What is the most critical challenge(s) faced in organization of your AI delivery team?

There are a few use cases which are 'low hanging fruits', eg chatbots. However, for many other use cases, the 'success' of Gen AI projects depends on
1) expectations of the outcomes - many projects require refinement and perhaps the project may be less than successful. A tolerance for the time and effort required is important. 
2) quality of the information  - the better the quality of the information, the better the outcome. Clean data is a consistent challenge that organizations face.

Internal AI/GANAI Governance ,  which is being wrapped around the use of AI and GENAI  in a regulated industry, where the rules are only just emerging or taking shape. Slowing adoption in a risk averse industry (Life sciences) . Having to leverage internal AI/Genai rather than public which is running at least one release version ahead.

From the user/customer, facilitating the definition of expectations and requirements for AI solutions. For example, in the legal sector there is currently a lot of discussion around the rates at which GenAI powered legaltech solutions hallucinating and generating inaccurate responses. However, there does not seem to be a clearly defined set of expectations from these solutions that, once met, will lead to adoption of these solutions at scale among the legal community. And that lack of defined expectations, in turn, means there also needs to be more thought in defining how the technology will be used and the implications on people and processes.

From the internal standpoint in selling/delivering AI solutions to users/customers, I think we need to find a robust way to quantitatively model the ROI of adopting new AI solutions tailored to each customer. This is an extremely difficult challenge, particularly at a PaaS level (e.g. marketing a platform to fine-tune a generative AI model for a specific domain) where the benefits can be more indirect, harder to attribute and measure (almost impossible to do in advance). The silver-lining is that, as more organizations adopt these solutions and we 'see them play-out', we will have more case studies with quantified value benchmarks to estimate key metrics in tech sales like TCO comparisons and expected ROI.

Aligning expectations to reality 
Clearly communicating the data and privacy concerns around the use of AI tools

The most critical challenge faced in the organization of our AI delivery team is bridging the gap between the overinflated expectations that most users have for AI and their limited knowledge about the technology. 
AI delivery team possesses a deep understanding of AI capabilities and limitations. Therefore, I believe it is important when forming an AI delivery team to ensure that the team understands these users' profiles and implements measures such as educating the users about AI.