There’s something about Software Crafters Cambridge - and it’s not just that I can never remember the order the words go in the name. They’re a quiet group of software professionals, but the talks and workshops are always interesting, and it quickly becomes apparent that there’s a lot of knowledge and experience in the room.
I’ve known Carl Farmer for several years. I helped him set up SyncIpswich back in the mid-teens. However, I hadn’t seen him for ages, so I was keen to come along and hear what he had to say. I wasn’t disappointed.
Carl works in AI at Microsoft as a Forward Deployed Engineer - sales, basically - so he’s seen it all. He took us expertly through how to measure the real ROI of AI. It turns out it’s the same way we measured it with human engineers; we just mustn’t forget to keep doing it that way. Carl then explored ways of working with AI, how to get the most out of prompts, and how to minimise costs. I came away with a lot of notes and useful tips.
The part that resonated most with me was the final section of his talk on quality and the human in the loop. It appears that many, or at least some, engineers are forgetting to engineer. They’re so excited about what AI can do that they’re forgetting to check what it did. In most cases, code still needs to be well crafted, it still needs automated tests, and it still needs to be reviewed just as thoroughly as a human engineer’s code would be.
There was a lot of discussion from a few people in the room throughout the talk, which added quite a lot of value. More than ninety minutes passed in a flash, and I realised I hadn’t been so engaged in a presentation for a very long time.
I’ve known Carl Farmer for several years. I helped him set up SyncIpswich back in the mid-teens. However, I hadn’t seen him for ages, so I was keen to come along and hear what he had to say. I wasn’t disappointed.
Carl works in AI at Microsoft as a Forward Deployed Engineer - sales, basically - so he’s seen it all. He took us expertly through how to measure the real ROI of AI. It turns out it’s the same way we measured it with human engineers; we just mustn’t forget to keep doing it that way. Carl then explored ways of working with AI, how to get the most out of prompts, and how to minimise costs. I came away with a lot of notes and useful tips.
The part that resonated most with me was the final section of his talk on quality and the human in the loop. It appears that many, or at least some, engineers are forgetting to engineer. They’re so excited about what AI can do that they’re forgetting to check what it did. In most cases, code still needs to be well crafted, it still needs automated tests, and it still needs to be reviewed just as thoroughly as a human engineer’s code would be.
There was a lot of discussion from a few people in the room throughout the talk, which added quite a lot of value. More than ninety minutes passed in a flash, and I realised I hadn’t been so engaged in a presentation for a very long time.

Comments
Post a Comment