09 Apr How to Successfully Introduce AI Into Oil and Gas
As the oil and gas industry slowly explores artificial intelligence (AI) solutions, it will have to confront critical management issues that hinder faster and more impactful adoption.
The Industry is Curious About AI
I’ve been invited (again) by COSIA (Canada’s Oil Sands Innovation Alliance), to participate in a panel discussion on the role of AI in oil and gas.
The first instance was in October of 2017, at COSIA’s annual associate member meeting, a gathering of the science-based institutions that work on common oil and gas issues in land, water and air. The first panelist was Cameron Schuler, who heads the Edmonton-based Alberta Machine Intelligence Institute who focused on the underlying science of AI and how it was evolving. The second panelist was Robin Auld, a consultant who was carrying out market research on AI readiness for oil and gas.
Being a practical sort, I decided to focus on how AI was already making a difference in oil and gas by sharing four specific use cases. My goal was to stimulate the attendees into thinking about where AI could be applied in their fields of work. You can find my presentation on Slideshare, but if your organisation blocks access to free learning, just send me a note and I’ll get you the material directly.
The four use cases that I set out are all in the public domain (that is, vendors sell their AI solutions in the market, or the media or the company has commented on the use case). The four include:
- the use of unmanned aerial vehicles (drones) to monitor oil and gas infrastructure from the air and to take action based on condition.
- the use of cameras and a massive image database to monitor facilities on the ground and to take independent action based on events.
- the use of very sophisticated math to predict plant operational and economic performance, in the case of tank farms and windmills, to improve operational outcomes.
- the use of a language interpretation engine to access huge libraries of unstructured content to support engineering decisions.
AI raises a number of management issues, which will be the subject of my remarks at the next panel discussion. I suspect these issues hinder the adoption of AI in the industry, and hence the interest in hearing what they may be and how to offset them.
Top AI Management Issues
In no particular order, here’s what I see as the key management issues that hinder the adoption of AI in oil and gas.
High hurdle rates for funding
Now that the industry has reset its cost structure to match its revenue model, returns to the industry have achieved the levels when the price of oil was at $100 per barrel (at least in Shell’s case). As a result, producing assets drive high returns, which tends to drive up the hurdle rates for all investments. This makes sense – anything that yields a return that is less than drilling another well should not be funded, following shareholder value theory. However, high hurdle rates crowd out innovation in new areas, like applying AI and other digital solutions to the industry. One super major has revealed that their hurdle rate for any digital investment must be $1b or more, in either improvements in cost or growth in revenues.
Offsetting economic performance metrics
A very significant pressure on CEOs, driven by Boards and shareholders, is to convert balance sheet assets (or reserves) into cash as prudently and as quickly as possible. Not only do these assets make no money sitting idle on the balance sheet, but the existing producing assets generate less value each year because of reservoir decline curves. This economic pressure gives drilling programs priority over all others. These performance metrics crowd out initiatives and investments that could also produce strong returns. A lack of understanding of AI specifically, and digital more generally, among management and Boards also blocks receptivity to AI.
Following the 2014 price collapse in oil and gas, and with uncertainty when (or if) prices will return to their previous highs, oil and gas companies embarked on their normal playbook to cope with commodity price downturns. Capital was slashed, price concessions were extracted from vendors, and employees were dismissed. Some 300,000 professionals have left the industry. There’s little appetite in oil and gas to add staff back given the uncertainty, and even less to add skills in areas that are technology centric rather than hard core oil and gas. Meanwhile the demand for AI knowhow has accelerated in many industries, leading to skills shortages in key areas related to AI and, importantly for oil and gas, the field of data science.
Data management practices
AI consumes a lot of data. In fact, the more data, the better the AI engines can perform. And not just the volume of data, but the accuracy and timeliness of that data. Oil and gas generally does not have the kind of robust data culture that we would see in other industries. There is usually no c-suite owner of data, and data isn’t viewed as a balance sheet item (it’s an expense). Field data is still horribly manual and paper based for many companies. The low prices demanded of service companies robs them of the capital necessary to improve data quality. Data silos inside companies prevents the kind of data sharing that the digital industry takes for granted. Automated data measurement devices have been viewed as too expensive and many producing assets lack the kinds of automation that other asset industries (utilities, aircraft, manufacturing) have in place. Early AI trials will stumble on data issues.
Risk averse culture
The culture of oil and gas places strong emphasis on safety, reliability, operational excellence and environmental sensitivity. Another term for this culture is risk averse. Oil and gas is distinctive in how slowly it embraces change, usually waiting for some courageous outfit to try something different and prove that it works, usually over a multi-year time horizon. Change must be nearly perfect in its implementation and effect, and tolerance for ambiguity is low. This risk aversion is applied to all change uniformly, not just those changes that demand it. There’s a clear preference for “products” that have little implementation uncertainty and very easy adoption. For the most part, AI is none of these yet.
Supply chain restrictions
The technology companies that make up the oil and gas supply chain have exploited the industry’s risk averse nature by convincing the industry of the merits of walled gardens of technology that are proprietary, not open source, with few interconnectivity options. Creative AI solutions that originate outside the industry, or beyond the normal supply chain, or need to interact with existing technologies, struggle to gain traction. One super major is using a defense contractor to design its next petroleum plant to achieve more openness in plant automation systems.
There’s been few if any fields that have not benefited from AI, but there is a clear risk that the early adopters will be forced to come to grips with the fact that AI is inherently better at some tasks (or many tasks) than humans. This makes AI a hard sell – no one wants to admit that a computer could run the assets much better than they can. One AI vendor has told me that the biggest hurdle they encounter is simply convincing some grizzly veteran with 30 years of experiencing managing an asset that AI can better his performance at all, let alone by 10% or more.
Getting Ready for AI
In my experience, the management challenges I’ve outlined are sufficient to stymie most efforts at introducing AI into oil and gas. Here’s some actions that forward-thinking companies will consider to move ahead.
Tilt the playing field
Since hurdle rates block innovation, change the hurdle rate for AI projects to encourage trials. Set aside specific budgets for AI projects, that cannot be allocated to other ventures. Force specific targets for AI, such as reduction in cost, or improvement in productivity, and hold managers accountable for delivery.
Concentrate where competitiveness gap is greatest
AI may be applied in many aspects of an oil and gas operation, but receptivity is probably highest where the competitive gap is the greatest. Focus on areas that manage or improve CO2 and methane emissions, or improve cost of operations.
Assign a data czar
With data so critical to AI success (and digital more broadly), set up an executive role responsible for digital and give them real clout to influence data sources, data management and digital investments.
Take a page from Woodside and create a small AI team focused on driving experiments and value creation using AI. Leverage the ecosystem of vendors and consultants to build that capability quickly. Equip the team with the tools to do the job.
Twist the culture
Highlight your early AI adopters and over-reward their efforts, regardless of results.
Oil and gas is not the first industry to struggle with adopting new ways of working. Just watch the movie Moneyball, a Hollywood flick about the 2002 Oakland A’s and their unconventional math-based approach to assembling a killer baseball team. AI is going to be the moneyball story for oil and gas.