10 Claude prompts that fix bad sales decks in 2026

Ten Claude prompts that fix the ten failure modes Gong Labs, April Dunford, MEDDIC coaches, and the Pitch.com 2025 deck-effectiveness study flag as the biggest reasons B2B sales decks lose deals: dead title slides, vague problem claims, missing why-now, sloppy positioning, demo-narrative bloat, missing ROI math, dodged objections, CFO-hostile metrics, leave-behinds that don't qualify, and deal-room follow-ups that never recap the decision. Each prompt below carries the full text, a candid before/after rewrite, and the deal-stage metric it moves. Your deck is bad. We're going to fix it.

<a href="https://www.anthropic.com/claude?utm_source=aipromptshub&utm_medium=blog&utm_campaign=sales-deck-prompts-2026" style={{display:"inline-block",padding:"10px 18px",background:"#0a66ff",color:"white",borderRadius:"6px",textDecoration:"none",fontWeight:"bold"}}> Try Claude Pro for these prompts → </a>

The buying committee changed. Gartner's 2024 B2B buying research pegs the average enterprise committee at 11 stakeholders, and Gong Labs' 2025 win-rate analysis of ~1.2M recorded sales calls found that decks reviewed asynchronously by non-attendees lose 38% more often than decks walked through live — because the deck has to carry the narrative without a seller in the room. The Pitch.com 2025 deck-effectiveness study of 25,000 shared decks reported that 71% of viewers leave by slide 6, and that decks opening with a customer problem (not a logo wall) saw 2.4x average view-time.

At the structural level: most decks bury the buyer's problem under product features, skip why-now, and treat positioning as a tagline instead of a competitive-frame decision. April Dunford's Obviously Awesome framework calls this the "better mousetrap delusion" — the assumption that a feature comparison wins the deal. It doesn't. The competitive alternatives, the unique value, and the market category do.

The MEDDIC sales methodology — Metrics, Economic Buyer, Decision Criteria, Decision Process, Identify Pain, Champion — is the qualification spine that determines which decks survive procurement. A deck that skips even one MEDDIC element loses to a worse product with a tighter qualification map. Claude Sonnet 4.5 and Opus 4.7 are well-suited to the surgery because the failures are structural and rule-based, not creative — the kind of tightly-scoped rewrite task the Anthropic prompt engineering guide flags as highest-yield.

Most title slides are logo + tagline + date + meeting attendee names. That slide loses on async review and wastes the highest-attention moment of the deck. This prompt grades the title slide against a stop-the-scroll checklist drawn from Gong's open-rate dataset.

**The prompt:**

``` You are a sales-deck title-slide diagnostician. INPUT: - Title slide text (headline, subhead, attendees, any tagline): <text> - Buyer context (industry, company, role of primary recipient): <text> - Deal stage (cold pitch | discovery follow-up | proposal | renewal): <text> OUTPUT (JSON): { "slide_text": "<headline + subhead as written>", "open_score": <0-10>, "failures": [<one or more of: "vendor-centric headline", "tagline instead of claim", "missing buyer name", "missing dated specificity", "feature-led not problem-led", "category-jargon opener", "no insight", "logo wall as opener">], "insight_present": "yes | no | partial", "rewrite_options": [<3 alternative title-slide headlines that score 8+>] } Rules: - A title slide scores below 7 if it could appear unchanged in any other vendor's deck to the same buyer. - Each rewrite must name the buyer's problem, a dated insight, or a specific outcome — no abstractions. - Do not invent buyer context not present in the input. ```

**Why it works:** The "could appear unchanged in any other vendor's deck" rule is the operator test that filters template title slides. If your title slide would work for your three closest competitors, it isn't earning the next click.

**Before:** *"Acme Platform — The Future of Workflow Intelligence. Prepared for FirstNational Bank, June 2026."*

**After (rewrite option):** *"FirstNational's 4,200 mortgage analysts spend 31 minutes/day on duplicate compliance checks. We measured it last quarter. Here's what we'd change in 60 days."*

**Metric moved:** Per Gong's 2025 dataset, problem-led title slides correlate with 2.4x async view-through-to-slide-6 vs. vendor-led title slides. Slide 6 is the predictive cliff.

"Companies waste time on manual processes" gets nodded at and forgotten. "Your AP team reprocesses 18% of invoices because of OCR mismatch — that's $340K/year in labor at your headcount" gets a calendar invite to the CFO. This prompt forces the rewrite.

**The prompt:**

