“Companies with flaky master data would be wise to cleaning that up first before embarking on DI.”, a comment on my last blog about Decision Intelligence said.

I surely agree, data is absolutely important when implementing technology. It already was when I was implementing ERP and APS systems 20+ years ago. And it still is. Garbage in is still garbage out indeed.  

Maybe more so these days when applying AI, Decision Intelligence, and high levels of automation.  The stakes are maybe a bit higher when we rely more and more on automation and machine intelligence.

It’s probably also a bit psychological. We feel we have less control of what the machine does. It is harder to intervene. Our fear fires up the amygdala, and activates our fight or flight response, which kept us alive for thousands of years. Groupthink makes fear spread amongst the leadership team. Fear for the unknown is hardwired and can be hard to overcome.

On top of that there is uncertainty. GenAI still hallucinates, and there are plenty of examples where AI shows significant bias, makes some weird predictions or conclusions and there are memories of a chatbot turning into a scary freak within a day! I think these examples are actually great because it shows the human will (almost) always need to be involved.

Given all this uncertainty, data sometimes seems to be turning into the ultimate excuse to maintain the status quo and do nothing. Whole leadership teams are throwing their hands up in the air saying in chorus; “We don’t have data quality; we can’t do this.”

When we think about AI and Decision Intelligence, I can’t say it any better than Lorien Pratt, co-inventor of Decision Intelligence;

“Data, technology and AI must take a back seat to diligent understanding of the decision that they support.”

 So, before you through your hands up in the air with your leadership team think about the following:

• You already make decisions with your current data: So why not make faster, more accurate, higher quality decisions using the same data for a starter?
• Your data decays: An IDC study revealed that 50% of respondents says data loses value within hours. Only automation to increase decision velocity can extract value out of this data.

• Data quality is never 100%: Data is like a living organism. It is always moving. It is never 100% complete, accurate and consistent.

• Some data is becoming probabilistic: supply chain parameters, like a lead time, are becoming probabilistic, so the field doesn’t necessarily need maintenance. Instead, a distribution function will continuously assess the best parameter to use.

• What data needs improved quality? Be specific. Does every bit of data need improvement? 90% of predictive capability sits in 10% of your data (Lorien Pratt). Focus on relevant data for your decision, the rest is noise.

• Think decision back, rather than data forward: work your way back from the decision, options, analytics needs and only then data requirements. Data quality needs are specific to decision & action needs. No need to clean all your data and dump it in your data lake when you’ll never use it.
• Data maintenance can be automated: for data that needs significant quality improvement, there are ways to automatically detect quality issues and update those.

Don’t panic about your data! Think decision first, data and AI later!

Picture source