Why the rubber is not meeting the road when it comes to AI and the Energy Industry ?

85% of 52 global chief executives from the oil and gas industry say they have either already adopted AI or are in the process of adoption in a KPMG report.

It is also predicted that the AI for oil and gas is expected to reach US$2.85 billion by 2022.

Clearly the energy executives are investing in AI to transform their business, the Technology executives are investing in AI to seize the big opportunity and of course with the current price of oil, there is genuinely a massive need for AI to deliver major efficiencies.

With all of these going in the right direction for AI in the Energy Industry, why is that there are not many AI examples at industrial scale.

Apparently we touch 70+ AI algorithms every single day when we look at Google Search or Amazon recommendations or Facebook face recognition or Uber pricing or Google Maps or Email Spam and so on.

We looked at the typical journey of AI projects to explore the key roadblocks and potential solutions that have emerged over the last 3 years.

Assuming a 100 Proof of Concepts got started on the back of exciting business use cases, only 20% of them reached the Pilot stage and a very small percentage got scaled across the organisation.

It would take a book to discuss all of the roadblocks and how they have all been overcome.

Three of these roadblocks still stand-tall.

1. At what prediction accuracy will you pivot from Pilot to Production?
2. What kind of guidance required for people to take action after AI has spoken?
3. How will you tag sensor data and align them all into the process flow diagram?

Netflix will not be putting anyone's life at risk if they decide to pivot to production when their Recommendation Accuracy is only 40%, however you need a robust system to help determine the accuracy percentage for your Oil Rig.

Will you shut down your Rig if AI predicted that the hammer is going to break?

We need a systematic way to explore the options, risks and a human approval even after the AI has spoken.

With 100,000 sensors screaming away data every second, it is important that we create an AI algorithm to collect and tag each of this field value pair with its metadata as well as organise the data in the right place in the process flow diagram. This automated first step in the data acquisition phase is extremely critical if we have to scale the idea of AI across 100s of plants or wells or refineries.