Choosing between a chatbot and a workflow

“We want a chatbot.” It’s one of the most common requests we get — and often it isn’t what the client needs. The right instinct isn’t to pick a technology, but to describe the problem. The shape follows from there.

Workflow: a known path

Choose a workflow when the steps are predictable and triggered by an event: a form submitted, an order placed, a file dropped. The system runs a reliable sequence, to the millisecond, thousands of times without tiring. No conversation, no chat interface — just a correct result, every time.

Chatbot: a fuzzy front door

A conversational assistant shines when the input is unpredictable and expressed in natural language: a customer question that can take a thousand forms, a search across a knowledge base, decision support. It absorbs human ambiguity — but that flexibility comes with a reliability cost you have to manage.

• Structured, event-triggered input → workflow
• Natural-language, varied input → assistant
• Often the best answer combines both

Usually: both

In practice, the most effective systems use an assistant as the interface — it understands the request — then hand execution to deterministic workflows behind the scenes. The user converses, but the sensitive action runs through a controlled, tested path. You get the flexibility of conversation and the reliability of automation.