AI automation
AI Chatbot and Agent Development Cost: A Complete Guide
What AI chatbots and agents cost to build and run, the factors that move the price, and the use cases that pay back fastest.
By Intelvue Engineering · Updated · 2 min read
The short answer: the cost of an AI chatbot or agent is driven far more by what it has to do than by the AI model. A bot that answers questions from your existing content is a modest project. An agent that verifies customers, reads and writes to several systems and needs approval workflows costs more, because integrations, guardrails and testing are real engineering work. Running costs are usually small next to the value of each automated conversation.
What are the types of AI chatbots and agents?
| Type | What it does | Relative effort |
|---|---|---|
| FAQ chatbot | Answers from your website and help content | Low |
| Lead capture bot | Qualifies visitors and books meetings | Low to medium |
| Support chatbot with hand-off | Answers, then passes complex cases to your helpdesk | Medium |
| Transactional agent | Takes actions: returns, bookings, account changes | Medium to high |
| Multi-system operations agent | Works across ERP, CRM, email and documents | High |
What drives the build cost?
- Integrations. Each system the bot talks to, such as Shopify, NetSuite, Zendesk or a custom API, adds design, development and testing.
- Actions and risk. Read-only answers are simpler than actions that change data. Actions need permissions, approval steps and audit logs.
- Channels. Website chat alone is simpler than chat plus WhatsApp, email and voice.
- Content readiness. Clean, current help content speeds things up; scattered or outdated content means a cleanup project first.
- Languages. More languages mean more testing.
- Evaluation. Building a test set from real conversations takes effort but is what makes the bot trustworthy.
What does it cost to run?
- Model usage: charged per amount of text processed. For typical business conversations this is a small cost per conversation, and choosing the right model size keeps it low.
- Hosting and monitoring: infrastructure, logging and alerting.
- Maintenance: updating content, refining instructions and reviewing conversations. Plan for a few hours a month at minimum.
How do you estimate the return?
A simple model:
- Count monthly conversations in the target use case.
- Estimate the share the bot can resolve, based on a pilot or a sample of past tickets.
- Multiply by your current cost per human-handled conversation.
- Subtract running costs.
Add the value that is harder to price: instant answers at any hour, fewer abandoned carts or sign-ups, and staff spending time on higher-value work.
Which use cases pay back fastest?
- Order status and delivery questions for eCommerce
- Appointment booking and rescheduling
- Lead qualification on high-traffic websites
- Internal IT and HR questions for larger teams
How can you keep costs under control?
Start with one use case and one channel. Use your existing content. Choose the smallest model that meets your accuracy target, and only reach for larger ones where they make a measurable difference. Build the evaluation set early, so every change is tested rather than guessed.
Want a realistic estimate for your use case? Our AI chatbot and AI agent teams can scope it with you in a short call.
Frequently asked questions
- Why do AI chatbot quotes vary so much?
- Because the work varies. A bot that answers from a few web pages is a small project; an agent that takes actions in several systems needs integrations, guardrails and testing, which is where most of the effort goes.
- What are the ongoing costs?
- Model usage per conversation, hosting, monitoring and regular tuning of content and prompts. Model usage is often the smallest of these for business use cases.
- How can we reduce cost?
- Start with one high-volume use case, reuse existing help content, pick the smallest model that meets your accuracy bar, and expand only after measuring results.