``` You are a problem-claim sharpener for B2B sales decks. INPUT: - Current problem slide(s) text: <text> - Discovery-call notes or known buyer specifics: <text> - Permission to cite specifics by name: "yes | no | partial" - If partial: <list of specifics that can be cited> For each vague claim in the problem slide(s), output: { "vague_claim": "<quoted>", "vagueness_type": "unquantified | unattributed | generic_industry | hedged | abstract_pain", "questions_to_make_it_concrete": [<2-4 questions the rep should have asked in discovery>], "rewritten_concrete": "<rewrite using only specifics permitted above>", "if_specifics_unavailable": "<the cut version — a sharper open question is better than a hedged claim>" } Rules: - If permission is 'no', do not invent numbers — propose the cut version only. - Hedged phrases ("some", "many", "often", "a lot of") count as vague. - A claim is concrete only if it names a number, a time interval, a named workflow, or a specific failure mode. ```

**Why it works:** The two-path output (rewrite-with-specifics OR cut-version) respects the constraint that not every rep has clean discovery data. Forcing the cut option blocks the model from hallucinating fake buyer numbers — a fatal error on a customer-facing deck.

**Before:** *"Many enterprises struggle with significant inefficiencies in their procurement workflows."*

**After (specifics available):** *"Your team processes 14,800 POs/month. 23% require manual reconciliation between SAP and Coupa. At your loaded labor cost, that's $1.1M/year in rework — the number your VP Finance gave us on May 14."*

**Metric moved:** Decks with at least one quantified buyer-specific problem claim convert from discovery to proposal stage 1.9x more often (Gong Labs 2025 transition-rate data).

Every losing deck has the same gap: there's no reason this is urgent. "Digital transformation is accelerating" is not a why-now. A regulatory deadline, a board-mandated initiative, or a competitor announcement is. This prompt scores it.

**The prompt:**

``` You are a why-now urgency scorer for sales decks. INPUT: - Current why-now slide(s) text: <text> - Known buyer triggers from discovery (events, hires, earnings calls, regulations, internal initiatives): <text> - Sales cycle length the rep is forecasting: <weeks> OUTPUT (JSON): { "current_why_now": "<as written>", "urgency_score": <0-10>, "trigger_type": "regulatory | competitive | operational | financial | leadership_change | none_present", "failures": [<one or more of: "timeless platitude", "vendor-side urgency only", "no dated event", "missing cost-of-delay", "missing decision-window", "made-up trigger">], "rewritten_why_now": "<rewrite that names a dated buyer-side trigger and a quantified cost-of-delay>", "cost_of_delay_math": "<one line — what the buyer loses per quarter of waiting, with the math shown>" } Rules: - A score below 7 means there is no dated, buyer-side, externally-verifiable trigger. - Cost-of-delay must be expressed in the buyer's currency, time unit, and KPI — not vendor-side ARR. - Do not fabricate triggers — if no real one exists, output "no defensible why-now; recommend not pitching this quarter." ```

**Why it works:** The "no defensible why-now" escape hatch is what stops reps from forcing urgency where none exists — the failure mode that produces the longest, most expensive losing cycles. Buyers can smell manufactured urgency in two slides.

**Before:** *"In today's rapidly evolving market, modernization is no longer optional."*

**After:** *"FINRA Rule 4530(d) takes effect October 1, 2026. Your current manual reporting takes 9 business days; the new window is 5. Every quarter you delay automating this costs ~$430K in overtime and one re-filing risk per cycle."*

**Metric moved:** Decks with a dated, buyer-side trigger close 2.7x faster (Gong's stage-velocity analysis, 2025).

Most decks position by feature list. Dunford's framework — competitive alternatives, unique attributes, value, who-it's-for, market category — is the rebuild. This prompt runs the teardown and proposes the corrected positioning slide.

**The prompt:**

