How much does a basic AI chatbot cost?

A basic AI chatbot for customer service typically costs $5,000-$25,000 to build, with monthly running costs of $100-$500 for API fees, hosting, and monitoring. The wide range depends on complexity: a simple FAQ bot sits at the lower end, while one with CRM integration and multi-channel support sits higher.

What are the ongoing costs of running an AI system?

Ongoing costs include API usage ($10-$600/month), cloud infrastructure ($50-$500/month), monitoring tools ($20-$100/month), and maintenance (roughly 10-20% of your build cost annually). You should also budget 2-5 hours per month of internal staff time for knowledge base updates.

Why is data preparation so expensive in AI projects?

Data preparation often accounts for 30-50% of total project time because most business data is not AI-ready. Documents exist in inconsistent formats, critical knowledge is undocumented, and existing content may contain contradictions. Cleaning, structuring, and validating data is manual, detail-oriented work.

Should I choose fixed-price or time-and-materials for an AI project?

Choose fixed-price if your requirements are well-defined and unlikely to change significantly. Choose time-and-materials if you are exploring, expect requirements to evolve, or need flexibility. Many successful projects use a fixed-price discovery phase followed by time-and-materials for the build.

The Real Cost of AI: What to Budget for Your First AI Project

The cost of an AI project depends on its complexity, but realistic ranges are: a simple AI chatbot costs $5,000-$25,000, workflow automation runs $3,000-$15,000, a RAG system sits at $15,000-$50,000, and a full custom AI platform starts at $50,000 and goes up from there. Beyond the build cost, you need to budget for ongoing API fees, infrastructure, monitoring, and the often-overlooked costs of data preparation and change management.

If you have been researching AI for your business, you have probably noticed that nobody wants to give you a straight answer on cost. Consultancies dodge the question. Vendors quote wildly different numbers. Blog posts say "it depends" and leave it there.

We are going to be more direct than that. Here is an honest breakdown of what AI projects actually cost in 2026, based on our experience delivering these projects for Australian businesses.

What Are the Core Cost Components of an AI Project?

1. AI Model and API Costs

Unless you are training your own model (you probably should not be), you will use a commercial AI model via API. Here is what that looks like:

Model	Input Cost (per 1M tokens)	Output Cost (per 1M tokens)	Typical Monthly Cost
GPT-4o	~$2.50	~$10.00	$50-$500
Claude 3.5 Sonnet	~$3.00	~$15.00	$50-$600
Gemini 1.5 Pro	~$1.25	~$5.00	$30-$300
GPT-4o Mini / Claude Haiku	~$0.15-$0.25	~$0.60-$1.25	$10-$100

What does "tokens" mean in practice? A token is roughly three-quarters of a word. A typical customer service chatbot interaction uses about 1,000-3,000 tokens. If you handle 500 conversations per month, that is roughly 750,000-1,500,000 tokens. At GPT-4o rates, that is approximately $5-$20 per month in API costs.

The key insight: API costs are usually the smallest line item. The expensive part is building the system around the API.

2. Development Costs

This is where most of the budget goes. Development costs vary massively depending on whether you build custom, use a platform, or work with a consultancy.

Approach	Cost Range	Timeline	Best For
No-code platforms (Zapier, Voiceflow)	$500-$5,000	1-4 weeks	Simple chatbots, basic automation
Low-code with consultancy	$5,000-$25,000	2-8 weeks	Custom chatbots, workflow automation
Custom development	$15,000-$100,000+	4-16 weeks	RAG systems, AI platforms, complex integrations

3. Infrastructure Costs

Your AI system needs somewhere to live and something to connect to:

Cloud hosting: $50-$500/month for typical business applications (AWS, GCP, or Azure).
Vector database (for RAG systems): $20-$200/month (Pinecone, Weaviate, or Qdrant).
Monitoring and logging: $20-$100/month (LangSmith, Helicone, or custom).
Data storage: $10-$100/month depending on volume.

For most small to medium projects, expect $100-$800 per month in infrastructure costs.

4. Ongoing Costs

This is where budgets get blown. Many organisations budget for the build but not for what comes after:

Monitoring and maintenance: 10-20% of the initial build cost annually. Models drift, integrations break, edge cases surface.
Content and knowledge base updates: 2-5 hours/month of internal staff time to keep your AI's knowledge current.
Model upgrades: When a better model releases (which happens regularly), you may want to upgrade and retest. Budget 1-2 days of development per quarter.
Support and bug fixes: Things break. Plan for it.

What Does Each Type of AI Project Actually Cost?

Project Type	Build Cost	Monthly Running Cost	Timeline
AI chatbot (customer-facing)	$5,000-$25,000	$100-$500	2-6 weeks
Internal knowledge assistant (RAG)	$15,000-$50,000	$200-$800	4-10 weeks
Workflow automation suite	$3,000-$15,000	$50-$300	1-4 weeks
Document processing pipeline	$10,000-$40,000	$150-$600	3-8 weeks
Full AI platform (multi-feature)	$50,000-$200,000+	$500-$3,000+	3-6 months

What Are the Hidden Costs Most People Miss?

Data Preparation (Often 30-50% of Total Project Time)

Your data is probably not AI-ready. Documents are in inconsistent formats. Knowledge is trapped in people's heads. Policies contradict each other. Cleaning, structuring, and preparing your data for AI ingestion is frequently the most time-consuming and expensive part of the project.

Change Management

Building the AI is one thing. Getting your team to actually use it is another. Budget for:

Staff training sessions (4-8 hours across the team).
Process documentation updates.
A "champion" within your organisation who drives adoption.
A feedback loop to capture what is working and what is not.

Scope Creep

Every AI project discovers additional use cases during development. "While we are at it, could it also..." is the most expensive phrase in AI consulting. Set a firm scope upfront and log additions for Phase 2.

How Should You Think About Total Cost of Ownership?

Use this framework to estimate your total cost over the first 12 months:

Build cost: One-off development investment (use the table above).
Monthly running costs: API + infrastructure + monitoring (multiply by 12).
Data preparation: Add 30-50% of your build cost if your data needs significant cleanup.
Change management: Add $2,000-$10,000 for training and adoption support.
Maintenance reserve: Add 15% of build cost for year-one maintenance.

Example: AI chatbot for a mid-sized Australian business

Build cost: $15,000
Monthly running: $250 x 12 = $3,000
Data prep: $5,000
Change management: $3,000
Maintenance reserve: $2,250
Total Year 1: $28,250

That is a realistic number. If someone quotes you $5,000 for the same scope, they are either cutting corners or not including the full picture.

How Can You Reduce AI Project Costs?

Start with quick wins: Prove value with $500-$3,000 automations before committing to larger builds.
Use smaller models where possible: GPT-4o Mini and Claude Haiku cost 90% less than their full-sized siblings and handle many tasks perfectly well.
Prepare your data before engaging a consultant: Organise your documents, consolidate knowledge, and clean up inconsistencies. This directly reduces billable hours.
Choose the right engagement model: Fixed price gives budget certainty. Time and materials gives flexibility. Choose based on how well-defined your requirements are.
Build incrementally: Ship a minimum viable product, learn from real usage, then expand. Do not try to build everything at once.

The organisations that get the best ROI from AI are not the ones that spend the most. They are the ones that scope tightly, start small, measure results, and scale what works. If you want help scoping and budgeting your first AI project, our strategy services are designed for exactly that.