Token Usage
Token Usage shows how much AI processing your workspace has consumed, broken down by time period, feature, and individual user. Use this page to monitor your usage, understand costs, and plan your subscription accordingly.
What Is a Token?
Tokens are the unit Saifa AI uses to measure AI processing. Every time AI reads a message, generates a reply, or processes a request, it consumes tokens — both for the input it receives and the output it produces.
Token basics (English):
1 word
~1.3 tokens
1 sentence
~20 tokens
1 paragraph
~100 tokens
1 page of text
~700 tokens
1 image
~1,200 tokens
📌 1 token ≈ 4 characters ≈ 0.75 words in English. Thai text may use more tokens per word due to character encoding differences.
AI Model Tiers
Saifa AI processes requests across three model tiers, each with different capability and token cost levels. Output tokens cost 4–6× more per token than input tokens across every tier.
Lite
Simple greetings, short FAQs, fast replies
29 tokens
Quick
Standard customer support, product questions
100 tokens
Think
Complex analysis, long conversations, detailed answers
350 tokens
Cost per unit — Input side:
1 word
0.03
0.13
0.67
1 sentence
0.5
2
10
1 paragraph
2.5
10
50
1 page
17.5
70
350
1 image
30
120
600
💡 Most customer service conversations (greetings, order status, FAQs) use the Lite or Quick tier. Complex multi-turn conversations or AI analysis features use Think.
🔢 How Token Usage Is Calculated
Every time Saifa AI processes a message, it uses tokens — small units of text that represent both the input and the output of a conversation.
What counts as tokens?
Each AI request includes:
Input tokens 📥 Everything sent to the AI, including:
Customer message
Previous conversation (chat history)
System instructions
Connected knowledge or data
Output tokens 📤 The response generated by the AI
💸 Why Token Usage Can Increase
Token usage may vary from message to message. Here are the main reasons:
1. Longer conversations = more context
As conversations continue, more previous messages are included to maintain context. 👉 This increases input tokens, even if the latest message is short.
2. More detailed responses = more output tokens
If the AI provides longer or more detailed answers, it uses more tokens. 👉 Output tokens are more expensive because they require more processing.
3. Language differences
Some languages (like Thai) use more tokens per sentence than English. 👉 The same message may cost more depending on the language used.
4. Complex requests use more AI processing
More advanced tasks (e.g. explanations, recommendations, or analysis) require:
More context
Longer responses
👉 This results in higher total token usage.
⚙️ How Saifa Calculates Cost
Saifa uses a weighted model:
Input tokens = standard cost
Output tokens = higher cost (typically 4–6× more)
So total usage is roughly:
Top-Level Metrics
Three summary cards at the top of the page show your workspace's current token consumption at a glance.
Total Tokens Used
The cumulative number of tokens your workspace has consumed since the start of the current billing period, shown alongside the percentage of your plan limit used.
Example: 10,617 tokens used = 10.6% of limit
Avg Daily Tokens
The average number of tokens consumed per day over the last 7 days, with a percentage change compared to the previous 7-day period.
Use this to spot unusual spikes — a sudden increase may indicate a new integration going live, a high-volume campaign, or unexpected AI usage.
Transaction of Token
The number of individual AI API calls made in the last 30 days, with a percentage change compared to the previous 30-day period.
Each transaction represents one complete AI request-and-response cycle — for example, one AI auto-reply sent to a customer, or one Saifa Assistant query.
Token Usage Limit
A progress bar below the summary cards shows your current usage against your plan's total token limit.
When you approach your limit, Saifa AI will notify you. If you exceed your limit, AI features may pause until your next billing cycle or until you upgrade your plan.
📌 To increase your token limit, go to Settings → Subscription Plan.
Usage Trends Tab
Path: Token Usage → Usage Trends
The Usage Trends tab shows token consumption and costs over time with two charts.
Monthly Token Usage
An area chart showing total tokens consumed per month over the last 7 months. Hover over any month to see the exact token count.
Use this to identify which months had high AI activity and correlate spikes with campaigns, product launches, or team growth.
Toggle between:
Token Usage — volume of tokens consumed
Compare — overlay with the previous equivalent period
Monthly Costs
A bar chart showing API cost in USD per month over the last 7 months. Hover over any bar to see the exact cost for that month.
Use this alongside Monthly Token Usage to understand your cost efficiency — if token usage increased but cost stayed flat, your conversations are using lighter model tiers effectively.
Daily Usage (Last 7 Days)
A line chart showing token consumption by day of the week for the past 7 days. Use this to identify your busiest days and ensure you have adequate AI capacity during peak periods.
Mode Breakdown Tab
Path: Token Usage → Mode Breakdown
The Mode Breakdown tab shows how tokens are distributed across the different Saifa AI features.
Usage by Mode
A donut chart and detailed breakdown table showing token consumption split by feature:
General Chat
Saifa Assistant conversations — writing, planning, analysis
Thunderstorm
Instant content generation feature
Customer Support Reply
AI auto-replies sent to customers in the inbox
Customer Support Suggestion
AI reply suggestions shown to agents (AI Assist ✨)
Understanding your mode distribution helps you identify which features consume the most tokens and whether usage matches your expectations.
User Request Table
Below the donut chart, a table shows token usage broken down by individual workspace member:
User
Team member name
Total Request
Number of AI requests made by this user
Last Request
Date and time of their most recent AI interaction
Use this to understand which team members are actively using AI features and whether usage is distributed appropriately across your team.
Recent Activity Tab
Path: Token Usage → Recent Activity
The Recent Activity tab shows a log of individual API calls — every AI request made in your workspace, listed in reverse chronological order.
Recent API Requests Table
Mode
The feature and sub-mode used (e.g. chat / general_chat)
Tokens
Total tokens consumed by this request (input + output)
Cost
USD cost of this individual API call
Datetime
Exact date and time the request was made
Status
Success (completed) or error status
Use this log to audit specific requests, investigate unexpected token spikes, or verify that all calls completed successfully.
You can control how many entries appear per page using the Show dropdown, and paginate through the full history using the arrows at the bottom right.
Tips for Managing Token Usage
Check Usage Trends monthly — compare your current month to the previous one to catch unexpected growth early, before you hit your limit.
Use Mode Breakdown to optimize costs — if Customer Support Reply is consuming most of your tokens, review whether your Knowledge Base can be made more concise so AI gives shorter, more efficient answers.
Monitor Daily Usage during campaigns — if you're running a promotion or product launch, check daily usage that week to ensure you won't hit your limit mid-campaign.
Review User Request to identify heavy users — if one team member accounts for a disproportionate share of General Chat tokens, check whether they're using Saifa Assistant efficiently or running unnecessary queries.
Upgrade before you hit the limit — don't wait for AI to pause. When your usage bar reaches 80%, go to Settings → Subscription Plan to review your options.
Last updated