« February 2017 | Main | August 2017 »

June 30, 2017

Summoning the Demon

Recently, we published an article on the "AI Coworker". Thanks for all your views. We talked about the role that "Buddy" might play in operations to increase productivity in the field and decision support center analysts to make sense of all the data that is being generated, and get a jump on predictive decisions that could create more value out of your existing assets. It sounds like a great opportunity to integrate human experience with artificial intelligence, but there are a few challenges along the way that we will discuss in this article. Will artificial intelligence take over in a digital world putting humans to the side? Will "Buddy" be a valuable partner, or will AI be considered "our biggest existential threat" as the entrepreneur, Elon Musk, said as he compared the research under way equivalent to "summoning the demon."

Warnings about the potential impact of artificial intelligence have recently been discussed by prominent business and technology leaders. Some warn the technology will destroy jobs while others point to ways it will create new jobs. In December 2014, in an article for the BBC, Professor Stephen Hawking said, "the development of full artificial intelligence could spell the end of the human race." Hawking goes on to summarize that artificial intelligence "would take off on its own and re-design itself at an ever-increasing rate. Humans, who are limited by slow biological evolution, couldn't compete and would be superseded." 

Now we are not going to go as far as suggesting we are nearing the "end of the human race." We have a few more practical challenges to consider first. Any successful implementation of new technology will be planned and implemented in stages, as the human and existing business culture get more comfortable with the AI agent. It is a matter of trust, a viable business case and leadership direction and encouragement. Of course, we are assuming that the technology works.

First, does the AI agent understand what I am asking for? Putting a natural language speech interface on the AI system is a start. This requires adoption of a common language, or semantic ontology, that covers even very specific terms and acronyms (can Buddy get the buzzwords down as he reads the comment columns of the field report?).

Second, start with easier challenges as a first step. Start Buddy off with responsibility for specific pieces of equipment or common problems (can Buddy recognize the precursors for equipment maintenance requirements, performance or potential failure, for example?). Does Buddy know how to navigate through your data ecosystem and retrieve the right data for your query from the multiple data stores where the best answers are kept? Does Buddy know how to recognize alarm conditions from your SCADA system and begin the response and triage process, based on established best practices?

As the comfort and trust levels expand and AI agent (Buddy) performs actions to meet expectations, more of the capability of the AI can be implemented, including the closed loop automation of some field processes.

There are skeptics out there and their concerns are relevant. With the advent of the Big Data explosion and the digitization of almost everything, large amounts of data are being collected and fed to algorithms to make predictions. What would happen if a computer could make predictions so accurate that they could beat the decisions that humans are making? What if we begin to depend too much on those algorithms but their models go in the wrong direction, mistaking a strong statistical correlation with a weaker one causes an industrial accident? Or worse? We still need humans in the loop to keep things on the right track.

How far could AI really go? The world, digital or not, is a complex environment, little is standard and many processes are difficult to model. There is a lot of uncertainty and reactions to events that make up a day in the life. The futurist and inventor Ray Kurzweil thinks true, human-level A.I. will be here in less than two decades. While it may well be more than that, however, it remains a real possibility and not a question of if, but when.

Here's something to think about. Researchers at the Facebook Artificial Intelligence Research (FAIR) lab describe using machine learning to train their dialog agents (bots) to negotiate. It turns out bots are actually quite good at deal making. At one point, the researchers write, they had to tweak one of their models because the bot-to-bot conversation "led to divergence from human language as the agents developed their own language for negotiating." They had to change to a fixed, supervised model instead because the model that allowed two bots to have a conversation--and use machine learning to constantly iterate strategies for that conversation along the way--led to those bots communicating in their own non-human language. A bit scary don't you think? Anybody remember what Steven Hawking said?

Some worry about who will "own" or control emerging technology. Many of advanced startup companies in Silicon Valley are backed by massive investments from international sponsors, including the Chinese. Does this become a political issue around intellectual property? Or a national security issue? Will the solutions be open to all or controlled by a very few individuals, companies or governments? As data becomes recognized as a valued asset, will algorithms become the new competitive advantage over other physical assets?

This will be a journey marked more by change management acceptance by human operators for the AI coworker's capabilities, and not on the advances in the algorithms (they will proceed faster with the data science team than in the operations center). It will take several steps, over several years, and proceed at different paces, in different industries and companies, but the transformation has already started.

 I believe humans do not get cut out of the picture, but our ability to monitor and optimize complex operations, will only grow as Buddy becomes a trusted coworker. Our productivity will rise as Buddy helps us shift through more data, faster and analyze exceptions to predicted behavior. Maybe we really can do more with less with Buddy's help.