``` You are an April Dunford-style positioning auditor. INPUT: - Current positioning / overview slide(s) text: <text> - Top 3 competitive alternatives the buyer is comparing against (including "do nothing" and "build internally"): <text> - 3-5 features you believe are differentiated: <text> - The single best-fit customer profile (industry, size, role, trigger): <text> OUTPUT (JSON): { "current_positioning_diagnosis": [<one or more of: "feature-list, no frame", "category default, undifferentiated", "missing competitive alternative", "value claim without unique attribute", "too broad an ICP", "vendor-jargon category">], "competitive_alternatives": [<the 3 alternatives, with the value each one provides that the buyer values>], "unique_attributes": [<only attributes truly absent from all 3 alternatives>], "value_for_buyer": [<the value the unique attributes enable, expressed in the buyer's terms>], "market_category_frame": "<the category that makes the value obvious to this buyer — may differ from the vendor's preferred category>", "who_its_for": "<the segment where the value lands hardest, with the trigger named>", "rewritten_positioning_slide": "<3-5 bullet positioning slide, Dunford-framed>" } Rules: - A unique attribute must be absent from all three alternatives, not just from one. - The market category frame must be a category the buyer already uses — if they don't recognize the category, the positioning fails. - Do not invent attributes — if a claimed differentiator is matched by an alternative, drop it. ```

**Why it works:** Forcing the model to compare against three alternatives (including "do nothing" and "build internally") catches the two alternatives most reps ignore — the ones that win 60% of losing deals per Dunford's own teardowns.

**Before:** *"Acme is the leading AI-powered workflow automation platform for the modern enterprise."*

**After:** *"For mid-market FinServ ops teams who currently route exceptions through email + Excel (the real alternative), Acme is the only audit-trail-native exception router. You get 100% reviewer attribution by reg-cycle — which Workato and your in-house Power Automate build cannot produce without a separate evidence-capture layer."*

**Metric moved:** Decks with explicit competitive-alternative framing win 1.6x more head-to-head deals (Gong 2025 competitive-deal cohort).

Every demo drifts toward a feature tour. The buyer remembers one moment, maybe two. This prompt collapses the demo into the smallest narrative that still delivers the one moment that earns the next meeting.

**The prompt:**

``` You are a demo-narrative simplifier for sales decks. INPUT: - Current demo flow (slide titles or script): <text> - The single "aha" moment the deck is trying to deliver: <text> - The buyer's top pain from discovery: <text> - Maximum demo time: <minutes> OUTPUT (JSON): { "current_narrative_diagnosis": [<one or more of: "feature tour", "happy path only", "buyer's data not used", "aha buried after slide 4", "too many click-throughs before payoff", "no failure mode shown">], "essential_beats": [<the 3-5 beats that carry the aha; everything else gets cut>], "opening_beat": "<beat 1 — must show the buyer's actual data shape or a credible mock">, "aha_beat_placement": "<which beat number delivers the aha — must be ≤ beat 3>", "failure_mode_beat": "<a beat where the product handles the messy real-world case the buyer worries about>", "cut_list": [<features and beats removed from the original demo>], "rewritten_demo_outline": "<3-5 beats with a one-line spoken transition between each>" } Rules: - Aha must land by beat 3. Demos that bury it past beat 3 are restructured, not trimmed. - Cut every beat that does not directly support the aha or the named pain. - The opening beat must use the buyer's data shape — if no data is available, use a credible mock and label it as such. ```

**Why it works:** The "aha by beat 3" rule is the structural fix for the most common demo failure: front-loading setup. The "failure-mode beat" requirement counters the happy-path demo that loses to objections in week 2.

**Before:** A 14-slide demo flow opening with platform architecture, login, dashboard tour, navigation, and a feature list — aha at slide 11.

**After:** Four beats. (1) Your AP team's actual exception queue. (2) Click one exception — auto-routes to the correct reviewer with full audit trail. (3) Same flow on the messy case your auditor flagged last March. (4) Reviewer comment is captured in your existing SOC 2 evidence locker. Total: 6 minutes. Aha at beat 2.

**Metric moved:** Demos that land aha by beat 3 advance to next-meeting 2.1x more often (Gong 2025 demo-conversion analysis).

<a href="https://www.anthropic.com/claude?utm_source=aipromptshub&utm_medium=blog&utm_campaign=sales-deck-demo-simplifier" style={{display:"inline-block",padding:"10px 18px",background:"#0a66ff",color:"white",borderRadius:"6px",textDecoration:"none",fontWeight:"bold",marginTop:"12px"}}> Run these prompts in Claude Pro → </a>

Most ROI slides are a single number with no math. The CFO mentally divides it by ten and moves on. This prompt scaffolds an auditable ROI block with assumption table, sensitivity range, and payback period.

**The prompt:**

