453- Make Approved AI Easier to Use w/Mason Herbel

Mason Herbel describes the shift from an AI startup to Parallel Systems, where the first questions were about governance and company data. Some employees already used AI. Others had barely started. His advice for IT leaders: give people a clear route to approved tools, then make that route easy enough to use at work. He suggests company enterprise plans, agreements about data handling, and a trial period for tools employees request. After the trial, check actual usage and whether another paid app already does the job. Blocking a website has limits when someone can reach for a personal phone. Mason puts it plainly: “So you may as well make it easy for people to use in a secured manner.” This is his proposed approach, rather than a report of a completed company-wide AI rollout. We also get into choosing infrastructure one workload at a time. Mason recounts moving an engineering firm away from aging local servers after checking its remaining applications and Microsoft licenses. At an AI company, he argued for examining owned GPUs as API bills grew. He describes that proposal and the conversations around it; he does not claim verified savings from a completed GPU deployment. For IT leaders trying to get a useful idea heard, Mason explains how he kept discussing the GPU proposal with a principal engineer. He also recalls asking to join a server installation early in his first job and learning by doing when the lead technician had to leave. The practical thread: ask how the work gets done, bring colleagues into the decision, and keep everyday IT needs covered while taking on new projects.