Real-World Token Usage Examples
Check out much token is needed for different real-world teams
Guide: Real-world token usage examples
Token usage in Astell is primarily driven by Processed Data ingestion (documents and media). Native Data (messages, tasks, emails, code) is not where teams usually burn through tokens. The best way to forecast usage is to convert what you ingest into pages/slides, images, and minutes—then do the math.
The token math you actually need
Use these rates as your baseline:
- Documents: 5 tokens/page (or 5/slide, 5/sheet)
- Images: 1 token/image
- Audio: 2 tokens/minute
- Video: 8 tokens/minute
- Meetings: 8 tokens/minute
- Web pages: 5 tokens/page
How to estimate your monthly usage in 3 steps
Step 1 — List what you're ingesting (Processed Data only)
Most teams' biggest line items are:
- Slide decks
- PDFs
- Images
- Recordings (video/meetings/audio)
Step 2 — Convert to units
Count:
- Slides/pages/sheets
- Images
- Minutes
Don't overthink precision—you want a directional budget.
Step 3 — Compare against your org's monthly pool
On Tree and above, tokens are pooled at the organization level (tokens/seat × seats). If usage goes beyond the included pool, it becomes pay-as-you-go and is billed on your next billing date at your plan's rate (Tree: $3/1,000; Grove: $2/1,000; Forest: custom).
Example 1: Development team (25 developers)
Most engineering activity is Native Data (free). Token usage usually comes from diagrams, specs, and short demos.
What they ingest (Processed Data)
- 50 architecture diagrams (PNG)
- 10 technical specs averaging 8 pages
- 5 demo videos averaging 3 minutes
Token estimate
- Images: 50 × 1 = 50 tokens
- Docs: (10 × 8 pages) × 5 = 80 × 5 = 400 tokens
- Video: (5 × 3 min) × 8 = 15 × 8 = 120 tokens
Estimated total: ~570 tokens/month
Engineering teams spike tokens when they backfill old specs, process long recordings, or ingest large design libraries.
Example 2: Marketing agency (20 people)
Marketing is document-heavy. Decks dominate, then you get a tail from images and short videos.
What they ingest (Processed Data)
- 100 slide decks averaging 12 slides
- 200 images
- 50 PDFs averaging 6 pages
- 20 videos averaging 2 minutes
Token estimate
- Decks: (100 × 12 slides) × 5 = 1,200 × 5 = 6,000 tokens
- Images: 200 × 1 = 200 tokens
- PDFs: (50 × 6 pages) × 5 = 300 × 5 = 1,500 tokens
- Video: (20 × 2 min) × 8 = 40 × 8 = 320 tokens
Estimated total: ~8,020 tokens/month
Month one is usually the peak. If you ingest an archive of old decks and reports, usage can jump 10×. Plan your token limit accordingly.
Example 3: Sales team (10 reps)
Sales token usage is driven by collateral (decks, PDFs, images). The rest is mostly Native Data.
What they ingest (Processed Data)
- 30 decks averaging 15 slides
- 20 PDFs averaging 4 pages
- 50 images
Token estimate
- Decks: (30 × 15 slides) × 5 = 450 × 5 = 2,250 tokens
- PDFs: (20 × 4 pages) × 5 = 80 × 5 = 400 tokens
- Images: 50 × 1 = 50 tokens
Estimated total: ~2,700 tokens/month
Costs are predictable unless reps dump large historical enablement libraries into the workspace.
How this changes with pooled tokens and pay-as-you-go
If you're on Tree or Grove, think in terms of your org pool, not per-user budgets. Example: a Tree org with 10 users has 100,000 included tokens/month.
If onboarding involves ingesting a large archive, you may exceed the included pool. At that point:
- If your token limit is low, ingestion may pause.
- If your token limit allows overage, ingestion continues and you'll see pay-as-you-go charges on your next billing date.
Your token limit is the spend-control lever.
Common pitfalls that make estimates wrong
- Underestimating slide decks (they add up fast).
- Treating onboarding like steady state (onboarding is usually the peak month).
- Forgetting video/meeting minutes are priced per minute.
Quick rule of thumb
- Mostly messages/tasks with a few docs → you'll rarely notice tokens.
- Lots of decks and PDFs weekly → tokens matter but are predictable.
- Years of archives or lots of recordings → plan for pay-as-you-go and set a token limit you're comfortable with.
Related Articles
Continue learning with these related help articles