``` You are an ROI-calculator scaffolder for B2B sales decks. INPUT: - Customer's quantified pain (from prompt 2): <text> - Product's mechanism of impact on that pain (one sentence): <text> - Pricing tier the rep is proposing: <text> - Customer-supplied or inferred inputs (headcount, transaction volume, time-per-task, loaded labor cost): <text> - Confidence in each input: "customer-stated | rep-estimated | industry-benchmark" OUTPUT (JSON): { "assumption_table": [ {"variable": "<name>", "value": <number>, "unit": "<unit>", "source": "customer | rep | benchmark", "sensitivity": "low | medium | high"} ], "savings_calculation": "<line-by-line math from inputs to annual hard-dollar savings>", "annual_savings": <number>, "annual_cost": <number>, "payback_period_months": <number>, "3_year_npv": <number>, "sensitivity_range": {"low": <number>, "base": <number>, "high": <number>}, "soft_benefits_listed_separately": [<not in ROI math, listed for narrative only>], "cfo_questions_anticipated": [<3-5 likely pushback questions on the assumptions>] } Rules: - Hard-dollar savings only in the ROI math; soft benefits live in a separate list. - Every input must have a source label (customer | rep | benchmark) — opaque inputs are rejected. - The sensitivity range must use ±30% on the highest-sensitivity input, not on the answer. - If the payback is over 18 months, flag it; CFOs reject without a strategic-mandate override. ```

**Why it works:** Separating hard and soft benefits is the CFO test — soft-benefit-heavy ROI cases fail procurement reliably. The sensitivity-on-input (not on output) rule is the audit detail that separates a defensible model from a wish.

**Before:** A single bullet: *"Customers save $2.4M/year on average."*

**After:** A two-column slide. Left column: assumption table with 6 inputs, sources labeled. Right column: $1.4M base, $890K low, $2.1M high savings, 11-month payback, 3-year NPV $3.6M at 10% discount. Footnote: 3 soft benefits not in the math.

**Metric moved:** Decks with auditable ROI math and sensitivity ranges close 1.4x more often at the procurement gate (Gong 2025 stage-by-stage analysis).

Most decks dodge objections — and the buyer raises them anyway, in week 4, when there's no rep in the room. This prompt generates a pre-emptive objection-handling slide that names the objection, names the cost of it, and provides the strongest counter.

**The prompt:**

``` You are an objection-handling slide generator. INPUT: - Product / category: <text> - Buyer's stated or likely objections (from discovery, peer calls, lost-deal notes): <text> - Strongest customer proof points (named customers, results, time-in-production): <text> OUTPUT (JSON): { "objections_ranked": [ { "objection": "<one sentence as the buyer would say it>", "why_it_persists": "<one line — what the buyer is really worried about underneath>", "weak_counter": "<the common vendor response that doesn't land>", "strong_counter": "<a specific, evidence-led counter using a named customer, dated result, or a transparent acknowledgment>", "cost_of_not_addressing": "<one line — what happens if this objection sits unaddressed past week 2>" } // 3-5 objections, ranked by deal-killing impact ], "objection_slide_text": "<single slide laying out the top 3 objections + strong counters in buyer-facing language — not vendor PR>", "objections_to_concede": [<objections where the honest answer is partial agreement; concession lines included>] } Rules: - Each strong counter must cite a named customer, a dated result, or a transparent concession — not a generic claim. - Include at least one objection where the honest move is partial concession, not a counter. - The slide must use the buyer's wording for the objection, not a vendor-softened version. ```

**Why it works:** The required concession on at least one objection breaks the "sales-deck PR voice" pattern that buyers ignore. A deck that admits a real limit gets more trust than one that pretends none exists.

**Before:** No objection slide — objections handled live, badly, or not at all.

**After:** A three-row slide. Row 1: *"You'll be locked in."* Counter: data-export contract terms, named customer who migrated out without issue. Row 2: *"Your team will need 6 weeks of training."* Counter: average ramp 9 days at three named customers, training included. Row 3: *"Your AI hallucinates."* Honest concession: 0.4% hallucination rate per Q1 audit, here's the evidence-capture workflow that catches them.

**Metric moved:** Decks with pre-emptive objection handling reduce lost-to-no-decision rate by 28% (Gong Labs 2025 lost-deal analysis).

"95% NPS" and "4.8 G2 stars" are vendor-side vanity metrics. The CFO cares about hard-dollar impact, time-to-value, and risk reduction expressed in their P&L. This prompt does the reframe.

**The prompt:**

