Oil and Gas Gets Wise To The Power of Artificial Intelligence
The oil and gas industry is getting more and more curious about artificial intelligence. What is AI, and why all of a sudden all the interest?
Occupying the place that I do, which is at the intersection of Digital Industry Way and Oil and Gas Avenue, I get to watch the passing traffic. There’s been a curious uptick in the frequency of discussion about artificial intelligence among the oil and gas towers in Calgary.
What kind of interest? Well, the company that I work for is a magnet for enquires on all manner of hot topics, including AI, and not just in my favorite industry, but from across the industrial landscape. Just in the past month or so, I’ve fielded an RFP to find out the state of play of AI, received an invite to speak at a conference on the topic, and have met several vendors who position themselves in AI for oil and gas.
Once the volume of noise, email traffic and requests for help in my company reaches a certain pitch, we up and assign someone the job of concentrating our resources and attention on the topic, and this week, we created the role of AI leader.
Therefore, I can say that AI is officially a thing. And a thing of interest to business.
If it’s a thing in Calgary, then it’s also a thing in other oil and gas towns, like Houston, Perth, Aberdeen and Abu Dhabi.
What is artificial intelligence?
We humans, arrogant and self centered as we are, tend to believe that intelligence is confined to our species. We define intelligence as certain special skills such as the ability to communicate with language, to translate between languages, to reason, to perceive the world visually, and to make decisions.
Therefore, an intelligent species can take an independent action (decision) based on a stimuli or some data (sound or visual). If I see something in my path of travel, I need to make some quick decisions - do I go over, under, through or around the obstacle, or do I simply halt and wait for it to move. Quite a bit of processing goes into such a simple decision. Animals also have this kind of intelligence.
Going a step further, an intelligent being can reason (if it’s raining out, I should carry an umbrella). It can learn from experience (when it was raining and I had an umbrella, I stayed dry). It can detect and act on patterns (I can see dark clouds, which is usually a sign of rain, and therefore I should carry an umbrella). It can invent language (Umbrella is too many syllables so let’s shorten it to brolley).
Something that is artificially intelligent should therefore be able to do some or a significant portion of the things that a naturally intelligent species can do, but using computers to process rules about the world, sensors to detect the world, historic data to provide reference to prior situations, and the ability to take some action based on the outcomes of applying the data to the rules.
An autonomous car, therefore, embodies quite a bit of artificial intelligence. They can detect and avoid obstacles, predict the likely path of other moving objects, and change their behavior according to the conditions, like slowing down in the rain or snow. Many robots in the future will feature this kind of navigational intelligence, learning and adaptability.
AI in Tax
My company has developed an artificial tax advisor to help with tax issues (yes, it’s not oil and gas, but hear me out first).
Firms in the EU have complex tax rules with which they need to comply, and one of the most complex areas is in the determination of the applicability and amount of value added tax (VAT, or the goods and services tax or GST, for my Canadian readers), applied against goods being shipped across EU borders.
No one likes to pay tax, and so companies frequently take their case to European courts to adjudicate on whether items are tax exempt, or meet the criteria for a lower tax rate.
The problem is that there are now thousands of such tax rulings. Just finding and reading up on all the relevant rulings takes hundreds of hours. Taking a case to court is expensive, but getting it wrong can also be costly. Companies want to know in advance their chances of getting a favorable ruling based on the specifics of their situation.
My Dutch colleagues have painstakingly loaded all of the thousands of accumulated case rulings into an AI engine from IBM, called IBM Watson. This AI engine can listen to a tax lawyer ask a complex tax question, instantly search the thousands of cases to zero in on just those that are relevant to the question, and can even provide the probability that a case before the courts will be successful based on the history of prior cases.
A good tax lawyer is 70% accurate in identifying the relevant cases, given enough time. Watson? 90% accurate, in seconds.
One of the secrets to AI success is to find those instances where AI can augment natural intelligence to accelerate analysis or improve the quality of human decision making based on its ability to process huge volumes of data using a set of rules or heuristics.
AI in Oil and Gas
Conditions look to be ripe for AI in oil and gas. Much grey-haired experience has left the industry, taking with it their grasp of the implicit rules of how the world works, and creating a need to capture and codify the rules. New recruits to the industry lack the experience and need to learn the rules quickly. The environment has grown considerably more complex (more rules, more regulations), meaning there’s a lot more data to process. Courts and regulators are unforgiving when the industry messes up, driving the stakes higher. The abundance of oil and gas resources has expanded supply, depressing prices and narrowing the margin for error.
You would think that the industry would be anxiously piloting AI wherever it can.
But the problem with AI in oil and gas is not where to apply it, but where to apply it most profitably at low risk. There are simply too many great use cases to prosecute.
Here’s a few of my favorite.
Contracts and Agreements
The number and diversity of contracts and agreements in your average up stream oil and gas outfit will number in the thousands. They’re all different, with often unique clauses, triggers and riders. They’re hard to understand (they’re written by lawyers, after all), and they can be in force for a very long time, outlasting their authors and original stewards. Knowing which are approaching key milestones is nigh impossible. Just finding them all can be a challenge.
AI could cut this paper problem down to size in short order, just as it has for the tax rules and laws. Imagine being able to just ask an AI-enabled contract manager to quickly find out which contracts need attention and why, and recommended actions to management.
Science fiction? Not at all. This is a parallel to my tax example. Another use case is Swiss Re, who use AI to help write insurance contracts in the first instance by using IBM Watson to recommend which clauses to use under the circumstances. Land professionals could really step up their efficiency in getting access.
Drilling Execution
Figuring out where to drill is tailor made for AI assistance. Engineers spend upwards of 40% of their time assembling the data to set up a drilling program, according to estimates by Woodside in Australia. They need data from prior drilling campaigns, cost estimates, infrastructure locations, well logs, seismic data, geologic interpretation, and so on. Not just any data, but the right data, and highlighting the specifics that require special attention. AI makes short work of this problem.
JV Accounting and Compliance
Oil companies still receive thousands of tickets from field services companies for services rendered in the field. People have to look at them, figure out which well they apply to, assign the right account codes, and sometimes suspend them for investigation. Sometimes they're hand written, sometimes they’re neatly typed out on a spreadsheet and submitted as a PDF file. Compliance reporting for commodities like water and emissions generates their own tsunami of paper. Some oil and gas companies actually employ more accountants and water usage compliance teams than geologists.
AI could take all of this mess, and convert it quickly and accurately into useful data, by using language processing to convert and interpret the text, identify and extract the right data, add data that's missing based on experience and rules, feed that data into the right systems, and make decisions to accept or dispute the charges.
Double counting would disappear, as would whole departments of accounting resources.
Science fiction? Nope. This is how AI is used in health care to deal with medical charges coming into insurance companies for reimbursement, against very complex rules (Obamacare, anyone?)
There’s no doubt that AI is a thing, that the conditions for adoption are ripe and there are lots of great use cases that will help lower costs and improve productivity and safety in the industry.