Last updated: August 2026
Custom AI chatbot development costs $5,000–$12,000 for a simple LLM-powered bot, $12,000–$30,000 for a support bot integrated with your helpdesk and grounded in your knowledge base, and $30,000–$75,000+ for multi-agent systems that execute complex workflows. Timeline runs from three weeks to four months, driven mostly by integrations and data readiness.
Ranges are ProCoders’ standard fixed-price scopes as of August 2026 — typical US/EU agency quotes for comparable builds run roughly double (see the published rate cards we compare against below). Freelancer and in-house comparisons included.
What’s included at each tier
Tier 1 — Simple LLM bot ($5k–$12k)
Discovery, prompt architecture, connection to a frontier or open-weight model, ingestion of your existing content (site, FAQs, PDFs), a web widget, human handoff, basic analytics, and testing against real conversation samples. What it doesn’t do: take actions in other systems or answer from data that changes hourly.
Tier 2 — Integrated support bot ($12k–$30k)
Everything above, plus a proper RAG pipeline (chunking, embeddings, retrieval tuning, freshness updates), two-way helpdesk/CRM integration (create tickets, look up orders, update records), role-based answer behavior, a second channel (usually WhatsApp or in-app), and an evaluation harness so you can measure answer quality before and after launch. This is the tier most businesses actually need.
Tier 3 — Multi-agent system ($30k–$75k+)
Multiple specialized agents coordinating: one that talks, others that verify identity, execute transactions, check policy, escalate. Adds workflow orchestration, audit trails, compliance controls (PII redaction, data residency), load handling, and serious testing. Timeline stretches with each system touched.
The five cost drivers (and which ones you control)
- Integrations — the biggest line item. Each connected system (Zendesk, Salesforce, Shopify, internal APIs…) typically adds $1,500–$4,000 depending on API quality. You control this: launch with the two integrations that matter, add the rest post-launch.
- Data readiness. RAG quality is a function of your documentation. Clean, current help articles → fast build. Scattered PDFs and tribal knowledge → add a content-cleanup phase. This is the most common source of quote variance between vendors.
- Model strategy. Frontier API models cost more per token but nothing to host; open-weight models flip that. The build cost difference is small; the running cost difference at volume is large (see monthly costs below).
- Compliance & security. SSO, PII handling, audit logs, data residency: add 15–30% at Tier 2+, unavoidable in fintech/health.
- Channels. Each additional channel (voice being the priciest — latency engineering + telephony) adds build and test time.
Agency vs freelancer vs in-house
How agencies structure these fees — fixed price vs retainer vs per-bot — is its own topic: chatbot agency pricing.
What it costs to run after launch
Build cost is half the story. Custom bots carry monthly costs of roughly:
- Hosting/infrastructure: $50–$500
- Model usage: $20–$400 — scales with volume, tunable 3–5× via prompt/model optimization
- Maintenance: expect 10–20% of build cost per year (content updates, model migrations, integration API changes), or a monthly retainer
Total: typically $150–$750/month — flat, regardless of conversation volume. That flatness is the economic argument for custom at scale (full hidden-costs breakdown).
Where the money goes in a typical Tier 2 build
Scope: support bot with helpdesk integration, RAG over your knowledge base, web + one extra channel.
Integrations and testing carry half the budget — which is why “just a chatbot” quotes that skip them come back as change orders later.
When custom beats SaaS (and when it doesn’t)
If your volume is low and your use case is standard, a $200/month SaaS plan beats every number in this article — rent it. Custom wins when per-conversation fees at your volume exceed amortized build cost (usually somewhere past 3,000–5,000 conversations/month), when the bot must act inside your systems rather than just chat, or when your data can’t live on a vendor’s infrastructure. The full decision framework with crossover math: build vs buy a chatbot. For overall market pricing context, start with the complete AI chatbot cost guide.
Get a scoped quote
Every range above narrows to a fixed number after a discovery call: we scope integrations, assess your data, and quote fixed-price with timeline. You own the code and the data.
Methodology: figures are ProCoders’ standard pricing as of August 2026; market comparisons from published agency rate benchmarks. We sell custom development — where SaaS is the honest answer above, we’ve said so.
Ahmad R.
Engineer at ProCoders. Spends most of the day shipping production AI systems for clients across SaaS, FinTech, and consumer. Writes here when something is worth a writeup.
Connect with Ahmad R. on LinkedIn →