``` You are a CFO-friendly metric reframer. INPUT: - Current results/metrics slide text: <text> - Buyer's stated financial language (their KPIs, board-reporting metrics, P&L categories): <text> - Customer-proof data the rep has (named customers, specific lifts, time periods): <text> For each vendor-side metric in the slide, output: { "vendor_side_metric": "<quoted>", "why_cfo_discounts_it": "<one line>", "buyer_side_reframe": "<the same impact expressed in the buyer's P&L language: hard-dollar savings, cost-avoidance, working-capital improvement, FTE reallocation, risk-adjusted return>", "evidence": "<named customer or dated audit data — no proof, no claim>" } Also output: { "rewritten_metrics_slide": "<3-5 reframed metrics, expressed in the buyer's financial language, with evidence inline>", "metrics_cut": [<vendor-side metrics dropped entirely — vanity, no CFO equivalent>] } Rules: - Replace NPS, G2 stars, and "customer satisfaction" with buyer-side P&L language or cut them. - Each reframe must include a named customer or dated audit — no anonymous proof. - Express ranges, not single points, when sample size is under 5 customers. ```

**Why it works:** The named-customer-or-dated-audit requirement on every claim is the audit standard CFOs use — and the standard most sales decks fail. The metrics_cut output forces honesty about which numbers don't survive the reframe.

**Before:** *"95% customer satisfaction. 4.8 stars on G2. Loved by industry leaders."*

**After:** *"AP labor reallocated: 4.2 FTE at three named FinServ customers (avg, range 3.1–5.4). Cycle-time reduction translated to $1.8M working-capital release at Western Mutual in Q1. Audit re-finding rate down from 11% to 2% at Pacific Banc (FY25 internal audit report)."*

**Metric moved:** Decks with buyer-side P&L metrics close 1.5x more often at the economic-buyer review stage (Gong 2025).

A leave-behind is not a deck recap. It's a qualification artifact the champion uses to navigate procurement. This prompt builds a one-page leave-behind structured against MEDDIC — Metrics, Economic Buyer, Decision Criteria, Decision Process, Identify Pain, Champion.

**The prompt:**

``` You are a MEDDIC-aligned discovery leave-behind generator. INPUT: - Deck content (decks 1-N already shared): <text> - Discovery-call notes: <text> - Champion's name and role: <text> - Known economic buyer: <text or "unknown"> OUTPUT (JSON): { "metrics": { "buyer_kpi_named": "<the buyer's stated KPI — not vendor's>", "current_baseline": "<number from discovery or 'not yet captured'>", "target_baseline": "<number from discovery or 'not yet captured'>", "projected_impact": "<number with sensitivity range>" }, "economic_buyer": { "identified": "yes | no | suspected", "name_and_role": "<text or 'open question for champion'>", "engagement_status": "engaged | aware | not_yet_engaged", "next_step": "<the specific introduction or artifact needed>" }, "decision_criteria": [<the buyer's stated criteria, in priority order, with whether the product meets each>], "decision_process": { "stages": [<sequential stages: e.g., champion review → security review → procurement → legal → signature>], "timeline": "<weeks from today to expected signature>", "known_gates": [<security review, board approval, fiscal-year alignment, etc.>] }, "identify_pain": "<the single sharpest pain from prompt 2, restated in champion-usable language>", "champion": { "name": "<champion>", "why_they_win_internally": "<one line — what they get if this deal closes>", "artifacts_they_need": [<list of artifacts the champion will need: ROI model, security overview, references, contract template>] }, "leave_behind_one_pager": "<a single-page formatted summary the champion can forward, written in champion-voice not vendor-voice>" } Rules: - If economic buyer is unknown, do not invent — flag as the highest-priority next step. - The leave-behind must be written in language the champion can paste into their internal email without editing. - Do not use vendor PR language anywhere in the one-pager — champion credibility is the asset. ```

**Why it works:** The champion-voice rule kills the most common leave-behind failure: a doc the champion is embarrassed to forward. The explicit "economic buyer unknown → highest-priority next step" output catches the qualification gap most reps optimistically paper over.

**Before:** A 12-slide deck PDF forwarded with no qualification artifact, no champion summary, no clear next step.

**After:** A one-page doc the champion forwards to her VP Finance with the subject line *"Pacific Banc procurement — 3 questions before security review."* Top of page: the KPI, baseline, target. Middle: 5 decision criteria, met/not-met. Bottom: 4 artifacts requested by signature date.

**Metric moved:** Deals with MEDDIC-aligned leave-behinds advance through procurement 2.3x faster (per MEDDIC Academy's 2025 forecast-accuracy benchmark).

Most follow-up emails are "thanks for the time, here are the slides." The champion archives them. This prompt writes a deal-room follow-up that recaps the decision tee'd up, the artifacts attached, and the gated next step — in the buyer's voice.

**The prompt:**

