# Is the Oil & Gas industry ready for AI?

## Big data is not always useful for AI solutions!

In this article I will share some tips for preparing wisely for the amazing algorithms AI has to offer. In the past decades Oil & Gas companies have been collecting enormous amounts of data. Unfortunately not all of the data that was collected has been useful. In some of the AI algorithms the goal is to learn from past data to make better decisions in the future, such as buying the right assets for production and using the right equipment. When I relate to learning, I relate to similar learning processes that people have. When facing a problem, we are trying to make sense from the information we already have about the problem to solve similar problems in the future.

#### Situations When Information Misleads Us

1. Some important data for understanding the problem is missing? (Important sensors are not installed, workers at the field don’t enter data that was required)
2. The data that was collected was actually wrong in a lot of the cases? (For example bad sensors, workers at the field that accidently entered manually wrong data)
3. The data that was collected was collected every time based on different terminology or based on different units? Isn’t it a mess?

I will show now a simple example that will help you understand what lack of good information can cause. Imagine that a new math operation marked as @ is shown to you with the following information.

• 2@2 = 4
• 1@1 = 1
• 2@1 = 2

Now you are asked to solve the next problem

• 4@2 = ?

Based on this data you can think about a couple of options

1. @ is actually * because 2*2 = 4, 1*1 = 1, 2*1 = 2, make sense right?
2. Wait, maybe it means the power of, i.e ^, because 2² = 4, 1¹ = 1, 2¹ = 2

Now if I add more information such as 3@2 = 9, we can be sure that @ is not * and it is more likely to be ^.

Imagine that you would only have the data that was first represented and you would have decided to use @ as * you would make a lot of wrong answers/decisions in the future. The same thing can happen in every kind of problem when significant data is missing.

1. Does your company collect significant data to solve the problem it aims to solve?
2. Think of a similar problem to the one with @ that could appear when the data collected was actually wrong, because of bad sensors or other problems. (I remind you that in the problem with @ the data was accurate but there wasn’t enough data to really understand the problem)

#### Tips for Collecting Quality Data

Now I will conclude with some significant tips that will help you collect better data for learning algorithms, but I’m sure you understand the point from what you read until now:

1. Collect data in an automated way rather than manually as it will probably be more accurate.
2. Make sure you collect the data under the same terminology and same measuring units. When you mean by depth of a well, is it only vertical depth or vertical and horizontal depth?
3. Make sure you are using good sensors. Are some of the sensors broken? Are the sensors collecting the data in-hole or on surface?
4. Are you collecting all the data that is needed to solve the problem. For example, are you collecting all that data that is necessary to evaluate the production of a well? Think of all the data that would help you solve the problem and make a better decision.

### That’s It For Now!

Data is very crucial to making AI, collecting better and cleaner data should always be one of your main focuses regrading AI. Collecting good data is most likely to be more important that the algorithm that is used to make better decisions. If the data is bad no algorithm could make it good. Remember more data doesn’t necessarily mean better data. Leave a message here for any information about our AI tools or any ideas.

Roi Shabshin is a co-founder of Unsist, a Calgary-based technology company that provides optimization products and services to the Oil & Gas industry through innovation and artificial intelligence expertise. Find out more on social media!