Even if we are not of the digital native generation, we still can learn new lessons. While "resistance may be futile" (using a Star Trek metaphor), collaboration might be productive. Digital technology should be a welcomed partner for the human, enabling us to become more productive and gain greater insights into our business challenges. We need not fight the demon like some digital luddite, but embrace the coming changes.

OK, Buddy let's give this a chance.

Jim Crompton is a thought leader for Noah Consulting, an Infosys Company, who is helping pioneer the relationships between complex Upstream processes and enterprises with automation to create competitive advantage.  His experience over numerous decades combined with the development capability of Infosys is working to ensure successful alignment of man and machine.

June 13, 2017

AI Coworker

With everyone talking about Big Data, Advanced Analytics and the Industrial Internet of Things, I have been trying to look beyond the hype, think about what's next and the adoption challenges all this brings. These new technologies are coming at a time when the oil and gas industry is trying to see if it is safe to lift their heads out of the bunker they have been in for the last several years due to low commodity prices. Some, especially those invested in unconventional plays in the Permian Basin, have already left the bunker and are charging ahead with lease acquisitions, mergers and new drilling programs. Oil production is going up, inventories are going up, more pipelines are getting built, oil field service rates are rising on higher demand, but their prices are staying down.

Others are more cautious and many are still trying to "fix" their asset portfolios, selling properties to repay debt, (many majors reducing downstream assets for a few examples), cutting capital budgets and projects from their annual plans, leaning on suppliers to keep prices low and some are even trimming staff a little more. Exploration is down and close attention to the operating budget is still essential.

But the technology advances are not waiting for commodity prices to rebound. The cry for digitization and new business models ring from consultants' speeches at every conference. Are we looking at these new technology advances in the right way? Are they just new and more capable tools or is there another way to think about adoption and transformation?

My thought process runs to a convergence rather than to the new capabilities themselves. I am more of a broad integrator then a deep specialist anyway. Here's my logic. What if we are able to integrate the new analytics capabilities, the new data lake capabilities, the cloud computing infrastructure, the new robotic process automation advances, with new interfaces enabled by mobile and voice technologies? Instead of buying a bunch of new tools could you think about Artificial Intelligence (AI) as a new coworker?

With the demand for increases in productivity (doing more with less), your new coworker, let's call him, Buddy, can interface with you by voice command and response (just like Apple's Siri®, Microsoft's Cortana®, Amazon's Echo®, IBM's Watson® and others). Unlike these others, Buddy knows the oil business (domain knowledge) and speaks the language that your job responsibilities require. Buddy knows how to navigate the diverse data environment, understands the business processes and authority chain unique to your company. Buddy is enabled by process automation and data discovery technology. At the appropriate stages, Buddy can answer queries with data from relevant sources and perform requested data processing, analysis and display steps. Buddy can even complete standard reports, escalate requests and recommendations to the proper level of authority.

Most importantly of all, Buddy saves time. Think about what else you could do while Buddy performs your time-consuming tasks of finding the right "trusted" data, performing data quality checks and updates, complying with standard work processes, processing data through sophisticated analytics tools and procedures. With all that happening behind the scenes, you are free to evaluate, to prioritize and just to think about what your data-driven world has for you.

You may think that this is a science fiction vision. You are right, but new projects like IBM Watson® working to understand stuck pipe, Amelia® from IPSoft, working on back office accounting tasks for Shell and Baker Hughes and others that haven't yet come out of the R&D labs, we may not be as far from this vision as you think. I recently worked with my team at Noah/Infosys developing an initial pilot of a "Buddy" like AI co-worker using Amazon's Alexa as the user interaction platform. It was both a great manifestation of the AI co-worker idea, but also served taught us several lessons on the complex challenges associated with deploying this type of technology. For example, in order for Buddy to know where to go get clean data, you need to have clean data, meaning many of the age-old data management challenges will still need to be addressed for AI co-workers to be implementable. Next step is to further leverage Infosys' Big Data and AI platform, Nia, to take the Buddy concept to the next level.

If all this new technology is going to truly transform our industry instead of just adding more tools to our already overwhelming collection of applications, we need to challenge our view of new technology. The use of automation has to take care of routine tasks, not just pile more data and more email requests on our desks (maybe they are text messages and tweets instead of emails these days). We already have more data than we can cope with and more is coming. We need this technology to find, sort, filter, process, model and simulate and then display the interesting stuff.