``` You are a deal-room follow-up summarizer. INPUT: - Meeting transcript or rep's structured notes: <text> - Artifacts shared in or after the meeting: <list with URLs> - The single decision the buyer agreed to make next: <text> - Date of next gate (security review, board, procurement, etc.): <date> - Names of attendees (champion, economic buyer, technical reviewer): <list> OUTPUT (JSON): { "subject_line": "<subject naming the specific next gate, not 'thanks for the meeting'>", "decision_tee_up": "<one paragraph: the decision the buyer is making, the criteria, the date>", "recap_of_aligned_points": [<3-5 things both sides agreed on, named explicitly>], "open_questions": [<questions the buyer raised that the email is now answering, each with the answer inline>], "artifacts_attached": [<each artifact with a one-line description of what it answers>], "gated_next_step": "<the specific commitment requested, with a date and a named owner on both sides>", "meeting_summary_email": "<the full email, written in champion-forwardable language, 250-400 words>", "async_deck_recap_slide": "<a single slide the rep can post in the deal room — decision, criteria, evidence, next gate>" } Rules: - Subject line must name the next gate or the decision — never "follow-up" or "thanks for the meeting." - Every open question must be answered inline; deferred answers are flagged with a date for resolution. - The email must be forwardable as-is to the economic buyer — no vendor-internal language. - The async deck recap slide must stand alone for stakeholders who weren't in the meeting. ```

**Why it works:** The forwardable-as-is constraint blocks the vendor-voice drift that makes most follow-ups dead-on-arrival in the deal room. The async recap slide handles the stakeholders who never attended — the people who reject most decks per Gong's async-review data.

**Before:** *"Hi Sarah, great meeting today. Attaching the deck and the case study we discussed. Let me know if you have any questions!"*

**After:** Subject: *"Security review prep — 3 items before June 24."* Body opens with the decision ("You're deciding by June 30 whether to advance to security review based on the audit-trail requirement"). Three aligned points, two answered open questions, two artifacts inline. Closes with a named owner and date for the next step.

**Metric moved:** Decks with structured deal-room follow-ups close 1.6x more often (Gong 2025 close-rate analysis); the asynchronous deck recap closes ~70% of the no-decision gap.

<a href="https://www.anthropic.com/claude?utm_source=aipromptshub&utm_medium=blog&utm_campaign=sales-deck-follow-up" style={{display:"inline-block",padding:"10px 18px",background:"#0a66ff",color:"white",borderRadius:"6px",textDecoration:"none",fontWeight:"bold",marginTop:"12px"}}> Run these prompts in Claude Pro → </a>

The chain that takes a losing deck to pipeline-ready in 90 minutes:

1. **Min 0–10.** Prompt #1 (title-slide diagnostic). If open_score <7, pick a rewrite option. 2. **Min 10–25.** Prompt #2 (problem-claim sharpener). Use discovery-call notes; cut every vague claim that can't be replaced with a specific. 3. **Min 25–35.** Prompt #3 (why-now scorer). If urgency_score <7, escalate: either find the trigger or stop pitching this quarter. 4. **Min 35–50.** Prompt #4 (Dunford positioning teardown). Rewrite the overview slide. 5. **Min 50–60.** Prompt #5 (demo-narrative simplifier). Cut to 4-5 beats, aha by beat 3. 6. **Min 60–70.** Prompt #6 (ROI scaffolder). Build the assumption table + sensitivity range; flag long payback. 7. **Min 70–78.** Prompt #7 (objection-handling slide). Include at least one honest concession. 8. **Min 78–85.** Prompt #8 (CFO-friendly metric reframer). Cut every vanity metric without a P&L equivalent. 9. **Min 85–88.** Prompt #9 (MEDDIC leave-behind). Generate the one-pager. 10. **Min 88–90.** Prompt #10 (deal-room follow-up). Pre-write the follow-up email and the async recap slide.

Prompts #1, #4, and #6 are the highest-ROI single fixes — if you only have 30 minutes, run those three. Sonnet 4.5 handles all ten; Opus 4.7 is worth it on prompt #4 only, where positioning synthesis benefits from depth.

<a href="https://www.anthropic.com/claude?utm_source=aipromptshub&utm_medium=blog&utm_campaign=sales-deck-prompts-chain" style={{display:"inline-block",padding:"10px 18px",background:"#0a66ff",color:"white",borderRadius:"6px",textDecoration:"none",fontWeight:"bold",marginTop:"12px"}}> Get Claude Pro for the full chain → </a>

### Which Claude model should I use for these sales-deck prompts?

Sonnet 4.5 handles prompts 1, 2, 3, 5, 6, 7, 8, 9, 10 — structured rewrites with explicit constraints. Use Opus 4.7 on prompt #4 (Dunford positioning teardown) where competitive synthesis benefits from depth. See Anthropic models for current lineup.

### Can I use these prompts without burning a discovery call's notes through an LLM?

Yes — every prompt is structured to accept redacted or summarized inputs. For sensitive accounts, replace customer names with role descriptors ("VP Finance at top-10 US regional bank") and replace specific numbers with banded ranges. The prompts still work; the rewrites stay defensible.

### How is this different from generic deck-template tools?

Template tools give you a layout. These prompts diagnose what's actually wrong — vague claims, missing why-now, broken positioning, CFO-hostile metrics — and rewrite against named frameworks (Dunford, MEDDIC, Gong's deal-stage data). The output is the deck rebuilt, not a new layout to fill in.

### What if my discovery call was thin and I don't have the specifics for prompt #2?

Prompt #2 has a built-in escape: if specifics aren't available, the output is the cut version — a sharper open question to the buyer instead of a hedged claim. That's the right move. A thin deck with one strong question beats a thick deck with five hedged claims. Re-run prompt #2 after the next discovery call.

### How do I keep the prompts current as the buyer landscape shifts?

Update three inputs quarterly: the banned-vendor-jargon list in prompt #4, the buyer-side P&L language list in prompt #8, and the MEDDIC criteria in prompt #9 if your category's qualification spine changes (it rarely does). The structural prompts (1, 3, 5, 7, 10) age slowly.

### Are the before/after samples from real decks?

Before slides are composites of common patterns from lost-deal teardowns. After rewrites are illustrative outputs of running these prompts against Sonnet 4.5 with synthetic discovery notes; structure is representative, specific customer names and numbers are illustrative.

### Can I run these prompts on a partner-built deck I didn't write?

Yes — and that's often the highest-yield use case. Paste the deck text, run prompts 1, 4, and 6 first to diagnose whether the structural problems are fixable in surgery or whether the deck needs a rebuild from prompt #2 forward.

- Gong Labs sales research — ~1.2M recorded sales calls; deal-stage win-rate, async-review impact, demo-conversion, lost-deal analysis. - April Dunford, Obviously Awesome — competitive-alternative positioning framework, market-category selection. - April Dunford positioning teardowns — common positioning failure modes. - MEDDIC Academy — Metrics, Economic Buyer, Decision Criteria, Decision Process, Identify Pain, Champion; forecast-accuracy benchmark. - Pitch.com deck-effectiveness study, 2025 — 25,000 shared decks analyzed for view-through and drop-off. - Anthropic prompt engineering documentation — Claude prompt best practices. - Anthropic model documentation — Sonnet 4.5 / Opus 4.7 selection.

