4. Improve your data over time. Don’t wait until you have the data just right to get started or you never will. “When it comes to AI projects, quality data is a myth. You won’t know what data you need and the form that you need it in until you know how you are going to use it—and vice versa. Instead, work with the data that you can get hold of rapidly, drive the value you can quickly and use the success to advocate for the next round of investment in your data assets and pipelines,” the report said.5. Improve AI capabilities over time. Just like with data, most successful DSMLAI projects start small and build on successes to scale. “That often means buying horizontal or vertical point solutions with embedded AI capabilities first and then going beyond the capabilities of these solutions using custom models and applications,” the report said.6. Worry about human bias first, then AI. Because AI is a tool developed by humans it will likely contain built-in biases. The best way to avoid bias is to carefully screen the data you use to train your AI models. “Above all, test multiple hypotheses, validate models, and monitor them over time for bias and, when applicable, fairness. If you do, your resulting models will almost certainly be less biased than human decisions. If you don’t, you risk reinforcing and proliferating bias,” the report said.

7. Do not let AI projects linger. Because they are poorly understood, implemented or abandoned by their executive sponsor, AI projects are subject to relegation. The best way to avoid this outcome is to kill them off sooner, rather than later. “Empower your teams to kill projects but capture the learnings and resurrect them in new, more viable incarnations,” the report said.