The honest range is $3,000 to $80,000. That spread is real, not a cop-out — it reflects the fact that "AI chatbot" means three different things depending on what you're building.
TL;DR
- A basic FAQ chatbot (answers questions from your docs): $3K–$15K. A lead-qualifying chatbot (asks questions, scores intent, hands off to your CRM): $15K–$40K. A full agent (qualifies, routes, books meetings, no human in the loop): $30K–$80K.
- Off-the-shelf tools like Intercom or Drift run $500–$2,000/month in SaaS fees, but they use generic qualification logic. You'll hit their limits within 6 months if your ICP is specific.
- The real cost drivers: CRM integration complexity, RAG pipeline depth, model API costs at scale.
- Don't start with the build. A $497 AI Profit Leak Audit maps your exact use case before you spend $20K on the wrong chatbot.
Why the range is so wide
"AI chatbot" covers a lot of ground. The cost difference between the three main types is roughly 10x, so it matters which one you're actually asking about.
A FAQ bot answers questions from a fixed knowledge base — your documentation, product pages, support articles. It uses RAG to search your content and return accurate answers. It can't take actions, it has no memory of the user across sessions, and it doesn't touch your CRM.
A lead-qualifying chatbot runs a conversation with a website visitor, asks qualification questions (budget, company size, urgency, use case), and routes qualified leads to your CRM or calendar. This needs more than a RAG setup — a conversation state machine, your qualification logic baked in, and usually a CRM integration to push qualified leads somewhere useful.
An AI agent does the full thing autonomously: qualify, route, book, confirm. It can look up information, update records, send emails, check calendar availability, and loop back on its own reasoning until the task is done.
The price you get from a vendor tells you which of those three they're actually building. If someone quotes under $5K for "a lead-qualifying chatbot with CRM integration," they're building option one and calling it option two.
Cost breakdown by tier
Tier 1 — FAQ / support chatbot
$3,000–$15,000 · 2–4 weeks
You get RAG set up over your existing content, a conversational interface (embedded widget or standalone), basic conversation logging, and a way to update the knowledge base without a rebuild.
What's not in this tier: CRM integration, lead scoring, memory across sessions, human handoff routing. If you need any of those, you're already in tier 2.
This works well when your support team handles the same 50 questions every week and you want to deflect volume. Or when you want to give website visitors 24/7 answers without hiring another rep. It breaks as soon as you want the chatbot to qualify who it's talking to and respond differently based on the answer.
Tier 2 — Lead-qualifying chatbot
$15,000–$40,000 · 4–8 weeks
The chatbot asks structured qualifying questions, scores responses against your ICP criteria, and routes qualified leads to your CRM, booking link, or rep queue. Unqualified visitors get a different response — a resource, a lower-tier offer, or an honest "not a fit."
This tier includes: qualification logic built to your criteria, CRM integration (Salesforce, HubSpot, or your stack), basic conversation memory, handoff logic, and reporting. Autonomous calendar booking isn't in scope — that's what pushes you into tier 3.
Good for when inbound volume is high enough that manual qualification is eating rep time. The math is simple: stop reps spending an hour on a $500 deal when there's a $50K one in the queue.
Tier 3 — Full AI agent (qualify → route → book)
$30,000–$80,000 · 6–12 weeks
No human in the loop. The agent qualifies the lead, pulls their company data from your CRM or enriches it from a third-party source, picks the right rep or meeting type, checks calendar availability, books the slot, and sends the confirmation — inside one conversation.
This is agent architecture, not chatbot architecture. Multiple tools the model can call, logic for edge cases, error recovery when an API call fails, observability so you know when something breaks. The $80K end includes multi-tenant deployment, an eval pipeline, and ongoing monitoring.
Custom build vs. off-the-shelf tools
| Off-the-shelf (Intercom, Drift, HubSpot Chat) | Custom-built chatbot | |
|---|---|---|
| Upfront cost | $0–$500 | $15K–$80K |
| Monthly cost | $500–$2,000+ | $200–$800 (API + hosting) |
| Qualification logic | Generic — their rules engine | Your ICP criteria, built in |
| CRM integration | Native for major CRMs | Depends on your stack |
| Customization ceiling | Low — capped by their features | None — you own the code |
| Year-1 total cost | $6K–$24K in SaaS fees | $15K–$80K build + $2–10K running |
The 3-year math usually flips in favour of a custom build at real volume. But the more important question is fit. Off-the-shelf tools use qualification logic designed for the median SaaS company. If your criteria are specific — and they almost always are — you'll be hacking around platform limitations within 6 months.
If your criteria fit neatly into a standard lead form, start with off-the-shelf. If you're qualifying on company size, specific tool stack, or workflow complexity, you need something built to your criteria.
The cost drivers vendors don't always mention
CRM integration complexity. A native HubSpot integration is a weekend project. A custom CRM with a non-standard API, or one requiring field mapping across 40 custom properties, can add 2–4 weeks. Get clarity on this before you sign.
