Pick what you're building, roughly how complex it is, and what else it needs. This gives a realistic range based on what similar engagements have actually cost, not a marketing lowball.
An estimate is only as good as the assumptions behind it. Our blog covers how long an AI agent build realistically takes, and why cost estimates for agents so often run low.
Already scoped the project and want to model ongoing inference cost? Estimate per-request and per-day token spend across Claude, GPT, and Gemini.
Try the token cost calculator →The estimator starts from a base range for the type of project, multiplies it by a complexity factor, then adds a percentage for each extra requirement you tick. These are planning ranges set by GreyScript AI for early budgeting. They are not a market survey, an industry average or a quote.
| Project type | Base cost (USD) | Base timeline (weeks) |
|---|---|---|
| Workflow automation | $6k to $20k | 2 to 5 |
| Chatbot or copilot | $8k to $25k | 3 to 6 |
| AI integration into an existing product | $10k to $30k | 3 to 7 |
| MCP or agent tooling | $15k to $40k | 4 to 8 |
| AI agent system | $20k to $65k | 5 to 10 |
| AI-native mobile app | $25k to $80k | 6 to 14 |
Complexity: simple is 0.75 times the base range, moderate is 1 times and complex is 1.5 times. Add-ons are added together as percentages of the base: custom model fine-tuning +25%, RAG or vector search +20%, on-device inference +15%, enterprise compliance needs +15%, ongoing MLOps and monitoring +10%.
Worked example: a complex AI agent system with RAG and compliance needs. The base is $20k to $65k, times 1.5 for complexity, times 1.35 for the two add-ons (1 + 0.20 + 0.15). The estimator shows about $41k to $132k over 8 to 15 weeks.
The ranges cover the build only. Running costs such as model usage depend on traffic and the model you pick; the token cost calculator estimates those.
It depends mainly on what you are building. In this estimator's planning ranges, a workflow automation starts around $6,000 to $20,000, a chatbot or copilot $8,000 to $25,000, an AI agent system $20,000 to $65,000 and an AI-native mobile app built from scratch $25,000 to $80,000, before complexity and add-ons. These are rough planning ranges set by GreyScript AI, not market averages and not a quote.
Complexity and scope. In this estimator, a complex project costs 1.5 times the moderate range and a simple one 0.75 times. Each add-on raises the total: custom model fine-tuning by 25 percent, RAG or vector search by 20, on-device inference by 15, enterprise compliance needs by 15 and ongoing MLOps and monitoring by 10. Real projects also vary with data quality, the number of integrations and how much testing the use case needs.
No. The ranges are for building the project. Per-request model usage is a separate running cost that depends on traffic, prompt length and the model you choose; the multi-model token cost calculator on this site estimates it.
No. It is a planning range for early budgeting. A real quote needs a look at your data, your integrations and your requirements, which is what a scoping conversation is for.