Lessons Learned from Assessing Trustworthy AI in Practice