ChatGPT retrained users for conversational support

ChatGPT changed customer support less by inventing better bots than by retraining users to trust natural language software. Before 2022, most support bots worked like rigid phone trees or keyword matchers, so customers learned to avoid them. Once ChatGPT showed that a plain English request could usually get a useful answer, companies like Intercom could launch bots that people would actually try, and support volumes, product roadmaps, and pricing models all started moving around that new behavior.

• Intercom had already built earlier bot generations, first button driven flows, then intent matching bots that could reach roughly 50% resolution only with heavy setup. Fin was different because it let customers ask open ended questions in their own words, using existing help docs as the source of truth instead of manually programming every path.
  2 sacra 3 sacra
• The strategic effect was immediate inside SaaS companies. Intercom said ChatGPT made older roadmap items feel irrelevant, and Zapier described it as the moment teams stopped dabbling and started rewriting plans. That matters because buyer demand and employee conviction arrived at the same time, which is rare in enterprise software transitions.
  1 sacra
• Once users accepted talking to bots, the economics of support changed fast. AI agents now commonly price per resolved conversation, around $0.99 to $1.50, instead of per seat, because the buyer is paying for the answer to happen, not for a rep license to exist. That opened the door to newer players like Sierra and Decagon, while pushing incumbents to rebuild around AI and workflow depth.
  3 sacra 5 sacra

This points toward software interfaces becoming more conversational by default, but the winners will not be the chat box alone. The durable advantage will come from owning the workflow behind the conversation, the help desk, the knowledge base, the customer context, and the actions the agent can take after it understands the request.