The real enterprise AI value is in agentic workflows, the chatbots are not even worthy.
After watching people spend eighteen months basically building the same generic chat interference.. I was done. You can't expect business metrics to improve when their tools only answer basic questions. While I was trying to fix this, I read about the Avenga case studies where they basically shift from passive chat assistants to active automated agents that handle the entire workflow. When one builds a dedicated sequence for clinical trial auditing or fraud detection you finally connect the model to a real operational value. It might take some tweaking to get that running smoothly though but hey at least it's generating a return on investment.
Btw, how did you guys manage to transition your projects from simple conversational interfaces to fully automated workflows? What did your engineering stack look like and did it actually do something for your team?