What we learned implementing AI in manufacturing

We recently implemented an AI System for a manufacturing business.

Some of the lessons learned along the way.

• Data will have problems. Problems you hadn’t seen, problems that didn’t exist before, and problems that you thought you had fixed.
• AI is not magic; it’s where you use it. We spent considerable time observing and interacting with their team. By the beta phase, they were excited rather than nervous about AI replacing them.
• The input data could shift, making the output less relevant. Also, an accurate output is not necessarily a usable output.
• Start conservatively. We aimed for 15-20% improvement and hit 23-30%. That’s better than the other way around!
• Very important to focus on “how can it fail.” Everything else can be improved later.

PS “What happens during holidays?” Always a good question to ask!