TLDR
- Off-the-shelf chatbot widgets (Intercom Fin, Crisp, Drift, etc.) are great when your support content is in one place and your conversations are predictable.
- Custom chatbots are worth it when you need them to take actions (look up orders, schedule meetings, file tickets), integrate with internal systems, or use proprietary content the off-the-shelf tools cannot access.
- Off-the-shelf is faster to ship (days) and starts at $50-$300 a month. Custom takes 4-8 weeks to ship and starts at $15K-$40K plus $50-$300 a month in API costs.
- The biggest cost driver is not the chatbot. It is the work of preparing and maintaining the content the chatbot draws from.
- Buy a widget, deploy it, see what conversations actually happen, then decide if a custom build is worth it.
Every business with a website is asking the same question: should we add a chatbot, and if so, do we buy one or build one? Here is the practical framework.
What Off-the-Shelf Chatbots Are Good At
Modern chatbot widgets (Intercom Fin, Crisp, Tidio, Drift, ChatBot.com) are mature products. They handle the visual chat UI, the WebSocket plumbing, the conversation logging, the escalation to human agents, and the integration with help center content. You drop a script tag on your site, point them at your help docs or FAQ, and they work.
For a business with:
- A reasonably complete help center or FAQ
- Predictable, mostly informational customer questions
- An existing customer support team they want to extend
- No need to take actions inside other systems
Off-the-shelf is the right answer. You will ship in days. You will pay between $50 and $500 a month depending on the vendor and volume. The functionality will be more polished than what you would build in your first iteration.
Where Off-the-Shelf Hits the Wall
The limits show up when you need the chatbot to do anything beyond answering questions. Common scenarios that push teams toward custom:
- "Look up my order status" (requires API access to your order system)
- "Schedule a demo with our team" (requires calendar integration)
- "Refund this transaction" (requires write access to your billing system)
- "Show me my account usage" (requires authenticated access to user data)
- "Help me configure this product" (requires knowledge of internal documentation that never goes into a public help center)
Some off-the-shelf platforms support custom actions through webhooks or plugins. The integration work is real, the UX is constrained, and you usually end up paying for both the widget and a custom backend layer to power the actions.
What a Custom Chatbot Looks Like
A custom chatbot for a business website is typically:
- A small chat UI component embedded in the page
- A backend API endpoint that handles user messages
- A retrieval system that finds relevant content from your knowledge base
- Tool definitions that let the model call your internal APIs (orders, accounts, scheduling)
- A conversation log so you can review and improve over time
- An escalation path to a human agent (often via Slack or your existing helpdesk)
The advantage is full control. The chatbot can do anything your APIs allow it to do. It can use proprietary content. It can handle authenticated requests. It can be tuned to your brand voice.
The Honest Cost Comparison
Off-the-shelf:
- Setup: 1-3 days of configuration
- Monthly: $50-$500 depending on volume and vendor
- Annual: $600-$6000
- Customization: limited by what the vendor supports
Custom (typical scope: handles top 10-15 question types, looks up account data, escalates to human):
- Build: $15K-$40K depending on integrations
- Monthly LLM API costs: $50-$500 depending on volume
- Annual ongoing: $600-$6000 in API costs plus maintenance
- Customization: full control
Year one favors off-the-shelf if you do not need actions. Year three custom is often comparable or cheaper, especially if the actions reduce human support workload.
The Real Cost Driver Nobody Mentions
Whether you build or buy, the dominant cost is content preparation. The chatbot is only as good as the knowledge it has access to. If your docs are scattered, outdated, or written in a way that confuses people, the chatbot will produce confused answers.
Plan for a real content audit before launch. Many of our chatbot projects spend more time on content cleanup than on the chatbot itself. The good news is that this work pays off in human-facing channels too. Better docs are better for everyone.
The Pragmatic Recommendation
Buy first, build later. Deploy an off-the-shelf widget. Review the actual conversations after a month. Identify the questions where the widget falls short and the actions you wish it could take. If those gaps are significant, that is the spec for your custom build. You will build the right thing instead of guessing.
At Stunzer Digital, we build custom AI assistants regularly. We are also happy to recommend an off-the-shelf widget when that is genuinely the right call. The wrong move is the consulting-firm answer of "yes, you need a custom enterprise AI platform" without first proving that the simpler thing does not work.
Tags
Related service
Want this built? See how we work on AI & Automation.
