2019 OilComm Conference and Exposition Houston

AccessIntel/OilComm conference hears from the Houston office of the FBI, from Invatare on AI/ML ‘not delivering as expected’ and from Datarobot on automated AI in oil and gas.

A word from the Houston office of the FBI…

James Morrison of the FBI’s Houston office cited former FBI chief James Comey as saying that the private sector is both a key player in cyber security and a likely victim, ‘the private sector possesses the knowledge, expertise and information to address cyber intrusions and cyber crime in general’. The problem is that, as ‘a survey’ has found, ‘60% of oils lack adequate cyber defense’ which leaves them open to exploits such as NightDragon, ShapeShift, the Havex ICS trojan, CrashOverride (caused the Ukrainian power outage) and the Trisis ICS malware. Other exploits target wireless systems such as Zigbee and LTE cellular networks. Morrison sees three ways forward: 1) blockchain-based systems that prevent data manipulation and fraud, 2) a ‘zero trust’ model of ‘adaptive security and visibility’ and 3) instant messaging to replace email which ‘will be obsolete by 2020, replaced by Slack’.

Invatare – AI/ML not delivering as expected

Trond Ellefsen, CEO of digital transformation specialist Invatare described digital transformation as both a ‘business risk and opportunity for oil and gas’. Landmark and Equinor started their ‘painful’ transformation journey early and today, both companies have reached an ‘impressive maturity point’ that will continue to accelerate both companies’ advances and distance them from their competitors. ‘We are experiencing a head-spinning and profound moment in time where everything is affected by everything else’. This is a ‘cross-industry self-fueling process’ which will, over the next five years ‘expand, accelerate and create new ripples in the fabric of the world we live in’.

Ellefsen provided a digital transformation status report for year-end 2019. Despite the promise, current investments in digital technologies, AI, machine learning etc. ‘do not seem to be delivering as expected’. Projects are not gaining traction, not generating the expected ROI and not changing behavior. The silos are not being broken down, the expected better answers are not coming and speed to delivery has not changed significantly. A 2019 Harvard Business Review study, found that of a ‘staggering’ $ 1.3 trillion spend on digital transformation journeys in 2018 across industries, close to $850 billion ‘went to waste’ and ‘80% of all digital projects are considered a failure’.

In an analysis that echoed the lead story in this issue* Ellefsen puts these failures down to the application of ‘small adjustments on top of a legacy architecture that was never meant for digital hyper connections and massive interrogation of connected historical and real-time data’. In order to succeed the oil and gas industry need to take ‘deliberate, differentiated and foundational approaches to its digital effort, incrementalism is no longer adequate’. Industry is in a deep transition which for the past few years has been driven by consultancies with more focus on their own revenue growth agenda than on better industry solutions.

What is needed is a cross functional integrated open platform. Ellefsen sees a ‘need for common concepts and cross the industry efforts’ to solve the pressing issues the industry is facing. In this context Ellefsen cited Halliburton’s Open Earth Community an example of ‘a future open architecture where data can be interrogated across functions and systems’. Cross company and community-like collaborative for modules like blockchain, security and other efforts that should not be handled by any one company alone. The OEC is an open industry collaborative concept available for ‘effective development of digital muscles’.

* On McKinsey’s advice for oil and gas CIOs.

Datarobot - Automated AI in oil and gas

Rajiv Shah’s (Datarobot) starting point was another digital fail, the fact that, as Rexer Analytics has found, ‘only 13% of data science projects reach production’ and worse, even fewer generate real business value. There are many reasons for this. Many AI projects are poorly defined from the start, with unclear goals and no measure of current results to benchmark against. The reliance on data scientists makes for long model build times, long review cycles and the creation of model documentation for review. Currently, model building involves tedious, manual work that requires deep experience of data science and coding. This leads to shortcuts, especially with junior data scientists! Finally, most data science projects fail when the models are integrated with business processes. DataRobot offers an alternative approach, automating the model building, validation, and deployment process with embedded data science best practices and guardrails incorporated directly into the platform. On example is geological facies classification from well logs. In the DataRobot environment, well logs are compared with labeled core samples to set-up a supervised machine learning model. The model is then generalized to predict facies types from log data alone. More on AI in geoscience from the DataRobot blog.

Next year’s OilComm Conference will be held from October 14-15, 2020 in Houston.

Click here to comment on this article

Click here to view this article in context on a desktop

© Oil IT Journal - all rights reserved.