During my Africa ICT Right internship I built SupportIQ, a chatbot for the organization's website. The technical build took weeks; earning the team's trust took longer — and taught me more.
The turning point was admitting the bot's limits in the design itself: every answer links its source, and anything uncertain escalates to a human with full context attached.
Three trust patterns
Ground everything: no retrieved passage, no answer. 'I don't know — here's who to ask' is a valid and valuable response.
Show receipts: source links on every answer turned skeptics into reviewers. The team started curating content because they could see exactly what the bot used.
Escalate with context: handoffs include the conversation so far. Nothing kills trust like making a user repeat themselves to a human.
Chatbots are UX projects
Model choice mattered less than empty states, error messages, and the first three suggested questions. Most users need permission to start — good prompts are onboarding.
Ship the humble bot. The confident one gets turned off.
Enjoyed this? Let's talk about it.
I write to think clearly. If this sparked an idea — a project, a disagreement, a collaboration — I want to hear it.
Get in touch