What to expect
What the AI delivers, what you finish in PowerPoint, and how to review your output
Deliverables AI gets you 90% of the way to a finished deck. You handle the last 10% — reviewing citations, adjusting a few data points, and adding any final touches in PowerPoint. Here's what that looks like in practice.
What the AI delivers
When generation completes, you'll download a fully editable PowerPoint file that includes:
- Professional narrative — A logical story arc tailored to your deliverable type (CIM, pitch deck, board deck, etc.)
- Real data — Charts, tables, and metrics pulled from your sources with proper citations
- Consistent design — Branded colors, typography, and layouts applied across every slide
- AI-generated imagery — Custom images that match your brand aesthetic and slide content
- Native PowerPoint objects — Every chart, table, text box, and image is a real PowerPoint element, not a flat image
This isn't a screenshot of a presentation — it's a working PowerPoint file. Click on a chart and a live Excel workbook opens. Select a text box and edit it directly. Move, resize, or delete any element. This native PowerPoint output is a key difference from web-native tools — see how we compare.
What you review and finish
No AI system produces a perfect, client-ready deck on the first pass. Plan to spend 15-30 minutes reviewing and polishing your output. Here's what to focus on:
1. Check citations and data
The AI cites its sources, but you should verify that key numbers are accurate and properly attributed. Look for:
- Financial figures — Confirm revenue, EBITDA, margins, and other metrics match your source documents
- Market data — Verify industry statistics and competitive claims
- Dates and names — Double-check that company names, executive names, and dates are correct
2. Look for markers
When the AI can't find specific information in your sources, it flags it rather than making something up:
| Marker | What it means | What to do |
|---|---|---|
| Illustrative | A reasonable estimate based on context, not a verified fact | Replace with actual data or remove if not needed |
| TBD | Information that needs to be filled in manually | Add the real data from your own knowledge or sources |
These markers are a feature, not a flaw — they tell you exactly where to focus your attention.
3. Refine the narrative
The AI creates a strong starting narrative, but you know your deal and your audience better:
- Adjust emphasis — Promote the points that matter most to your specific audience
- Add insider knowledge — Details from conversations, negotiations, or observations that weren't in your source documents
- Tighten language — Fine-tune wording to match your firm's voice and the specific context
4. Polish the design
The design is production-ready out of the box, but you may want to:
- Swap an AI-generated image for a specific photo
- Adjust chart formatting for your firm's conventions
- Move or resize elements to your preference
Why 90% is the right target
A fully automated deck with zero human review would be dangerous — especially for deal materials where every number matters. The 90% model gives you:
- Speed — Hours or weeks of work compressed into minutes of generation plus minutes of review
- Accuracy — Human eyes on every critical data point before it reaches a client or committee
- Control — You make the final call on narrative, emphasis, and presentation
Think of it as having a highly capable analyst who prepares the first draft. You still review, refine, and sign off — but the heavy lifting is done.
Tips for faster reviews
- Upload comprehensive sources — The more context the AI has, the fewer gaps it needs to flag
- Be specific in your prompt — Mention exact metrics, audience, and structure to reduce post-generation edits
- Use Express mode for iteration — Generate a quick draft first, review the structure, then regenerate in Comprehensive mode with a refined prompt
- Build a review checklist — For recurring deliverables, create a standard checklist of items to verify each time
The first time you use Deliverables AI, budget a bit more time for review as you learn what the system does well and where you want to add your own touch. After a few generations, most users develop a quick review routine that takes 15 minutes or less.