A lot of projects, AI and otherwise, forget some basic design principles that can be used to craft an effort with higher likelihood of success. My five heuristics are:Prisms Separate Light Into Its Important Consituent PartsgettyPoliticsRhythmIdentityStaffMetricsSome of these are so obvious I am almost embarrassed to mention them, but I have sat through dozens if not hundreds of funding pitches that miss these simple, fundamental ideas.Politics: Make sure the cost and benefit of a new project sit under the same operational leader.At the CEO level all costs and benefits roll up under the same leader, but as one works their way down into the organization there are many departments and functions who are motivated to optimize their domain first and the rest of the organization later. As one senior partner at a consulting firm said to me, As soon as I hit my sales number, Im a team player. If that happens six months before the end of the year Im a team player for six months; if it happens a month before the end of the year, Im a team player for a month. In that spirit, I have sat through many project funding proposals in which the cost for a project sits in one department, say the IT department, and the benefit sits in another, say marketing. This is a big red flag for any project. Theres also a classic Academy of Management Journal by Steven Kerr on this topic entitled, On the Folly of Hoping for A While Rewarding B.Rhythm: Choose projects with a fast cycle-time for benefits realization.Customer service, digital marketing, trading operations, usually have fast cycle times and if your Generative AI projects helps make customer service representatives more productive or automates their tasks, it is easy to see benefit, quickly. This is one of the reason some scientists who research reproduction study May flies who are born, live and die in 24 hours. They do not investigate sea turtles who can live 100 years or more. I believe that the best projects deliver some initial value within the year, and even within six months.Identity: Tie the project to the identity of the organization.Simon Sinek has become famous by noting the Power of Why. Tying any project to the identity of the organization helps to answer why this project. For example, after Steve Sinfosky from Microsoft was snowed in for two days at his alma mater wrote the Computing at Cornell memo to Bill Gates, which started a process that transformed the firm. Gates subsequently penned his famous Internet Tidal Wave memo that tied this academic network to the core why of Microsoft. This type of linkage to identity makes the reason for doing any project more obvious. For instance, anyone working in an insurance firm today can avail themselves of the fact that insurance firms were the first AI firms in the sense that they always have been competing on the quality of their decision models and Generative AI is just the next extension of that identity. This connection is worth making.MORE FOR YOUStaff: Any investment in helping staff understand AI is not wasted.Ethan Mollick in his wonderful book C0-Intelligence makes the point that todays AI is the dumbest AI we will ever have. The machines are getting smarter every day and at a brisk pace. When making the case for a project, it is face valid that every organization and employee will want to have some facility with AI as a user, creator of chat bots, or maybe a deep expert in the coming years. This means that any investment in a project concerning AI will educate the staff on an inevitable content area. As long as the people stay with the firm, it will be a worthwhile expense to train your talent, independent of project outcome.Metrics: Dont invent new ones, tie to the existing ones!It is often tempting for individuals doing something new to create new measures for their efforts. The problem is that establishing new metrics can be as hard, or harder, than delivering the project itself. If you notice, it is very rare that organizations and functions change their metrics. Therefore if you are proposing a new AI project, it is vital that you make the linkage to existing measures explicit and simple. Moreover, if you are delivering a call center agent to the firm, dont focus on return on assets or NPV, focus on how the new technology can help drive calls per representative, length of calls, one call resolution and customer satisfaction the types of metrics that any call center executive manages to every day.Questions to Ask YourselfWhen considering AI projects remember to look at them through the PRISM (sorry about that, but I just had to say that corny idea). This simple framework can help you choose less risk projects or to modify your approach to existing ones. Ask these three questions:How does my project rate on the five PRISM dimensions?Can I make the linkage to each clear and credibly?Based on these heuristics are there projects I should modify, cancel, or accelerate?These questions are easy to ask early, and so difficult to ask once you are into the project. I hope you will ask them briskly and as you begin not as diagnostic when overseeing a post mortem!