Short answer: an AI consulting engagement with a working pilot costs about US$8,000 to US$30,000; a production chatbot or assistant US$20,000 to US$80,000; a generative AI workflow US$30,000 to US$120,000; an AI agent system from US$50,000; and a custom machine learning model US$40,000 to US$200,000. Model usage fees are extra.
AI development cost by solution type
| Solution | Typical US range | Timeline |
|---|---|---|
| AI consulting and pilot | US$8,000 - 30,000 | 3-6 weeks |
| RAG chatbot or assistant | US$20,000 - 80,000 | 6-12 weeks |
| Generative AI workflow | US$30,000 - 120,000 | 2-4 months |
| AI agent system | US$50,000 - 250,000+ | 3-6 months |
| Custom ML model | US$40,000 - 200,000 | 3-6 months |
| Computer vision | US$60,000 - 300,000 | 3-9 months |
What makes AI projects expensive?
- Data readiness: cleaning, labelling and structuring data is often the largest cost.
- Evaluation: building a test set from real examples and measuring accuracy properly.
- Integration: connecting the model to CRMs, ERPs, help desks and document stores.
- Guardrails: permissions, approvals, human hand-off and audit logs, especially for agents.
- Compliance: data residency and consent under GDPR, the EU AI Act, UAE and Saudi PDPL or Singapore PDPA.
The running cost of AI
Every request to a large language model has a usage cost. For most assistants that is cents per conversation, but at scale it matters. Model the cost per task at your real volume before launch, cache repeated answers, use smaller models for simple steps and set budgets with alerts.
Where to start
Pick one frequent, measurable process such as support questions, document extraction or lead research. Prove it with a pilot on real data, compare it with the current process, and only then build for production.