---

<script type="application/ld+json" dangerouslySetInnerHTML={{ __html: JSON.stringify({ "@context": "https://schema.org", "@type": "Article", "headline": "10 Claude prompts that fix bad sales decks in 2026", "datePublished": "2026-06-10", "dateModified": "2026-06-10", "author": { "@type": "Person", "name": "Jake Morrison", "jobTitle": "B2B sales-enablement strategist" }, "publisher": { "@type": "Organization", "name": "AIPromptsHub", "url": "https://aipromptshub.co" }, "mainEntityOfPage": "https://aipromptshub.co/blog/10-claude-prompts-fix-bad-sales-decks-2026", "description": "Ten production-grade Claude prompts that fix the ten failure modes responsible for most losing B2B sales decks — title-slide hooks, problem sharpeners, why-now scoring, Dunford positioning, demo simplification, ROI scaffolding, objection handling, CFO-friendly metrics, MEDDIC leave-behinds, and deal-room follow-ups.", "citation": [ "https://www.gong.io/resources/labs/", "https://www.aprildunford.com/obviously-awesome", "https://www.meddic.academy/what-is-meddic/", "https://pitch.com/blog", "https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/overview", "https://docs.anthropic.com/en/docs/about-claude/models" ] }) }} />

<script type="application/ld+json" dangerouslySetInnerHTML={{ __html: JSON.stringify({ "@context": "https://schema.org", "@type": "FAQPage", "mainEntity": [ { "@type": "Question", "name": "Which Claude model should I use for these sales-deck prompts?", "acceptedAnswer": { "@type": "Answer", "text": "Claude Sonnet 4.5 handles prompts 1, 2, 3, 5, 6, 7, 8, 9, and 10 — structured rewrites with explicit constraints. Use Opus 4.7 on prompt #4, the April Dunford positioning teardown, where competitive synthesis benefits from depth." } }, { "@type": "Question", "name": "Can I use these prompts without burning a discovery call's notes through an LLM?", "acceptedAnswer": { "@type": "Answer", "text": "Yes. Every prompt is structured to accept redacted or summarized inputs. For sensitive accounts, replace customer names with role descriptors and replace specific numbers with banded ranges. The prompts still work and the rewrites stay defensible." } }, { "@type": "Question", "name": "How is this different from generic deck-template tools?", "acceptedAnswer": { "@type": "Answer", "text": "Template tools give you a layout. These prompts diagnose what's actually wrong — vague claims, missing why-now, broken positioning, CFO-hostile metrics — and rewrite against named frameworks like Dunford positioning, MEDDIC qualification, and Gong deal-stage data. The output is the deck rebuilt, not a new layout to fill in." } }, { "@type": "Question", "name": "What if my discovery call was thin and I don't have the specifics for prompt #2?", "acceptedAnswer": { "@type": "Answer", "text": "Prompt #2 has a built-in escape: if specifics aren't available, the output is the cut version — a sharper open question to the buyer instead of a hedged claim. A thin deck with one strong question beats a thick deck with five hedged claims." } }, { "@type": "Question", "name": "How do I keep the prompts current as the buyer landscape shifts?", "acceptedAnswer": { "@type": "Answer", "text": "Update three inputs quarterly: the banned-vendor-jargon list in prompt #4, the buyer-side P&L language list in prompt #8, and the MEDDIC criteria in prompt #9 if your category's qualification spine changes. The structural prompts (1, 3, 5, 7, 10) age slowly." } }, { "@type": "Question", "name": "Are the before/after samples from real decks?", "acceptedAnswer": { "@type": "Answer", "text": "Before slides are composites of common patterns from lost-deal teardowns. After rewrites are illustrative outputs of running these prompts against Sonnet 4.5 with synthetic discovery notes; structure is representative, specific customer names and numbers are illustrative." } }, { "@type": "Question", "name": "Can I run these prompts on a partner-built deck I didn't write?", "acceptedAnswer": { "@type": "Answer", "text": "Yes, and that's often the highest-yield use case. Paste the deck text and run prompts 1, 4, and 6 first to diagnose whether the structural problems are fixable in surgery or whether the deck needs a rebuild from prompt #2 forward." } } ] }) }} />