I am not suggesting that your company start a robotics resources group to complement your human resources group just yet. I am asking you to think about all this new technology in a different way. It may be science fiction right now and many early pilots will probably fail to some degree, but things are changing fast. Self-driving cars are an example of technology potentially changing the future of transportation. Drones are changing methods of visual inspection across industry (Cyberhawk) and even how countries fight wars with more uses being discovered each day. Drones are controlling themselves as remote operated vehicles (ROV) and autonomous underwater vehicles (AUV). Gamification is using Virtual Reality (VR) and Augmented Reality (AR) to turn training sessions into video "games" with scoring and even competitions.

There are new "smart technologies" transforming the oil field at an increasing rate of speed. Some of these technologies will help with the daily challenge of trying to find data to make better decisions and "do more with less".  Unlike the movie, "2001, A Space Odyssey", the AI Hal, might be a glimpse of how it might look, but I hope it isn't trying to get rid of you somewhere in the plot.

Ok, Buddy, let's get to work.


June 2, 2017

Battle for the Digital Core

While no one is popping corks on the champagne yet, things are looking a little better in the oil patch these days. Oil prices have recovered to a range between the mid $40s to the low $50s. Drilling and production costs in some basins have fallen to the point where $50 oil can mean a positive cash flow again. The Permian Basin looks healthy with lease prices in the best trends near $60K per acre, and increased drilling activity has added hundreds of drill rigs to the fleet over the past few months. OPEC has announced a production cut of over a million barrels of oil per day and crude oil inventories have fallen a little. A few major capital projects have been approved including:

·         Chevron's Tengiz Future Growth Project (Caspian Basin)

·         Statoil's, Johan Sverdrup (Norwegian North Sea)

·         BP's Mad Dog 2 (Gulf of Mexico)

·         Eni's Zohr (offshore Egypt)

This appears to be a sign we have reached the bottom of the commodity cycle and can start planning for better times.

I am not brave enough to try and predict future oil or natural gas prices or even global demand for fossil fuels, but I do want to talk about some of the interesting new developments in the digitization world that apply to our industry. I want to introduce a concept I call "the battle for the digital core." If the industry is preparing for new investments, investing in data as an enterprise asset and integration capabilities to help each employee become a more productive and better data analyst can have profound, long-term returns.

The challenge of the digital core is all about integration: the variety of data required; the dozens of technologies and vendors; and the people involved in a complex work process. In a recent SPE paper, "How Do We Accelerate Uptake and Fulfill the Value Potential of Intelligent Energy?", (SPE 181091 by H. Gilman, T. Lilleng, E. Nordveldt and T. Unneland), the authors described four types of integration that are needed to enable the digital oilfield as:

1)      transaction- oriented

2)      linking cross-disciplinary workflows

3)   collaboration between different locations (onshore and offshore and between similar assets in different basins)

4)      bringing together different time frames (matching history and real-time)

Who are the competitors in the battle for the digital core? What does winning look like? And most importantly, How are you going to manage the challenge of integration - the type we are talking about here? It is one thing to link one application to a specific data source with an API or a bit of ETL code, but the digital core is much more complex. Most programmers think functionality first and integration later, if at all. Commercial applications often create proprietary data models instead of trying to use industry or even company standards. Well understood, common definitions of critical information objects are often lacking, even between functions in the same company. Technology companies want to bring something new and cool to the market, therefore, avoid reuse of proven solutions or legacy investments. Upgrade churn and no backwards compatibility seem to be a cruel joke on technology users. Integration is an unnatural act in the digital oilfield, but it is a critical missing element. When the opportunity for collaboration comes, existing solutions often present more barriers than open doors. But I see that things are changing.

I see five different sets of technology suppliers trying to do a better job of integrating their own products and applications, but they are not stopping there. They all see the need and opportunity to be the first to provide the full asset lifecycle integration platform, or digital core, for the next generation of cross-functional, cross-asset, holistic enterprise solutions.  The platform will connect people, processes and technology, with the data foundation needed to become a data-driven, analytics-based organization. I will try to briefly describe what each of the five players (IT, OT or operations technology, ET or engineering technology, oilfield services vendors and back office ERP vendors) are trying to do.

First, let's take a look at IT. IT is the enterprise steward of shared computing and communications infrastructure and the database technology used across the enterprise. Their view of the digital core is a very technology-centric one. They have been responsible for structured databases, document management systems, email and messaging platforms, enterprise data warehouse and now the data lakes. They are the programmers that write the point-to-point applications links. The ones who have succeeded in building bridges between traditional technologies and the ones who have unintentionally created the complex spaghetti of links without a central design that reinforce the information silos we have today. IT is trying to leverage enterprise architecture design methods to belatedly develop an enterprise view of all digital assets and improve search and data discovery.