Model API costs at scale. 1,000 conversations/month at GPT-4o pricing runs roughly $200–800/month in API fees. At 10,000 conversations/month, that's $2,000–8,000. Architecture choices — caching common responses, routing early-stage questions to a cheaper model — can cut this by 60–70%. Ask your vendor specifically how they plan to handle it.
Conversation memory. A stateless chatbot treats every message as a fresh conversation. A stateful one remembers context across sessions — which matters for lead qualification when you don't want to ask the same questions twice. Memory adds engineering complexity and cost, but it also improves conversion for anything beyond a basic FAQ bot.
Human handoff logic. When to route to a human, how to transfer conversation context, what to do when the handoff fails — this is more work than it sounds. If your use case needs reliable handoff, scope it explicitly before anything gets built.
Eval infrastructure. A chatbot that works in testing and breaks in production is a real failure mode, and it's more common than people expect. Any vendor skipping an eval pipeline is cutting a corner you'll pay for later. Proper eval adds 1–2 weeks and $3–8K to a build. It's not optional if you care about production reliability.
Our fixed-price chatbot engagements
We don't bill hourly. Hourly billing means slower work costs you more — wrong incentive for everyone.
For AI chatbot development, our engagements work like this:
- Scoping sprint — 1 week, $2,500. We define your qualification criteria, map conversation flows, spec the integrations, and tell you exactly what tier you need before any build starts.
- Tier 2 build — 4–6 weeks, $15K–$35K fixed. Lead-qualifying chatbot with CRM integration and qualification logic built to your criteria.
- Tier 3 agent build — 6–10 weeks, $35K–$65K fixed. Full qualify → route → book automation, with observability and a 90-day support window.
Written acceptance criteria before payment. If we miss them, you get a full refund — that's the ClearShip guarantee.
Scope it before you spend
Three questions that determine most of what your chatbot will cost:
What does "qualified" actually mean at your company? Write it in plain English — company size, role, industry, urgency signal, budget range. The more specific this is, the faster and cheaper the build. Vague criteria are where scope creep starts.
What happens after a lead is qualified? Does it book a meeting directly? Flag a rep in Slack? Push to a CRM pipeline? The answer changes the integration requirements significantly.
What's the failure mode? What should the chatbot do with someone it can't qualify? The unqualified flow matters as much as the qualified one, and it's where most off-the-shelf tools completely fall apart.
If you can answer all three, you're ready for a real quote. If you can't, any quote you get will be wrong and you'll find out about it 6 weeks into the build.
If you're not sure what you need
The most expensive chatbot project is the one that gets scoped wrong and rebuilt. It happens when "we should have a chatbot" goes straight to "sign the contract" without anyone defining what qualification means, what the integration looks like, or what success actually is.
The AI Profit Leak Audit is a 30-page assessment ($497, 7-day delivery) that maps your operations, identifies where a chatbot or agent generates the most ROI, and tells you what tier you need before you spend $20K figuring it out yourself.
Most clients come in thinking they need a full agent. About half find a well-built tier-2 chatbot does the job. About a quarter find their highest-ROI AI play is somewhere else entirely. Better to know that before the build starts, not six weeks in.
How much does a basic AI chatbot cost to build?+
A basic FAQ chatbot that answers questions from your documentation using RAG runs $3,000–$15,000, with a 2–4 week build time. CRM integration, lead scoring, and conversation memory across sessions aren't included at this tier. If you need any of those, you're in tier 2 pricing ($15K–$40K).
What's the difference in cost between a chatbot and an AI agent?+
A chatbot generates responses. An agent takes actions — booking meetings, updating CRM records, routing tickets. The architectural difference is real and so is the cost: a production chatbot runs $10K–$40K, a production agent runs $30K–$80K. Most businesses should start with a chatbot unless they have a specific multi-step workflow to automate end to end.
Is off-the-shelf cheaper than a custom AI chatbot?+
In year one, yes — Intercom or HubSpot Chat cost $0 upfront and $500–$2,000/month. A custom build costs $15K–$40K upfront but runs $200–800/month after that. The economics usually flip around year two. More importantly, off-the-shelf tools use generic qualification logic. If your ICP is specific, you'll hit their limitations fast.
What makes AI chatbot development more expensive?+
CRM integration complexity is the biggest one — each external system adds 1–3 weeks. Model API costs at scale come second (GPT-4o isn't cheap at 10K conversations/month). Conversation memory, human handoff logic, and eval infrastructure all add scope. Vendors who skip eval to hit a lower price point are passing the cost to you in production failures.
How long does it take to build an AI chatbot?+
A tier-1 FAQ bot is 2–4 weeks. A tier-2 lead-qualifying chatbot is 4–8 weeks. A full AI agent that qualifies, routes, and books is 6–12 weeks. These assume scope is defined before build starts. Vague requirements add weeks — which is exactly why the scoping sprint exists.