What are the three biggest barriers you're facing when it comes to adopting GenAI applications and services?

There are various attention points in reality:

1) Security Posture of your assets and IP
2) Shadow costs especially for the running ones where AI services are rolled out via API;
3) Training for staff and executive
4) Brand reputation risks
5) Change Management for the usage of these services

I for one will be glad when the hype bubble around genAI settles to realistic levels. I prefer to think of GenAI as a personaly productivity tool, a good knowledge elicitation tool in the hands of a good user, and a next gen chatbot (with all of the limitations on such services). I am not a luddite on the matter and am carefully tracking the research and development in the area, but when I think of the big word "Enterprise Applications" and services, I hesitated in answering as I was searching for reliability, explainability, logic and reasoning capability. I have never believed in free lunches, especially not in the IT field, and at the end of the day someone has to code the business logic. There is just no way around reasoning and these models cannot reason. 
Personal productivity is great, but with the current tech I estimate it gives our organisation a 10% bump in productivity at best - and is helping with clarity of writing, but at the expense of a lukewarm muddle of emails, proposals, briefings, papers and documentation that all sound the same.
It's my view and it doesn't stop our innovation and testing, but I would love to learn of a performing and sustainable genAI application at enterprise scale. I haven't yet come across it.