A new term is surfacing this describes this approach called the "data fabric." The collection of technologies that make up the data fabric enables the IT department to integrate, secure and govern various data sources through automation, simplification and self-service capabilities. The data fabric follows key work processes to create smoother information pathways for often used and high-value requirements to build a "get-my-data" portal for easier and a more intuitive access to data.

The need to make faster decisions requires that organizations incorporate real-time data into the process. Operations Technology (OT) includes process control, SCADA, historians and operational data stores, equipment health monitoring, predictive maintenance applications. For many years, this class of technology was the responsibility of electrical engineers and automation specialists and never crossed IT's path. But the proprietary technology has morphed to commercial IT technologies for lower costs, and the capability to bring the field data into the home office now exists, as does the ability to augment this data collection with mobile data capture methods. The process automation vendors see the digital core opportunity as do original equipment manufacturers (OEM), as the products they sell are getting more information-rich and services to optimize equipment performance brings a higher profit than just selling the equipment these days.

The surprise player in this space for me is Engineering Technology (ET). The facilities engineers have been diligently creating digital solutions for facilities design (3D CAD/CAM), construction and commissioning tasks often called the "digital plant." Vendors can talk about a "digital twin" of the actual physical facility and are proposing that this platform can be of value to operations and maintenance throughout the asset lifecycle. Here is where lessons learned from manufacturing can be leveraged by exploration and production. Document-centric work processes can be related to a dynamic 3D model of the topside processes and training can be enhanced by gaming technology and virtual (or augmented) reality visualization. They want to be part of the digital core as well.

Due to my geoscience background, I have a bias towards the familiar subsurface platforms offered by oilfield service software vendors. OFS solutions bring the subsurface characterization, reservoir simulation and drilling design and surveillance suite of tools, that are already widely adopted in the industry. A few tweaks to their existing platforms and they join the playing field in the digital core competition.

Finally, we can't count out the back-office transaction-oriented ERP platforms. This isn't an academic or geoscience or engineering world, it is a commercial world after all. Any asset manager has to bring in the information that financial transactions, procurement orders, regulatory permits, land contracts and HR records provide. From a financial perspective, all this activity is brought into the ERP system for final profit and loss reporting and review anyway, so the ERP platform has a role to play.

Designing and developing a digital core will not be an easy task (either from the technical elements, to the commercial adoption challenge). The industry not only has the task of looking at different vendors in each space but the potential of one, or several, platforms working together to become the digital core. This maybe a task that many companies are either too small, have limited capabilities or just want to leverage others, so that they can concentrate on their core business of producing oil and gas.

When you look under the technology cover to envision what the digital core might look like, I can see one of three scenarios playing out.

Facebook scenario: In this scenario, I can see one company dominating with a cloud computing infrastructure, offering the digital core integration capability as-a-service. You get assurance that your data is secure, you get a webpage portal (link) where you can discover your data and a workspace and tools to build your models. Off you go with only a monthly usage based charge to pay. Upgrades of software functionality are taken care of automatically, infrastructure additions are behind the curtains and not on your capital budget. You can focus on your business with the benefits of a holistic perspective of your asset performance and asset lifecycle. Sounds tempting, doesn't it?

Duplo scenario: It is possible that the digital core platform can be developed with a standards-based, inter-operability approach. Developing and adopting standards for data and integration will allow an operator to mix and match vendors and applications. They can assemble, rather than program, the parts they need and can host their version of the digital core inside their firewalls and still communicate with external stakeholders seamlessly. This sounds tempting as well, but the industry is going to have to put in some hard work in standards groups to make this viable.

Silo's win:  This scenario is the bad news. If it proves too difficult to agree to the inter-operability needed, open integration cannot be achieved - proprietary solutions will dominate, with too many vendors competing with differentiated functionality and no common core. Barriers to integration still stand, both from a commercial, cultural and technical perspective.  In this scenario, the digital core remains the recommendations of consultants (like me) on their PowerPoint decks, with little ability to help the industry.

So, what will the future bring? I wish I could predict, just as I wish I could predict the oil price in five years' time. I could make a fortune on the options market if my crystal ball was working well enough. I think the promise of the digital core is substantial and the efforts from the five technology markets tend to support that viewpoint. The Facebook scenario has many challenges and unless everyone agrees to buy this vision-as-a-service and let one company take the risks, a lot of collaboration will be required. Operators might want to pay the subscription fees only when the commercial digital core-as-a-service is ready enough for them and until that time, wait patiently on the sidelines. There are technology companies working hard to win their business. The Duplo scenario means we will all have to work together to bring this to fruition. The Silo scenario, we can keep doing what we have always done including complaining that we cannot find our data, create value from all the data we collect, and ultimately suffer the consequences of limited insight and lower productivity.