My AI chatbot project derailed when it started thanking users for their patience
I was testing a support bot in my kitchen around 3pm and it kept saying that after every question, even ones like 'what's your return policy'. Turns out I fed it a dataset full of canned phrases. I rewrote the training data to use plain responses and re-ran it, but now it sounds robotic. Has anyone else dealt with a bot that mixes up polite filler with actual answers?