"This whole AI thing is so easy, we'll just do it ourselves"

I really enjoyed reading this blog post ("Beyond the AI MVP") over at technology news site The Stack from Lawrence Jones, an engineer at incident.io.

In the post, Lawrence sets out how 'deceptively simple' it look to deploy AI. Lawrence wanted a little system to sort through customer incidents and knocked out a working prototype which, 'looked amazing'.

He explains:

You build something in days that looks almost finished. The demo is incredible. Your team is excited. Then it meets reality, and everything falls apart. The system that looked so polished in demos starts making basic mistakes that seem obvious in hindsight.

It's a serious trap that a lot of organisations fall into, especially when their engineering team create something off the side of the desk.

This has historically been a challenge in many companies – the vendor walks in and presents their capabilities. Someone in IT says, "I could do that in a few days."

So you say goodbye to the vendor(s) and the IT guy gets to work. And when it comes to AI, it is really easy nowadays for the ultra-test-version to look amazing.

It's moving it to production quality – that's the science bit. Then it's trying to keep it there. Quickly that becomes a full-time, all encompassing job itself.

If you'd like to get a good overview of how things are in today's engineering world for many companies, check out Lawrence's post and his descriptions!

Nice work Lawrence and great initiative by The Stack!