR is now a cron job for the cost of the APIs.
The catch: this is the most expensive L4 build to get wrong, because the failure mode is public — bad cold emails at scale damage your sender reputation, your domain, and your brand simultaneously. The per-prospect Claude call quality is the difference between personalized outbound and the highest-volume spam your category has ever produced. Your won-deal log and ICP discipline are doing the load-bearing work here; if those are weak, the pipeline amplifies the weakness.
Build this last. Pipeline Watch first, then Reply Manager, then this. If your Pipeline Hour isn't producing wins, the automated outbound won't either.
If you're imagining any of these three right now — pause. The honest path is: spend a quarter at L2–L3 with the Pipeline Hour. By the second quarter you'll know which of the three fits your motion and which would be a toy. The build is then weeks of work, not months. When you're at that point, that's a good call to have.
Sales is the role where AI compounds fastest and destroys brand fastest. The wins and the losses share the same root: every prospect interaction is public. Here are the patterns to watch in yourself.
"Hi {{first_name}}, I noticed your company {{company}}…" Sending mass outbound that's clearly AI-drafted, with no real signal, is the single fastest way to burn your sender reputation and your brand.
Every outbound message needs a real signal (Prompt 03) or it shouldn't go. The L4 outbound pipeline above works because step 4 — read the actual page — is non-negotiable.
The project gets built with a marketing ICP ("operations leaders at growth-stage SMBs") instead of the won-deal log. Every lead score is generic. The integration looks great in demos and produces noise in practice.
The won/lost deal log is the load-bearing section. Without it, no scoring prompt is worth running.
Prompt 01 takes 90 seconds. You skip it on the "easy" call. You get caught off guard by something you'd have known if you'd run it. Repeat enough times and you've quietly normalized walking into calls cold.
Run it before every external call for two weeks. By week three you won't take a call without it — and your close rates will tell you why.
The Pipeline Hour exists to kill deals weekly. If your pipeline never shrinks, you're not running the Pipeline Hour right — or you're not honest enough in the "deals I'm dragging" list.
A clean pipeline closes faster than a bloated one. The deals you walked from this quarter are the reason next quarter's deals close — your attention had room.
Letting Claude draft the proposal, the negotiation reply, the discount counter — and sending it without judgment-edits. AI is a prep tool and a draft tool. It is not a closer. The buyer can tell when you let it close.
Verify everything that leaves your business. Drafts auto-generate. Sends — especially anything with pricing or commitment in them — stay human.
Building the Intent-Signal Outbound Pipeline before your won-deal log is real and your manual outbound (Prompt 03) consistently produces replies. The pipeline amplifies whatever it's amplifying — and if that's a thin voice on a fuzzy ICP, you've just industrialized your worst sales work.
L1–L3 first, including a quarter of running the Pipeline Hour. The ladder exists for a reason.
Download the template. Fill in the won/lost log honestly. Save the prompts. Brief every call. Walk every Friday. In 90 days your close rate will tell you everything the AI consultants won't.