For this week, I’d like to tell a little story. It’s a story about “Jeff” (not his real name) who’s a software engineering manager at a well-known Automotive Tier 1 supplier. Because of all the AI-hype, Jeff was thinking that AI can do just about everything, right? He explored different use cases for using AI with his engineering team, but unfortunately, found that the commonly-used LLM he tried was pretty good at general questions – but was occasionally wrong and made things up – while sounding very authoritative. He decided it was a bad combination.
Jeff had been feeling pressure from his employer to find ways to “work the AI magic.” Jeff had a problem: The day-to-day life in the trenches of his engineering team was hectic, and didn’t seem to present itself well for AI automation.
But Jeff started thinking about requirements, specifications, and standards. He started asking his team about how much time they spent, pouring through thousands of pages, looking for guidance and answers. The numbers were alarmingly high. In some cases, his engineers spent 1/5 of their time just trying to determine which requirements were applicable to the project they were working on. This translated into hundreds (or thousands) of hours.
Jeff thought about his organizational goals for 2024 – 2025 and one stood out: Reduce development costs. He knew there had to be a solution here. But what? Not only was his team responsible for developing the systems OEMs need but he and his team were supposed to stay current with technology trends, align to shifts in customer architectures, and adhere to changing requirements – all while pressured to reduce costs.
Enter Humaxa. When we encountered Jeff, he was stuck. He couldn’t see an easy way to get started with AI to solve his “real world” problems. We asked him not just about getting answers about what requirements and regulations were important for certain tasks, but where there were gaps – and who should be responsible for what. Those are the things generic LLMs, not trained on industry-specific data, will ever be able to do. We worked closely with Jeff to design workflows that made his work life infinitely easier.
Organization of requirements and “to-do” lists were hot topics for Jeff. We fashioned a way to pull recommended “to do” lists from a multitude of requirement documents. We built a way to pull all the requirements and classify them into type, who should be responsible for each one, and any pertinent recommendations – all through AI and all generated almost instantly.
Jeff’s eyes got big. He looked up and said “Do you realize that what your AI did in minutes would have taken my team days? Or months?
When it came time to calculate the cost savings, Jeff struggled – the numbers came in beyond expectations, but would people feel like AI was taking their jobs?
We at Humaxa pointed out how important “human-in-the-loop” is regarding the AI platform. The AI, specifically tailored for Jeff’s team, is also being trained by Jeff’s team. Only a human can do that. At the bottom of each answer are three buttons: “Yes that was helpful”, “No it was not helpful”, and “Add Correction.” Through the last option, the AI gets better and better through human verification and training.
Implementing a new, powerful technology can be daunting, but Jeff found that by working with a trusted partner, focused on a specific industry, dreams really can come true.
We’d love to work with you too!
Carolyn Peer
CEO/Co-founder, Humaxa
