Professional Services9 min read

AI Slide Generation for Consultants: Claude Frameworks That Actually Work

Turn raw research into McKinsey-style slides with Claude. Pyramid principle, SCR storyline, and chart-pick prompts. With a free template you can use today.

Haojun See

Founder & Director, On The Ground

Updated 1 May 2026

The structure that wins

Three frameworks dominate consulting decks. Pyramid principle (Barbara Minto). Top-down: state the answer, then the supporting arguments, then the evidence. Each level is MECE and answers the question of the level above. SCR storyline (Situation–Complication–Resolution). A narrative arc that takes the audience from current state → tension → recommended action. Action titles. Every slide title is a complete sentence stating the slide's takeaway, not a label. Compare *"Revenue trends"* (label) vs. *"Revenue grew 18% YoY, driven mainly by enterprise contracts"* (action title). Claude is excellent at all three when given clear prompts. The combined effect: a deck that's structured before any slide is rendered.

The pyramid prompt

*"You are structuring a partner-quality consulting deck. Topic: [TOPIC] Audience: [AUDIENCE — e.g. 'CEO and management committee of a SG mid-cap retailer'] Key recommendation: [YOUR RECOMMENDATION — one sentence] Evidence available: [PASTE NOTES, DATA SUMMARY, OR ATTACHMENTS] Build a pyramid: 1. The answer (your recommendation, restated as the apex) 2. 3 supporting arguments (the 'reasons why' that prove the answer) 3. 2–3 sub-points per supporting argument (the evidence that proves each reason) Output as a nested list. Make every node a complete sentence. Make the supporting arguments MECE. Use Singapore commercial English."* Output: a clear pyramid you can convert into a deck spine. ~20 minutes of work into 3 minutes.

The slide-by-slide prompt

Once you have the pyramid: *"For each node in the pyramid below, write one slide: - Action title (full sentence, ~12 words, states the takeaway) - 3 supporting bullets (each a complete sentence, MECE, parallel structure) - Suggested chart or visual (chart type + variables to plot, or a description of a non-chart visual) - Source citation (placeholder if not yet sourced) Format each slide as: [SLIDE N]\nTitle:\nBullets:\nVisual:\nSource:\n Pyramid: [PASTE]"* You now have a slide-by-slide spec ready to render. Most consulting teams stop here and let an associate render in PowerPoint. Or use a script that renders to .pptx programmatically.

The SCR storyline prompt

For decks that need narrative tension (board updates, change-management cases): *"Build a Situation–Complication–Resolution storyline for this deck. Situation: where the audience is now (3 sentences) Complication: what's changed or what tension exists (3 sentences) Resolution: what we recommend and why (3 sentences, ending with a clear ask) Then list the 8–12 slides in order, mapping each to one element of S/C/R. Output as a numbered list with action titles."* This produces decks that flow as a story rather than a slide-pile.

Chart-picking prompt

Claude is opinionated and good at chart selection. Use: *"Below is a data set. For each insight or claim I want to make, recommend the best chart type and the specific variables to plot. Insights: 1. [INSIGHT] 2. [INSIGHT] 3. [INSIGHT] Data: [PASTE] For each insight: chart type, x and y variables, any series breakdown, suggested title in action-title format, and one-sentence rationale for why this chart over alternatives."* Output: a chart-by-chart plan. Reduce the cognitive load of chart selection by 90%.

Pitfalls to watch

Generic recommendations. Claude defaults to generic advice if you don't give it real context. Always paste your actual notes, your client context, and your evidence — not just a topic. Over-MECE. Sometimes Claude will create six supporting arguments when three would be sharper. Force discipline: *"Reduce to three supporting arguments. Combine or drop the weakest two."* Verbose action titles. A 25-word action title isn't an action title. Force: *"Trim every action title to ≤14 words. If it can't be trimmed, the underlying idea isn't sharp enough — restate."* Singaporean English drift. Claude's default register can read American. Add: *"Singapore commercial English. No 'utilize'. No 'leverage' as a verb. No 'going forward'. Reuse 'we' or 'the firm' rather than 'our team'."*

Where to take this next

For broader strategic-advisory frameworks, see Claude for Strategic Advisory: Frameworks for Useful AI Output — coming soon. For CFO-specific deck patterns, see Claude for CFOs: From Raw Ledger to Board Pack. For audit-committee report writing, see Claude for Auditors: Workpaper Review Prompts. OTG can build a custom slide-generation pipeline for your firm — we've done it for SG-based consulting boutiques. Book a free 30-minute call. For prompt patterns across functions, see The Singapore Prompt Library.

Frequently asked questions

Can Claude actually create slides?

Claude generates slide CONTENT (titles, bullets, chart specifications, storyline). Rendering into PowerPoint or Google Slides is done programmatically (via python-pptx, Google Slides API, or a tool like Marp). The structured content is the hard part; Claude does that very well.

Will the slides look McKinsey-quality?

The structure (pyramid principle, action titles, MECE bullets, single-message-per-slide) — yes, with the right prompt. The visual polish (typography, alignment, brand templates) — that's a separate template-application step you do once and reuse.

Is this PDPA-safe for client work?

Use Claude Pro/Team (no training on input). Redact specific client names and confidential figures before prompting if the deck is for a different audience. Standard professional-services confidentiality applies.

Does it work for non-strategy consulting (audit, accounting, recruitment)?

Yes — the same frameworks (pyramid, action titles, SCR) translate to any analytical deliverable: audit committee reports, management letters, recruitment shortlists, advisory memos. The prompts adapt.

How long does a deck take with Claude?

A 12–15 slide partner-quality deck: 3–4 hours with Claude (vs. 8–12 hours from scratch). The big saving is in the structuring and writing; visual polish remains a human task.

PreviousClaude for Auditors: Workpaper Review Prompts (ACRA-Aligned)Next Executive Search with Claude: PDPA-Safe CV Parsing at Scale

Want to Apply This to Your Business?

We're a Singapore AI development and automation agency. Let's discuss how we can help solve your specific challenges.

Get in Touch View our services