Calculator Inputs
Enter deployment, model, staffing, and infrastructure assumptions. The form uses a three-column layout on large screens, two columns on smaller screens, and one column on mobile.
Example Data Table
These sample scenarios show how costs can shift with volume, automation depth, and human support requirements.
| Scenario | Monthly Conversations | Avg. Messages | Avg. Tokens per Message | Handoff Rate | Monthly Estimate |
|---|---|---|---|---|---|
| Starter FAQ Bot | 2,500 | 8 | 650 | 1.5% | $612.00 |
| Support Assistant | 12,000 | 10 | 790 | 4.5% | $2,967.11 |
| Enterprise Operations Bot | 50,000 | 12 | 980 | 6.0% | $11,846.40 |
Formula Used
Monthly Messages = Monthly Conversations × Average Messages per Conversation
Input Token Cost = (Monthly Messages × Average Input Tokens per Message ÷ 1,000,000) × Input Cost per 1M Tokens
Output Token Cost = (Monthly Messages × Average Output Tokens per Message ÷ 1,000,000) × Output Cost per 1M Tokens
Handoff Cost = Monthly Conversations × Handoff Rate × Handoff Cost per Case
Amortized Setup = One-Time Setup Cost ÷ Setup Amortization Months
Base Monthly Cost = Token Costs + Retrieval + Tools + Handoffs + Hosting + Monitoring + Training + Amortized Setup
Total Monthly Cost = Base Monthly Cost + (Base Monthly Cost × Contingency %)
This structure separates variable costs from fixed costs so you can understand scaling behavior. Variable costs grow with usage, while hosting, monitoring, training, and amortized setup stay more stable unless your architecture changes.
How to Use This Calculator
- Enter your expected monthly conversations and the average number of messages in each conversation.
- Add average input and output tokens per message based on your prompt design and model behavior.
- Enter provider prices for input tokens, output tokens, retrieval, and any tool or API calls.
- Estimate the share of conversations that need a human agent and the average cost of each handoff.
- Include hosting, monitoring, training, and one-time setup expenses.
- Choose amortization months to spread implementation cost over time.
- Add a contingency percentage for unexpected usage spikes, retries, vendor changes, or support overhead.
- Submit the form to see monthly totals, annual totals, cost efficiency metrics, and the Plotly forecast graph.
FAQs
1. What does this calculator estimate?
It estimates monthly and annual chatbot operating cost using conversation volume, token usage, retrieval expense, tool calls, human handoffs, hosting, monitoring, training, setup amortization, and contingency assumptions.
2. Which costs are variable in this model?
Input tokens, output tokens, retrieval, tool usage, and human handoffs are variable. These costs rise when conversation volume increases or when each conversation becomes longer and more complex.
3. Why is setup cost amortized?
Amortization spreads implementation expense over several months. This gives a more realistic operating view instead of making the first month appear unusually expensive compared with later months.
4. Should I estimate by message or by conversation?
Use both. Conversations help planning at a business level, while messages help estimate token consumption. Combining them improves budgeting accuracy and highlights whether short chats or long sessions drive cost.
5. How accurate is the forecast graph?
The graph is only as accurate as your assumptions. It uses your growth percentage and current fixed-cost profile. Refresh it whenever pricing, handoff rate, prompt length, or expected traffic changes.
6. Does this include human support escalation?
Yes. The handoff section estimates how many conversations need an agent and multiplies them by the average cost per case, making support escalation visible in the total estimate.
7. Can I use this for self-hosted or open-source deployments?
Yes. Replace token prices with your effective compute cost and enter infrastructure, monitoring, and support values that match your environment. The calculator still works as a planning framework.
8. What contingency percentage is reasonable?
Many teams start between 5% and 15%. Higher contingency is useful for new deployments, uncertain traffic, unstable prompt flows, or projects likely to add tools, languages, or support channels.