You can spin up a chatbot on a SaaS platform this afternoon. So why would anyone pay for a custom build? The honest answer: often you shouldn't — and knowing when you should is the difference between a good investment and an expensive detour.
We build custom chatbots for a living, so read this knowing our bias — but we'll tell you plainly when a platform is the right call, because a mismatched project helps nobody.
What platforms do well
Chatbot platforms (Intercom Fin, Tidio, Chatbase and the like) are genuinely good at the standard case: index your help centre, drop a widget on your site, answer common questions. Setup takes hours, not weeks, and pricing starts low.
If your need is "answer FAQs from our public docs on our website", a platform is probably the right first move. We say this to prospective clients regularly.
Where platforms hit their ceiling
The ceiling shows up at integrations and data. Platforms connect to the popular tools in the ways the vendor anticipated — but the moment you need the bot to check order status in your internal API, write structured records to your CRM with your rules, or reason over data behind your firewall, you're fighting the platform instead of using it.
Data control is the second ceiling: your documents, customer conversations, and embeddings live in the vendor's cloud on the vendor's terms. For regulated industries or privacy-sensitive teams, that's often a hard no.
The third is cost at scale: per-resolution and per-seat pricing looks small until volume grows, at which point platform fees can exceed what a custom build would have cost — with none of the ownership.
What custom actually buys you
A custom build means the chatbot does exactly what your workflow needs: your retrieval over your data sources, your integrations, your escalation rules, your tone, deployed in your cloud if you want it there.
It also means ownership. Code, prompts, vector stores, and infrastructure are yours — no per-conversation ransom, no feature roadmap you don't control, no vendor lock-in at renewal time.
And it means model freedom: swap OpenAI for Anthropic or an open-weight model when pricing or quality shifts, instead of waiting for a vendor to offer it.
The real cost comparison
Platforms: low entry cost, recurring per-seat/per-resolution fees that scale with usage, near-zero build time. Custom: real upfront build cost (see our pricing guide), then modest running costs you control.
The crossover point is usually volume plus integration depth. High conversation volume, deep integrations, or sensitive data push the math toward custom quickly. Low volume and standard needs keep platforms cheaper for years.
A decision framework
Choose a platform if: your questions are answerable from public docs, standard integrations cover you, and data residency isn't a concern.
Go custom if any of these are true: the bot must act on internal systems, your data can't live in a third-party SaaS, conversation volume makes per-resolution pricing painful, or the chatbot is core product rather than a widget.
And if you're unsure — start on a platform, learn what your users actually ask, and bring that data to a custom build later. Nothing about starting cheap is wasted.
Frequently asked questions
Conclusion
Platforms win on speed and entry cost; custom wins on depth, control, and unit economics at scale. Neither is universally right — the mistake is picking by marketing instead of by workflow.
Describe your use case to us and we'll tell you straight which side of the line it falls on — including when the answer is "just use a platform".